review core i3 takes on athlon ii

Core i3 takes on Athlon II

Coming up with a way to characterize a major comparison of desktop processors like this one isn’t always easy. Since our initial review of Intel’s Clarkdale processors, the desktop CPU market has shifted in a number of ways both big and small. For one, Clarkdale CPUs have proliferated everywhere, and we’ve gotten our hands on one model, the Core i3-530, that promises to be a much better value than the relatively high-end Core i5-661 that we first reviewed. To counter, AMD has introduced five new value-oriented CPUs, ranging from two cores to four, including the Athlon II X4 635, a potent value quad-core priced directly opposite the Core i3-530.

Naturally, we’ve tested these two direct competitors against one another, considered their power consumption, and plumbed the depths of their overclocking potential. But we’ve also had Damage Labs churning away around the clock to expand our CPU results substantially. We now have test results for processors priced as low as $74 and as much as a grand. Not only that, but we’ve looked back in time by testing a pair of older CPUs—including the always interesting Pentium 4 670 at 3.8GHz—to give potential upgraders a sense of what they might have to gain. Join us as we navigate a sea of test results and consider the best values available in today’s desktop processors.

The Clarkdale continuum
At a price of just $113, the Core i3-530 is much more reasonably priced for a dual-core processor than the Core i5-661 we considered in our first Clarkdale review. Yet because it’s a Clarkdale processor, the Core i3-530 has inherited much of the goodness from its elder siblings in the Core i7 line, including the ability to track and execute two threads per processor core (known as Hyper-Threading in Intel marketing parlance), an integrated DDR3 memory controller, and a CPU microarchitecture that does an awful lot of work in every clock cycle. In truth, Clarkdale processors are rather unique, because they’re really two chips in one package: a 32-nm dual-core Westmere processor and a second, 45-nm chip that houses a memory controller, an integrated graphics processor (IGP), and PCI Express logic.

Pentium E6500, Core i3-530, and Core i5-661

Clarkdale CPUs will plug into LGA1155 or LGA1156-style sockets, but you’ll need an LGA1155 motherboard to take advantage of the integrated graphics. Here’s a look at how the i3-530 fits into the broader Clarkdale lineup:

Model Cores Threads Base core
clock speed
clock speed
TDP Price
2 2 2.8 GHz 3MB 533 MHz 73W $87
Core i3-530 2 4 2.93 GHz 4MB 733 MHz 73W $113
Core i3-540 2 4 3.06 GHz 4MB 733 MHz 73W $133
Core i5-650 2 4 3.20 GHz 3.46 GHz 4MB 733 MHz 73W $176
Core i5-660 2 4 3.33 GHz 3.60 GHz 4MB 733 MHz 73W $196
Core i5-661 2 4 3.33 GHz 3.60 GHz 4MB 900 MHz 87W $196
Core i5-670 2 4 3.46 GHz 3.73 GHz 4MB 733 MHz 73W $284

One perk Intel has stripped out of the relatively inexpensive i3-530 is the Turbo Boost feature that raises clock speeds opportunistically when thermal headroom is available. Fortunately, the i3-530’s base frequency of 2.93GHz is pretty respectable all by itself. The i3-530 has the same 73W thermal design power rating as the rest of the lineup, except for the weirdo i5-661, whose higher-clocked IGP contributes to its 87W TDP. We’ve tested both the i3-530 and the i5-661, and through the magic of underclocking, we have simulated the Core i3-540, as well.

Of course, Intel’s product lineup extends well beyond Clarkdale, to the Core i5-700- and Core i7-800-series quad-core Lynnfield processors and into the beefy Core i7-900-series CPUs with triple-channel memory controllers.

Model Cores Threads Base core
clock speed
clock speed
TDP Price
Core i5-750 4 4 2.66 GHz 3.20 GHz 8MB 2 95W $196
Core i7-860 4 8 2.80 GHz 3.46 GHz 8MB 2 95W $284
Core i7-870 4 8 2.93 GHz 3.60 GHz 8MB 2 95W $562
Core i7-920 4 8 2.66 GHz 2.93 GHz 8MB 3 130W $284
Core i7-960 4 8 3.20 GHz 3.46 GHz 8MB 3 130W $562
Core i7-975
4 8 3.33 GHz 3.60 GHz 8MB 3 130W $999

We’ve included representatives from nearly every rung of the ladder, including one relative newcomer, the Core i7-960. At $562, the Core i7-960 costs just over half what the thousand-dollar Core i7-975 Extreme does, yet the difference between them is only 133MHz. (Well, the Core i7-975 Extreme earns its extremeness by offering an unlocked upper multiplier for the overclocking crowd, too. But you have to ask yourself: how much is an unlocked multiplier worth?) The Core i7-960 replaces the i7-950 at the same price point, a reshuffling no doubt prompted by the introduction of the very potent Core i7-870 at the same price on another socket.

A five-way refresh from AMD
Not long after the debut of the Clarkdale lineup, AMD conducted a freshening up of nearly its entire desktop CPU range. This move was unusual in a couple of ways: because it involved the introduction of five new CPU models at once, and because each one of them represents only a 100MHz clock speed increase over the prior incumbent at the same price point. The update was broad but incremental, serving as a bit of a price cut and a minor performance boost.

Top row: Athlon II X2 255, Athlon II X3 440, Athlon II X4 635.

Bottom row: Phenom II X2 555, Phenom II X4 910e.

The case is not without some singular features, though, as Sherlock Holmes would say. In fact, this calls for us to bust out our table of the current processor types, like so:

Code name Key
Cores Threads Last-level
cache size
Penryn Core 2 Duo 2 2 6 MB 45 410 107
Bloomfield Core i7 4 8 8 MB 45 731 263
Lynnfield Core i5, i7 4 8 8 MB 45 774 296
Westmere Core i3, i5 2 4 4 MB 32 383 81
Deneb Phenom II 4 4 6 MB 45 758 258
Propus/Rana Athlon II
4 4 512 KB x 4 45 300 169
Regor Athlon II X2 2 2 1 MB x 2 45 234 118

Athlon II X2 processors are based on the chip code-named Regor, a relative newcomer to AMD’s lineup. Regor features two Phenom II-class processor cores, each with 1MB of L2 cache. Like other Athlon IIs, it has no L3 cache. Regor may not impress the ladies with its neck-snapping acceleration, but it’s a relatively small chip that ought to draw power rather modestly and be cheap to produce. The Athlon II X2 255, for instance, has a healthy 3.1GHz clock speed, a 65W TDP, and a price tag of just 74 bucks—not a bad combo, all things considered.

We’ve picked up another code name in our table along the way, too: “Rana,” used to denote Athlon II X3 chips based on the Propus silicon that powers the value quad-core Athlon II X4 chips. Why we need another code name for the same silicon with a core disabled is beyond me, but I think the answer has to do with the fact that additional code names cost nothing. And marketing people need something to do with their time. The logic seems to be: “Yeah, I used to be called Bart, before I lost my leg in ‘nam and changed my name to Dexter.”

Regardless of the code-name chicanery, AMD’s value-oriented quad- and triple-core parts are a pretty savvy response to the excellent Core i3. AMD can’t offer you Clarkdale-class computing power in two cores, but they can give you more cores at the same price—at the cost of some additional power consumption and heat production. The first five lines of the table below detail the new models AMD recently introduced.

Model Cores Base core
clock speed
cache size
TDP Price
Athlon II X2
2 3.1GHz 2 x 1 MB L2 65W $74
Athlon II X3
3 3.0GHz 3 x 512 KB
95W $84
Athlon II X4
4 2.9GHz 4 x 512 KB
95W $119
Phenom II X2
2 3.2GHz 6 MB L3 Y 80W $99
Phenom II X4
4 2.6GHz 6 MB L3 65W $169
Phenom II X4
4 3.4GHz 6 MB L3 Y 125W $195

We’ve tested four of the five new parts, including all of the Athlon II chips. The Phenom II X4 910e is a low-power version of the Phenom II with a 65W TDP. We haven’t had time to run it through our full benchmark suite, but we have measured its power efficiency.

The one new model we’ve neglected, simply due to lack of time, is the Phenom II X2 555, a dual-core variant of the Phenom II that’s also a Black Edition chip with an unlocked multiplier. We happen to think the Athlon II X3 and X4 processors are potentially more compelling, but if you want to get a feel for how the X2 555 might perform, we have a full set of results for its younger sibling, the X2 550. If you squint hard enough when reading our graphs, you won’t be able to see the effect the extra 100MHz might have in the X2 555, anyhow.

I’ve included the Phenom II X4 965 in the table above because it represents the absolute top end of AMD’s desktop CPU line—at only $195. AMD can’t really ask any more than that given Intel’s current performance dominance, and the situation has led to a very compressed product stack. AMD does offer a host of Athlon II and Phenom II processors at varying price points and clock frequencies, but the differences between them are fairly small. One thing you won’t find in the new lineup: a Phenom II X3 processor. The triple-core options now appear to be confined to the Athlon II line, which probably makes sense under the circumstances.

Kicking it old school with LGA775
In response to CPU round-ups like this one, we often get requests for the inclusion of older processors, so folks can have a better sense of how an upgrade might serve them. Happily, this time around, we were able to test a couple of older processors as points of reference.

The newer of the two is the Core 2 Quad Q6600, which debuted just over three years ago at a very healthy price of $851. A 2.4GHz quad-core part based on 65-nm Conroe/Kentsfield silicon, the Q6600 became an enduring enthusiast favorite as its price dropped over time. I expect these CPUs are still at work in the systems of quite a few TR readers to this day. What’s more, the Q6600’s showing in single- and dual-threaded applications should essentially match that of the Core 2 Duo E6600, another popular pick from the same period.

If that’s not old school enough for you, how about some Prescott action? The Pentium 4 670 is a single-core, 90-nanometer CPU clocked at a heady 3.8GHz. The P4 670 first hit the market nearly five years ago and was among the fastest desktop processors Intel offered at the time. Of course, that was a rather dark time for Intel, because AMD had an iron grip on the performance lead—especially in games. Still, the P4 670 was essentially state of the art, with reasonably competitive performance overall. In fact, it’s pretty much as far back as we can reach in Intel’s product stable while maintaining compatibility with our 64-bit operating system and applications. The applications in today’s benchmark suite are much more broadly multithreaded, too, which could allow the P4 to take better advantage of its Hyper-Threading capability than it could back in the day.

Then again, I wouldn’t get too worked up about the prospects for a dual-threaded, 3.8GHz CPU from five years ago if I were you.

We’ve used older Intel processors for comparison not to rub Intel’s nose in a troubled period from its past, but because of the incredible track record of socket compatibility the firm has amassed during LGA775’s run. Older motherboards weren’t always capable of supporting newer CPUs, but we were able to drop both the Q6600 and the P4 670 in our new Asus G43 motherboard, boot up, and go to town. That enabled us to include these CPUs without the need to equip an additional test system.

Key contests to watch
Since we’ve tested such a broad range of processors, let me point out a few match-ups between Intel and AMD that are worth watching. The main one, of course, is our headliner, the contest between the Core i3-530 at $113 and the Athlon II X4 635 at $119. You know the outlines of that one. Pay careful attention to the contrasts in performance, power draw, and overclocking potential between these two processors, because I have a feeling this could be a close one, all told.

We don’t often test CPUs as cheap as the Athlon II X2 255, which rings up at just 74 bucks. But we have this time, and we’ve also tested its direct rival from Intel at the same price, the Pentium E6500. The E6500 is based on an older Penryn (Core 2 Duo) chip running at 2.93GHz on a 1066MHz bus with just 2MB of L2 cache. Those specs ought to put it very close to the X2 255, which ticks away at 3.1GHz and has a total of 2MB of L2 cache, as well. Penryn’s per-clock performance has generally been a little better than recent AMD processors, so the outcome is in no way assured. Both processors share the same 65W TDP rating.

I wish we had a Pentium G6950 to compare to the Athlon II X3 440, because those two chips are direct competitors. Perhaps we can snag one for testing next time around, but we at least have full results for the X3 440 now.

Test notes
We’ve mentioned that we underclocked the Core i5-661 to 2.8GHz in order to simulate the Core i3-540. Although we did change the core clock to the proper speed, the processor’s uncore clock remained at the i5-661’s stock frequency. We believe shipping Core i3-540 processors have a 2.13GHz uncore clock, while the i5-661 has a 2.4GHz uncore clock, so our simulated processor may perform slightly better than the real item due to a higher L3 cache speed. The differences are likely to be very minor, based on our experience with Lynnfield parts—the L3 cache is incredibly fast, regardless—but we thought you should know about that possibility.

Additionally, our Core i7-960 is an underclocked Core i7-975 Extreme, but in that case, we’re fairly certain all of the clocks match what they should, since Bloomfield gives us a little more control over such things. In order to run the Core i7-960’s memory at 1333MHz, we raised its uncore clock to 2.66GHz. That comes with the territory, and I expect many Core i7-960 owners have done the same.

As is our custom, we’ve omitted the simulated processor speed grades from our power consumption testing.

After consulting with our readers, we’ve decided to enable Windows’ “Balanced” power profile for the bulk of our desktop processor tests, which means power-saving features like SpeedStep and Cool’n’Quiet are operating. (In the past, we only enabled these features for power consumption testing.) Our spot checks demonstrated to us that, typically, there’s no performance penalty for enabling these features on today’s CPUs. If there is a real-world penalty to enabling these features, well, we think that’s worthy of inclusion in our measurements, since the vast majority of desktop processors these days will spend their lives with these features enabled. We did disable these power management features to measure cache latencies, but otherwise, it was unnecessary to do so.

Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and we reported the median of the scores produced.

Our test systems were configured like so:

Processor Athlon II X2 255 3.1GHz

Athlon II X3 440 3.0GHz

Athlon II X4 630 2.8GHz

Athlon II X4 635 2.9GHz

Phenom II X2 550 3.1GHz
Phenom II X4 910e 2.6GHz

Phenom II X4 965 3.4GHz

Pentium E6500 2.93GHz

2 Duo E7600 3.06GHz

Core 2 Quad Q6600 2.4GHz
4 670 3.8GHz

2 Duo E8600 3.33GHz
Core 2 Quad Q9400 2.66GHz
Motherboard Gigabyte
P5G43T-M Pro
P5G43T-M Pro
P5G43T-M Pro
North bridge 785GX G43
South bridge SB750 ICH10R ICH10R ICH10R
Memory size 4GB
(2 DIMMs)
(2 DIMMs)
(2 DIMMs)
(2 DIMMs)








8-8-8-20 2T 7-7-7-20 2T 7-7-7-20 2T 8-8-8-20 2T

Rapid Storage Technology

Rapid Storage Technology

Rapid Storage Technology
Audio Integrated

SB750/ALC889A with Realtek drivers

ICH10R/ ALC887 with
Realtek drivers

ICH10R/ALC887 with Realtek drivers

with Realtek drivers

Processor Core
i5-750 2.66GHz

Core i7-870 2.93GHz
i3-530 2.93GHz

i3-540 3.06GHz

Core i5-661 3.33GHz
i7-920 2.66GHz
i7-960 3.2GHz

Core i7-975 Extreme 3.33GHz
Motherboard Gigabyte
North bridge P55
South bridge ICH10R ICH10R
Memory size 4GB
(2 DIMMs)
(2 DIMMs)
(3 DIMMs)
(3 DIMMs)
Memory type Corsair








8-8-8-20 2T 8-8-8-20 2T 7-7-7-20 2T 8-8-8-20 2T

Rapid Storage Technology

Rapid Storage Technology

Rapid Storage Technology

Rapid Storage Technology
Audio Integrated

P55 PCH/ALC889 with Realtek drivers

H57 PCH/ALC889 with Realtek drivers

ICH10R/ALC888 with Realtek drivers

ICH10R/ALC889 with Realtek drivers

They all shared the following common elements:

Hard drive WD
ENGTX260 TOP SP216 (GeForce GTX 260) with ForceWare 195.62 drivers
OS Windows
7 Ultimate x64 Edition RTM
August 2009 update
Power & Cooling Silencer 610 Watt

I’d like to thank Asus, Corsair, Gigabyte, OCZ, and WD for helping to outfit our test rigs with some of the finest hardware available. Thanks to Intel and AMD for providing the processors, as well, of course.

The test systems’ Windows desktops were set at 1600×1200 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled.

We used the following versions of our test applications:

The tests and methods we employ are usually publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Power consumption and efficiency
We have pages and pages of performance data to show you, but let’s start with our power consumption tests, since that’s a very important characteristic of a CPU these days and is especially relevant to our key match-ups.

For these tests, we used an Extech 380803 power meter to capture power use over a span of time. The meter reads power use at the wall socket, so it incorporates power use from the entire system—the CPU, motherboard, memory, graphics solution, hard drives, and anything else plugged into the power supply unit. (We plugged the computer monitor into a separate outlet.) We measured how each of our test systems used power across a set time period, during which time we ran Cinebench’s multithreaded rendering test.

We’ll start with the show-your-work stuff, plots of the raw power consumption readings. We’ve broken things down by socket type in order to keep them manageable. Please note that, because our Asus H57 motherboard tends to draw more power than we’d like, we’ve tested power consumption for the Core i5-530 and the Core i5-661 on our P55 mobo, instead.

You’ll notice that the Pentium 4 670 couldn’t finish rendering the scene before our test period ended. That presented us with a real problem. In fact, we had to extend the P4’s test period to nearly twice the usual length in order to capture the full scene render.

We can slice up these raw data in various ways in order to better understand them. We’ll start with a look at idle power, taken from the trailing edge of our test period, after all CPUs have completed the render.

System power draw at idle is still very much influenced by how well the CPU is able to shut down parts of itself when dropping into lower power states, as these numbers attest. The latest Intel processors, both the Clarkdales and Lynnfields, come out looking very good. The newer LGA775 CPUs have nice, low power draw at idle, as well. The results for the Q6600 and the Pentium 4 670 tell the story of considerable progress for Intel over time, even in the same socket type.

Interestingly, our quad-core Phenom II-based systems just match our Q6600-based one, suggesting AMD is a couple of generations behind in terms of power reductions at idle. Fortunately, the Propus- and Regor-based Athlon IIs move the needle a bit. Our Athlon II X2 255-based test rig only pulls 7W more than its Pentium E6500-based rival.

Next, we can look at peak power draw by taking an average from the ten-second span from 15 to 25 seconds into our test period, when the processors were rendering.

Intel has a more pronounced lead when it comes to peak power draw. Consider the Pentium E6500 versus the Athlon II X2 255: the Athlon II-based system draws 28W more under load. Our Core i3-530 system pulls 50W less than our Athlon II X4 635-based one. The gap between the Core i5-750 and the Phenom II X4 965 systems is 40W, also in Intel’s favor. Only the Bloomfield Core i7 processors, with their extra memory channel, draw more power than the X4 965. AMD is capable of making CPUs that require fewer watts, as the results for the Phenom II X4 910e attest. Too bad those low-power Phenom IIs aren’t the norm.

We can highlight power efficiency by looking at total energy use over our time span. This method takes into account power use both during the render and during the idle time. We can express the result in terms of watt-seconds, also known as joules. (In this case, to keep things manageable, we’re using kilojoules.)

Notice that the Pentium 4 670 isn’t included above, since its total test period was much longer.

We can pinpoint efficiency more effectively by considering the amount of energy used for the task. Since the different systems completed the render at different speeds, we’ve isolated the render period for each system. We’ve then computed the amount of energy used by each system to render the scene. This method should account for both power use and, to some degree, performance, because shorter render times may lead to less energy consumption.

Let’s first pause to collect our breath and wonder at the amount of progress we’ve seen since on this front since the Pentium 4 days. The P4 670 system requires over seven and a half times the energy that our most efficient contender, the Core i7-870 system, does to render this scene. On the exact same motherboard, in the same socket, the P4 670 uses 6.2 times the power Core 2 Quad Q9400 does to complete the same work. The move from one core to four has no doubt played a big part in this progress, but obviously, advances of many types have produced these gains.

These efficiency numbers tell us Intel’s processors generally have solid performance to go with their relatively low power consumption. Notably, the Bloomfield Core i7-975 and i7-920 have moved into the upper ranks, because they’re able to finish the job more than quickly enough to compensate for their relatively high peak power draw. That same dynamic propels the Phenom IIs past the Athlon IIs among AMD’s offerings.

Memory subsystem performance
Now that we’ve considered power efficiency, we’ll move on to our performance results, beginning with some synthetic tests of the CPUs’ memory subsystems. These results don’t track directly with real-world performance, but they do give us some insights into the CPU and system architectures involved. For this first test, the graph is pretty crowded. I’ve tried to be selective, generally only choosing one representative from each architecture.

This test is multithreaded, so more cores—with associated L1 and L2 caches—can lead to higher throughput. If you follow one of the individual lines, say the Core i7-975 Extreme, you can see how bandwidth declines from the L1 caches to the L2, and from the L2 into the L3. I’d like to see finer-grained results, so we could discern between CPUs with, say, 2MB of L2 cache and those with more. The results we do have are somewhat enlightening, but don’t really show us all we’d like to see.

One result that’s easily discernible: the Pentium 4’s cache bandwidth isn’t exactly torrential. Then again, with only one core, the deck is stacked against it. Notice that the Core 2 Duo E8600’s L2 bandwidth, at points from 128KB to 1MB, is about twice the P4 670’s.

This graph becomes almost impossible to read once we get to the larger block sizes, where we’re really measuring main memory bandwidth. Stream is a better test of that particular attribute, though.

These results are separated neatly by CPU type. The triple-channel Bloomfield Core i7s are fastest, followed by the Lynnfields, Clarkdales, and Socket AM3 processors. The LGA775 CPUs, with their front-side bus bottleneck, are last—and even they are distinguished pretty cleanly according to bus frequency. Obviously, the move to integrated memory controllers has improved system bandwidth considerably: the Core i7-975 outperforms the Q6600 by a factor of four—and the Pentium 4 670 by, roughly, a factor of five.

Integrated memory controllers have also lowered memory access latencies substantially, but the Clarkdale processors are an exception. Their dual-chip-per-package arrangement, with the memory controller on a separate chip from the processor cores, apparently contributes quite a bit of additional latency. In essentially the same socket, the Core i7-870 needs about half the time to get to memory that the Core i3-530 does. Everything else with an integrated memory controller is much quicker.

This is my favorite game in a long, long time, so I had to use in it our latest CPU test suite. Borderlands is based on Unreal Engine technology and includes built-in speed test, which we used here. We tested with the game set to its highest quality settings at a range of resolutions. The results from the lowest resolutions will highlight the separation between the CPUs best, so I’d pay the most attention to them. The higher resolution results demonstrate what happens when the GeForce GTX 260 graphics card begins to restrict frame rates.

The GPU bottleneck serves to limit the disparities between the CPUs somewhat as the display resolution increases, but in this case, its impact isn’t especially dramatic. The finishing order changes little from one resolution to the next, and the frame rates hardly change at all in the bottom half of the pack. If you’re primarily concerned with relative CPU performance, you’ll want to focus on our lowest-resolution results.

Those show us that the Core i3-530 is a little faster in this game than the Athlon II X4 635, although the X4 635’s frame rates still average in the 50’s, which should be plenty fast. I wouldn’t pay too much attention to the minimum frame rates in this particular case; they don’t seem to be primarily determined by CPU speed, and they all fall within the same basic range.

I’ve been saying for ages that most of today’s games will run well on just about any desktop processor. That’s true in part because most games are co-developed for game consoles, whose CPUs are weaker than, well, me trying to climb the rope in gym class. Still, in the Pentium 4 670, we have finally found a processor that can’t run a contemporary game well. The Q6600 has no trouble, though.

I have two more things to note about Borderlands. One, this game doesn’t seem to benefit from having more than two cores available. The Core 2 Duo E8600 outperforms the Q9400, for instance, and all of the Athlon II chips hit similar frame rates. Two, the Phenom II processors appear to get a nice boost out of their L3 caches. The Athlon II chips suffer by comparison.

DiRT 2
This excellent new racer packs a nicely scriptable performance test. We tested at the game’s “high” quality presets with 4X antialiasing.

I’m growing tired of watching the GPU bottleneck take over, but I know some of you want to see all of these results. Generally speaking, this game makes good use of at least four cores (or threads), and the non-Clarkdale dual-cores cluster at the bottom of the charts. The Athlon II X4 and Core i3 processors maintain a pretty close parity here, although the Athlon IIs have a slight advantage.

Modern Warfare 2
With Modern Warfare 2, we used FRAPS to record frame rates over the course of a 60-second gameplay session. We conducted this gameplay session five times on each CPU and have reported the median score from each processor. We’ve also graphed the frame rates from a single, representative session for each. We tested this game at a relatively low 1024×768 resolution, with no AA, but otherwise using the highest in-game visual quality settings.

Practically speaking, any of these processors—except the Pentium 4 670—will serve just fine for Modern Warfare 2. Once more, the frame rate gap between the Core i3 and Athlon II X4 processors is negligible.

When minimum frame rates don’t drop below 50 FPS, you’re pretty much golden. If you have a Pentium 4 still, you’re kinda not.

Left 4 Dead 2
We tested Left 4 Dead 2 by playing back a custom demo using the game’s timedemo function. Again, we had all of the image quality options cranked, and we tested with 16X anisotropic filtering and 4X antialiasing. The game’s multi-core rendering option was, of course, enabled.

Ah, now here is an example of a GPU bottleneck truly taking over at higher resolutions. Then again, we’re talking about a bottleneck that appears to limit frame rates to the mid-90’s.

At lower resolutions, we have a dead heat between the Core i3-530 and the Athlon II X4 635, amazingly enough. They’re both right in the middle of the pack, though they take very different roads to get there. Judging by the overall results, frame rates in this game are influenced pretty heavily by cache size and core count. The top ranks are populated exclusively by quad-core CPUs with large caches.

Source engine particle simulation
Next up is a test we picked up during a visit to Valve Software, the developers of the Half-Life games. They had been working to incorporate support for multi-core processors into their Source game engine, and they cooked up some benchmarks to demonstrate the benefits of multithreading.

This test runs a particle simulation inside of the Source engine. Most games today use particle systems to create effects like smoke, steam, and fire, but the realism and interactivity of those effects are limited by the available computing horsepower. Valve’s particle system distributes the load across multiple CPU cores.

The newer Intel processors seem to gain quite a bit from Hyper-Threading in this test—witness the Core i5-661 nearly matching the HT-less Core i5-750. Astoundingly, the Core i7-975 Extreme achieves over nine times the throughput of the Pentium 4 670.

Productivity and general use software
We have, for quite some time now, used WorldBench in our CPU tests. Over that time, we’ve found that some of WorldBench’s tests can be rather temperamental and may refuse to run periodically. We’ve also found that some of the same tests tend to have inconsistent results that aren’t always influenced much by processor performance. Other applications in WorldBench 6, like the Windows Media Encoder 9 test, make little or no use of multithreading, despite the fact that such applications are typically nicely multithreaded these days. As a result, we’ve decided to limit our use of WorldBench to a selection of its applications, rather than the full suite.

MS Office productivity

This test fires up multiple applications from the Microsoft Office suite and switches back and forth between them in order to simulate a multitasking user session. Even so, it doesn’t demand or extract much from additional CPU cores, and really, there’s little separation between the CPUs overall. Even the Pentium 4 670 isn’t that slow in the grand scheme of things.

Firefox web browsing

AMD’s Athlon IIs struggle here compared to the Core i3 processors, likely due to their smaller caches. The Phenom II chips, with their L3 caches, fare pretty well. Heck, even the Athlon II X2 255, with its larger L2 caches, outruns the Athlon II X3 and X4 chips. Then again, the Athlon IIs finish in the order of their core clock frequencies, so who knows?

Multitasking – Firefox and Windows Media Encoder

This one encodes a video in the background while running the same Firefox browser test used above. Adding that background task changes surprisingly little about the overall results. Among the AMD chips, the finishing order is unchanged.

File compression and encryption

7-Zip file compression and decompression

This application scales nicely with multiple threads and cores, and the four true cores of the Athlon II X4 deliver a victory over the Core i3, although the margin of victory is thinner than one might think.

WinZip file compression

If you keep moldy, old executables like WinZip 10 around, here’s the sort of behavior you can expect. Without multiple threads, WinZip 10 relies on the computing power a single CPU core.

TrueCrypt disk encryption
Here’s a new addition at our readers’ request. This full-disk encryption suite includes a performance test, for obvious reasons. We tested with a 50MB buffer size and, because the benchmark spits out a lot of data, averaged and summarized the results in a couple of different ways.

TrueCrypt doesn’t yet support Westmere’s new encryption-related AES instructions. We’ll have to try this test again once the software is updated to make use of them.

I’m going to talk about the overall average results, since those are easy enough to grasp quickly. Feel free to browse through the more detailed results below if you’re interested.

For our purposes, the most obviously notable outcome is the Athlon II X4 635’s pronounced advantage over the Core i3-530. Also noteworthy: the fastest Core i7 manages more than eight times the throughput of the Pentium 4 670.

Image processing

The Panorama Factory photo stitching
The Panorama Factory handles an increasingly popular image processing task: joining together multiple images to create a wide-aspect panorama. This task can require lots of memory and can be computationally intensive, so The Panorama Factory comes in a 64-bit version that’s widely multithreaded. I asked it to join four pictures, each eight megapixels, into a glorious panorama of the interior of Damage Labs.

In the past, we’ve added up the time taken by all of the different elements of the panorama creation wizard and reported that number, along with detailed results for each operation. However, doing so is incredibly data-input-intensive, and the process tends to be dominated by a single, long operation: the stitch. So this time around, we’ve simply decided to report the stitch time, which saves us a lot of work and still gets at the heart of the matter.

Once more, the Athlon II X4 635’s four real cores grant it the edge over the Core i3-530. Six seconds of destiny! The Pentium E6500 outperforms the Athlon II X2 255, though.

picCOLOR image processing and analysis
picCOLOR was created by Dr. Reinert H. G. Müller of the FIBUS Institute. This isn’t Photoshop; picCOLOR’s image analysis capabilities can be used for scientific applications like particle flow analysis. Dr. Müller has supplied us with new revisions of his program for some time now, all the while optimizing picCOLOR for new advances in CPU technology, including SSE extensions, multiple cores, and Hyper-Threading. Many of its individual functions are multithreaded.

Recently, at our request, Dr. Müller graciously agreed to re-tool his picCOLOR benchmark to incorporate some real-world usage scenarios. As a result, we now have four new tests that employ picCOLOR for image analysis. I’ve included explanations of each test from Dr. Müller below.

Particle Image Velocimetry (PIV) is being used for flow measurement in air and water.
The medium (air or water) is seeded with tiny particles (1..5um diameter, smoke or oil fog in air,
titanium dioxide in water). The tiny particles will follow the flow more or less exactly, except may be
in very strong sonic shocks or extremely strong vortices. Now, two images are taken within a very
short time interval, for instance 1us. Illumination is a very thin laser light sheet. Image resolution is
1280×1024 pixels. The particles will have moved a little with the flow in the short time interval and
the resulting displacement of each particle gives information on the local flow speed and direction.
The calculation is done with cross-correlation in small sub-windows (32×32, or 64×64 pixel) with some
overlap. Each sub-window will produce a displacement vector that tells us everything about flow speed
and direction. The calculation can easily be done multithreaded and is implemented in picCOLOR with
up to 8 threads and more on request.

All of picCOLOR’s results are indexed against a reference system, based on a Pentium III 1GHz, whose performance equals a score of 1.0 in each test. For instance, in the Particle Image Velocimetry test above, the Core i7-975 Extreme is a staggering 36.2 times faster than the PIII 1GHz, and even the Pentium 4 670 offers 4.7 times the speed.

Real Time 3D Object Tracking is used for tracking of airplane wing and helicopter blade deflection and deformation in wind tunnel tests. Especially for comparison with numerical simulations, the exact deformation
of a wing has to be known. An important application for high speed tracking is the testing of wing flutter, a
very dangerous phenomenon. Here, a measurement frequency of 1000Hz and more is required to solve the
complex and possibly disastrous motion of an aircraft wing. The function first tracks the objects in 2 images
using small recognizable markers on the wing and a stereo camera set-up. Then, a 3D-reconstruction
follows in real time using matrix conversions. . . . This test is single threaded, but will be converted to 3 threads in the future.

Multi Barcodes: With this test, several different bar codes are searched on a large image (3200×4400 pixel).
These codes are simple 2D codes, EAN13 (=UPC) and 2 of 5. They can be in any rotation and can be extremely fine
(down to 1.5 pixel for the thinnest lines). To find the bar codes, the test uses several filters (some of them multithreaded). The bar code edge processing is single threaded, though.

Label Recognition/Rotation is being used as an important pre-processing step for character reading (OCR).
For this test in the large bar code image all possible labels are detected and rotated to zero degree text rotation.
In a real application, these rotated labels would now be transferred to an OCR-program – there are several good programs
available on the market. But all these programs can only accept text in zero degree position. The test uses morphology
and different filters (some of them multithreaded) to detect the labels and simple character detection functions to locate the text and to determine the rotational angle of the text. . . . This test uses Rotation in the last important step, which is fully multithreaded with up to 8 threads.

The Core i3-530 takes three of the four real-world tests from the Athlon II X4 635. The Core i3 looks to be slightly superior for this sort of image analysis. Meanwhile, though, the Athlon II X2 255 snags three of four over the Pentium E6500, and the fourth is a tie.

picCOLOR’s synthetic tests measure a number of the program’s individual functions, and the program then computes an average score, again indexed versus a 1GHz Pentium III. This should be a pretty good index of overall image processing performance. Although the Athlon II X4 635 comes out ahead of the Core i3-530, the gap between the processors remains thin.

Image handling looks to be one of those areas, much like our 3D gaming tests, where the competing CPUs from Intel and AMD are at approximate performance parity. Since these image processing programs are nicely multithreaded wherever possible, that’s quite an accomplishment for the dual-core Core i3-530—although we’ve come to expect such feats from Intel’s latest architecture.

Media encoding and editing

x264 HD benchmark
This benchmark tests one of the most popular H.264 video encoders, the open-source x264. The results come in two parts, for the two passes the encoder makes through the video file. I’ve chosen to report them separately, since that’s typically how the results are reported in the public database of results for this benchmark

If you’re into encoding video, you’ve no doubt come to appreciate the benefits of a fast multi-core processor. The Athlon II X4 635 finishes pass two of this process in about half the time that the Athlon II X2 255 or Pentium E6500 does, for example. Video encoding has been an optimization target for CPU architects for some time now, and we’ve seen marked progress, obviously, as the Pentium 4 670’s relatively abysmal frame rates illustrate.

Windows Live Movie Maker 14 video encoding
For this test, I used Windows Live Movie Maker to transcode a 30-minute TV show, recorded in 720p .wtv format on my Windows 7 Media Center system, into a 320×240 WMV-format video format appropriate for mobile devices.

Since these results are measured in seconds, they illustrate my point even better. If you pick an Athlon II X4 635 instead of a Core i3-530 for your system, you can expect an encode process like this one to finish about 30 seconds sooner. The justification for a high-end processor is pretty clear: you’ll save over four and a half minutes by going with a Core i7-960 instead of a Pentium E6500. If time is money, fast computer hardware may seem rather cheap in the big picture.

LAME MT audio encoding
LAME MT is a multithreaded version of the LAME MP3 encoder. LAME MT was created as a demonstration of the benefits of multithreading specifically on a Hyper-Threaded CPU like the Pentium 4. Of course, multithreading works even better on multi-core processors.

Rather than run multiple parallel threads, LAME MT runs the MP3 encoder’s psycho-acoustic analysis function on a separate thread from the rest of the encoder using simple linear pipelining. That is, the psycho-acoustic analysis happens one frame ahead of everything else, and its results are buffered for later use by the second thread. That means this test won’t really use more than two CPU cores.

We have results for two different 64-bit versions of LAME MT from different compilers, one from Microsoft and one from Intel, doing two different types of encoding, variable bit rate and constant bit rate. We are encoding a massive 10-minute, 6-second 101MB WAV file here.

Audio encoding is one of those areas where multithreading will only take you so far. Unless you’re batching up multiple songs and encoding them together at once, you probably won’t benefit much from having more than two cores.

3D modeling and rendering

Cinebench rendering
The Cinebench benchmark is based on Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. This test runs with just a single thread and then with as many threads as CPU cores (or threads, in CPUs with multiple hardware threads per core) are available.

Chalk up another one for the Athlon II X4 635 over the Core i3-530.

POV-Ray rendering
We’re using the latest beta version of POV-Ray 3.7 that includes native multithreading and 64-bit support.

This seems like an appropriate place to stop and marvel at the way render times have dropped since the Pentium 4 670. The chess2.pov scene is handled in fully parallelized fashion, so our multi-core processors crunch through it pretty quickly—as little as a tenth of the time required by the Pentium 4 670.

Still, although PC hardware gets faster over time, software often gets slower. If you go look at our review from back in the day, the Pentium 4 670 rendered this same scene in 309 seconds using a single thread. Now it’s taken over 600 seconds to do it with POV-Ray 3.7. Just to make sure we didn’t have a configuration problem, I installed an old version of POV-Ray 3.6.1 64-bit from March, 2005 on our LGA775 test system. Lo and behold, the P4 670 completed the render in about the same time we’d measured way back when. POV-Ray’s renderer has surely gained features in the interim, but it’s not nearly as quick as it once was.

3ds max modeling and rendering

The first 3ds max test measures 3D modeling speed, not rendering, so it’s a bit of a different animal. As you’ll note, the P4 670 doesn’t get so badly creamed here as elsewhere, partly because the graphics card is a speed limiter in this test.

Valve VRAD map compilation
This next test processes a map from Half-Life 2 using Valve’s VRAD lighting tool. Valve uses VRAD to pre-compute lighting that goes into games like Half-Life 2.

Almost regardless of which rendering software you choose, the Athlon II X4’s four real cores and their relatively strong floating-point math ability put the X4 635 ahead of the Core i3-530.

[email protected]
Next, we have a slick little [email protected] benchmark CD created by notfred, one of the members of Team TR, our excellent Folding team. For the unfamiliar, [email protected] is a distributed computing project created by folks at Stanford University that investigates how proteins work in the human body, in an attempt to better understand diseases like Parkinson’s, Alzheimer’s, and cystic fibrosis. It’s a great way to use your PC’s spare CPU cycles to help advance medical research. I’d encourage you to visit our distributed computing forum and consider joining our team if you haven’t already joined one.

The [email protected] project uses a number of highly optimized routines to process different types of work units from Stanford’s research projects. The Gromacs core, for instance, uses SSE on Intel processors, 3DNow! on AMD processors, and Altivec on PowerPCs. Overall, [email protected] should be a great example of real-world scientific computing.

notfred’s Folding Benchmark CD tests the most common work unit types and estimates the number of points per day that a CPU could earn for a Folding team member. The CD itself is a bootable ISO. The CD boots into Linux, detects the system’s processors and Ethernet adapters, picks up an IP address, and downloads the latest versions of the Folding execution cores from Stanford. It then processes a sample work unit of each type.

On a system with two CPU cores, for instance, the CD spins off a Tinker WU on core 1 and an Amber WU on core 2. When either of those WUs are finished, the benchmark moves on to additional WU types, always keeping both cores occupied with some sort of calculation. Should the benchmark run out of new WUs to test, it simply processes another WU in order to prevent any of the cores from going idle as the others finish. Once all four of the WU types have been tested, the benchmark averages the points per day among them. That points-per-day average is then multiplied by the number of cores on the CPU in order to estimate the total number of points per day that CPU might achieve.

This may be a somewhat quirky method of estimating overall performance, but my sense is that it generally ought to work. We’ve discussed some potential reservations about how it works here, for those who are interested. I have included results for each of the individual WU types below, so you can see how the different CPUs perform on each.

However, the individual results for each unit type tend to get a little wonky because the CPUs with Hyper-Threading are running multiple threads on each core. To get a clear sense of performance for all CPUs, you’ll want to focus on the final graph showing total projected points per day.

This is our final benchmark test (phew!), and it’s another case where the Athlon II X4 635’s four true cores give it the win. Even so, the Core i3-530 delivers more points per day than the Core 2 Quad Q6600, showing that two of today’s cores are sometimes superior to four of the prior generation’s.

I’m going to start by talking about my overclocking attempt for the Core i5-661, since I didn’t have time to include overclocking results in our initial review of that product. The i5-661 was the first of this bunch I attempted to overclock, followed by the i3-530 and the Athlon II X4 635.

As usual, our overclocking efforts involved some quick-and-dirty assessments of stability and a pretty vanilla suite of tweaks—changing the frequency by modifying the base clock speed, raising the core CPU voltage and some related values, lowering the memory multiplier to keep DIMM speeds sane, and keeping an eye on cooling. We conducted some quick stability tests with each new clock rate using Prime95, and we then benchmarked the best stable speed for each processor.

Using those simple methods, I was able to coax the Core i5-661 up to a frequency that makes my head spin: 4.5GHz at a very healthy 1.4V. That’s nearly a 50% overclock on the Core i5-661’s 3.3GHz stock speed (although the i5-661 can step up to 3.6GHz via Turbo Boost.)

Getting to this speed using our Asus P7H57D-V EVO motherboard was alarmingly easy. I just turned up the base clock from 133MHz, dropped the memory multiplier, and everything seemed to work right. Then I stepped up through various clock speeds waiting for the i5-661 to show signs of a problem. At 4.15GHz, I noticed the CPU was generating heat faster than Intel’s puny stock cooler for the Clarkdale processors could take it away. The i5-661 was stable during our stress test, but CPU temps climbed to 74° C before I pulled the plug. Next, I swapped in a Thermalright tower cooler, which kept temperatures comfortably in the 50s, and kept on pushing.

The reason the i5-661 overheated was the same reason overclocking seemed so strangely effortless: turns out the Asus mobo automatically adjusts several key voltage values, including the CPU voltage, when you’re overclocking. By the time the i5-661 hit 4.5GHz, the core was at 1.4V, well up from the 1.16V stock value. Having played with it some, I think the Asus BIOS is pretty intelligently tuned, and I’m convinced you’ll want a better cooler than the stock Intel one if you want to extract the most from a Clarkdale.

My success with the i5-661 made me excited to see what the i3-530 could do—and it did not disappoint.

The i3-530 was stable at 4.4GHz with a 200MHz base clock, again a nearly 50% overclock. For those of you who don’t know this fact, overclocking itself is magic, and when you reach a mark like 50%, your body releases endorphins into your brain, triggering a feeling of happiness and well-being. I was pleased to experience this sensation, which brought back fond memories of doing this a decade ago at frequencies an order of magnitude lower, when I overclocked a Celeron 300A to 450MHz.

The Asus board’s auto-voltage feature had taken the i3-530 up to 1.4V at 4.4GHz, but I was curious to see whether that much juice was really necessary. I started down at 1.2375V and tried to get the i3-530 to POST and boot at 4.4GHz, but I wound up stepping clear up to 1.3875V before the system was stable again. Obviously, Asus has done its homework on Clarkdale voltages, though you may want to tweak things yourself to keep power consumption in check.

The Athlon II X4 635 wasn’t quite as willing to rev, but it did make it up to 3.48GHz at 1.45V (stock is 2.9GHz at 1.4V). We were using a much beefier cooler on the X4 635, from a Phenom II X4 955, and it kept temperatures reasonable without us having to dig out a massive tower.

Here’s how the overclocked processors performed in a couple of benchmarks.

The Clarkdales are screamers at those clock speeds, matching up well against the fastest CPUs around. Heck, the overclocked i5-661 nearly ties the quad-core Core i5-750 and Phenom II X4 965 in Cinebench’s multithreaded test. Good grief.

Another question we wanted to investigate is what overclocking does to the power consumption of these processors. One reason today’s CPUs have so much clock speed headroom in them is that frequencies are usually limited primarily by power and thermal envelopes. Does overclocking a Clarkdale to ridiculous heights make any sense, or will you be left with a room-heating monster?

To find out, I stuck the overclocked systems on our power meter and ran our Cinebench power test. I was using the Asus H57 board at this point, remember, which draws a little more power than the Gigabyte P55 board we used for our main power consumption tests, so I’ve included and marked the H57 results in the graphs.

The Athlon II and its motherboard seem to do a better job of dropping voltage levels at idle while overclocked than the Core i3-530 on the Asus H57. The Asus simply holds the voltage steady at 1.3875V, regardless. That explains the high idle power draw for the overclocked Core i3-530.

At any rate, the overclocked systems are entirely reasonable, despite some big gains in power consumption. The overclocked Core i3-530 system’s peak power draw is a couple of watts lower than our Core i7-870 test system’s, and the overclocked Athlon II X4 635 is a few watts below our Phenom II X4 965. If you pay extra to Intel and AMD to get a faster CPU, you’ll still be in the same thermal territory as our overclocked specimens. You will need a decent air cooler in order to overclock them, but nothing too elaborate.

The value proposition
Now that we’ve buried you under mounds of information, what can we make of it all? One way to filter the information is to consider the value proposition for each CPU model. Exercises like this one are inherently fraught with various, scary dangers—giving the wrong impression, committing bad math, overemphasizing price, coming off as irredeemably cheesy—but our value comparisons have proven to be popular over time, so with the capable assistance of TR System Guide guru Cyril Kowaliski, I’ve taken another crack at it.

What we’ve done is mash up all of our performance data in one, big summary value for each processor. The performance data for each benchmark was converted to a percentage using the Pentium 4 670 as the baseline. We’ve included nearly every benchmark we used in our overall index, with the exception of the purely synthetic tests like Stream. In cases where the benchmarks had multiple components, we used an overall mean rather than including every component score individually. Each benchmark should thus be represented and weighted equally in the final tally. (The one case where we didn’t average together a single application’s output was WorldBench’s two 3ds max tests, since one measures 3D modeling performance and the other rendering.)

This overall performance index makes me a little bit wary, because it’s simply a mash-up of results from various tests, rather than an index carefully weighted to express a certain set of priorities. Still, our test suite itself is intended to cover the general desktop PC’s usage model, so the index ought to suffice for this exercise.

We then took prices for each CPU from the official Intel and AMD price lists or, in the case of the new Athlon II models, directly from AMD. Since the Phenom II X2 550 isn’t on AMD’s price page, we took its price from Newegg. For our historical comparison, we’ve also included the Core 2 Quad Q6600 and the Pentium 4 670 in a couple of places at their initial launch prices.

If we simply take overall performance and divide by price, we get results that look like this:

This bar chart does give us a strong sense of value, no doubt—and the Athlon IIs look excellent in this light—but it may focus our attention a little too exclusively on CPU prices alone. As I’ve mentioned, for many of us, time is money, and faster computer hardware is relatively inexpensive. What we really want to know is where we can find the best combination of price and performance for our needs. To give us a better visual sense of that, we’ve devised our nefarious scatter plots.

The faster a processor is, the higher on the chart it will be. The cheaper it is, the closer to the left edge. The better values, then, tend to be closer to the top-left corner of the plot. If you wish, you can find your price range and look for the best performer in that area.

With the data plotted in this fashion, we can see a few other contenders that might join the Athlon II X3 and X4 processors as value stand-outs at higher performance levels, including the Core i5-750, Phenom II X4 965, Core i7-920, and even the Core i7-960. The ghosts of the P4 670 and the Core 2 Quad Q6600 haunt our value scatter plot, as well, reminding us of the dismal CPU values in days past.

That gets us closer to the heart of the matter, but in reality, the price of a processor is just one component of a PC’s total cost, and the various platforms do have some price disparities between them. After an epic feud that involved pitchforks, shotguns, and various hurled insults, we finally agreed on some sample systems loosely based on the Utility Player build in our latest system guide for each platform type. Our goal was to achieve rough parity by selecting full-sized ATX motherboards with similar, enthusiast-friendly feature sets. Here are the components we picked for the different platforms, along with system prices:

Platform Total price Motheboard Memory Common components
AMD 790X $608.94 Gigabyte GA-770XT-USB3
4GB Corsair DDR3-1333
XFX Radeon HD 5770 1GB graphics card ($159.99), Western Digital Caviar Black 640GB hard drive ($74.99), Samsung SH-S223L DVD burner (28.99), Antec Sonata III case with 500W PSU ($114.99)
Intel P45 $623.94 Gigabyte GA-EP45T-USB3P
Intel P55 $603.94 Gigabyte GA-P55-USB3
Intel X58 $758.94 Gigabyte GA-X58A-UD3R
6GB Corsair DDR3-1600

What happens when we factor system prices into our value equation?

Whoa. Suddenly, you should buy a Core i5-750! That’s more my kind of recommendation. Print this one out and show it to your spouse and/or boss, folks. In the context of a beefy system like this one, going with a cheap CPU like an Athlon II X2 255 or a Pentium E6500 doesn’t make a heckuva lot of sense. You’d be paying relatively little more to get substantially higher performance from a faster processor.

Note that, in our main event, the Athlon II X4 635 comes out a little ahead of the Core i3-530, though the contrast between them is fairly minor. Here’s the scatter version.

The inclusion of total system prices alters the complexion of our scatter plot somewhat, too, mainly by making the LGA775 and LGA1366 processors look less attractive. The cheaper chips lose their luster, as well. The Core i5-750 and i7-870 remain nicely positioned, while the poorer values include the Core 2 Duo E8600, the Q9400, and the Core i5-661.

Performance per dollar isn’t the whole story these days, though. The power efficiency of a processor increasingly helps determine its value proposition for a host of reasons, from total system costs to noise levels to the size of your electric bill. We measured full system power draw and considered efficiency earlier in this article; now, we can factor in system prices to give us a sense of power-efficient performance per dollar.

By this measure, the Core i3-530 is near the top of the charts, and the Athlon II X4 635 is stuck in the middle of the pack—and you should still buy a Core i5-750. The scatter plot tells the story a little differently.

The power efficiency of Intel’s newer processors is especially evident here. At system prices around $700, the Core i3-530 is easily superior to the Athlon II X4 635, and at about $800, Intel has three offerings that prove more efficient than the Phenom II X4 965. AMD’s dual- and triple-core parts cluster near the bottom corner, cheap but inefficient.

In the past 17 pages, I have unconsciously worked like mad to atone for choosing a liberal arts major. The amount of information we’ve foisted upon you is otherwise inexcusable, I’m aware. If you will permit me, I’d like to turn things around by saying a few words about the primary subjects of our attention today.

The contest between the Core i3-530 and the Athlon II X4 635 boils down to a few simple considerations. For about the same price as the Core i3-530, the Athlon II X4 635 offers four cores that perform better in applications that rely heavily on multiple threads, such as video encoding, 3D rendering, and [email protected] In other uses, such as video games and image processing, these two CPUs perform almost identically. The Athlon II X4 635 leads slightly in overall performance and, as we established on the previous page, in terms of performance value. If that’s all you care about when choosing a processor, then your decision has been made.

The trade-off here is that our X4 635 system draws 50W more than the i3-530-based one when engaged, a consequence of the fact that Intel requires only two 32-nm cores to achieve throughput similar to four AMD cores at 45 nm. The X4 635’s power draw isn’t unreasonable by desktop PC standards, but the i3-530’s is exceptional. That fact alone might not be enough to persuade me to prefer the i3-530 for my own system. Combine it with the overclocking headroom we’ve seen from our two Clarkdale processors, though, and I’m sold. The prospect of a near-50% overclock, and the speed that goes with it, has me salivating. Either of these CPUs is a pretty good choice, but I’d make mine a Core i3-530—along with an embarrassingly tall tower cooler.

That’s what I’d choose between our two headliners, at least. I suppose I should say a word or two about the cheaper processors, as well. The Athlon II X2 255 and the Pentium E6500 pretty much tied in our overall performance index and, heck, in quite a few individual tests. They both passed our gaming tests with acceptable frame rates, and for general use, they’ll blow away a “nettop” processor like a dual-core Atom. They’d be practically interchangeable, if the X2 255 didn’t draw an extra 28W of power at peak. The E6500 may be older Intel technology, but it still has the X2 255 beaten.

Before we go, we can’t ignore the fact that our overall leader in both power efficiency and performance per system cost was the Core i5-750. If you’re purely rational about these things—and you can afford to spend nearly $200 on a CPU—the i5-750 is obviously the best choice among the processors we tested.

I’d also like to pour out a 40 for my homie who couldn’t be with us, the Core i7-860. At $284, the i7-860 improves on the i5-750 by adding higher clock speeds and Hyper-Threading to the mix at the same 95W TDP. We suspect the i7-860 might be Intel’s most appealing desktop processor. Perhaps soon we’ll have the chance to test one—you know, what we really need is more data—but based on a few things we already know, we wouldn’t hesitate to recommend one now.

0 responses to “Core i3 takes on Athlon II

  1. I found this article very informative and appreciate the planning, work, and dedication that it took to acquire so much data and reduce it into summarizations that made sense.

    I have a question about the Value Proposition section of the report. The test systems, as defined, were all based on Asus motherboards while the “recommended” systems all use Gigabyte motherboards. On which system costs is the value segment of the report calculated?

    Why do I ask? One example is because if someone were to copy-cat a system to achieve the results from this report, their trust in doing so will have been violated. The test data presented defines performance of the test system only. The “value” systems become subjective recommendations when any component is changed. The Gigabyte MB selection may have been an informed and knowledgable one, but ensuing performance cannot be associated with the reported test results.

    In conclusion, if test system costs were used to calculate relative value then I apologize for asking the question. If the “value” system costs were used, then I’d like to see a report addendum that applies to test system costs along with a clarification that use of a Gigabyte MB may yield different results; performance could improve or fall short of test data reported. Thank you.

  2. “Clarkdale CPUs will plug into LGA1155 or LGA1156-style sockets, but you’ll need an LGA1155 motherboard to take advantage of the integrated graphics. Here’s a look at how the i3-530 fits into the broader Clarkdale lineup”

    Nope. You will not be able to plug a clarkdale cpu into a LGA1155 socket. Motherboards based on H55/H57/Q57 chipsets are needed to utilized the integrated graphics of clarkdale CPUs, which are also LGA1156 socket boards. LGA1155 was reported to come along with the upcoming Sandy bridge CPUs and is not backward compatibility with Clarkdale/Lynnfield.


  3. Exactly, Alouette. The Athlon II X4 systems I’ve been building for my budget-conscious clients only cost me $385 for the whole thing (minus the OS), not $700 as the article would put it. I use Micro Center, but if one prefers Newegg, they have comparable prices and even more choices. Check it out:

    Athlon II X4 630: $100
    MSI AMD 785G Motherboard (770 chipset + HD 4200 graphics): $70
    3GB (2GB+1GB) Kingston or Crucial DDR2-800: $70
    Thermaltake 430W PSU: $35
    LG SATA DVD+RW: $25
    Gigabyte ATX Mid-tower: $25
    500GB Seagate 7200.12: $60
    Total: $385

    For parity with the article, add in a GTX 260 and thus the necessary 600W Thermaltake PSU, swap in 4GB of Corsair DDR2, and upgrade to a 1TB Seagate 7200.12, all for (160 + 40 + 30 + 30 = $260) more.

    After all that you’ve STILL got the ENTIRE AMD line standing a lot closer to the left side of the graph (and a hair lower due to DDR2) than the Intel offerings, at $645 for the X4 630 system or $725 for an X4 965 system.

    If one needed gaming performance beyond my $385 base system, I’d probably pair the lower-end X4 630 with an HD 5670 and a 500W Antec, making the whole system $560 (even with 4GB Corsair and a 1TB HD).

    These kinds of price differences may seem like splitting hairs, but when the whole premise of this section of the article is the silly notion that Intel somehow WINS on price-to-performance, I have to interject and state the facts: AMD cannot be beaten, from the low- up to the upper-mid-end of the spectrum, on the price-to-performance metric.

    To completely ignore the DDR2 option for these AMD CPUs (and the negligible 5% performance decrease that comes with it) is to completely ignore the POINT of price-to-performance comparisons. Obviously Intel has AMD beat above the i5 750 / X4 965 threshold, but at that point it’s not really about price, is it?

  4. Well, heh with the triple-crossfire boards I was being a bit facetious but I was just trying to make a point. I don’t know for sure that Gigabyte is necessarily more respected than ASRock seeing how ASRock comes from ASUS which to my knowledge has always been legendary. I never really worried myself about what my motherboard brand was any more than I worry about the brand of my vidcard (unless I can have a lifetime warranty for maybe $10 more but no mobo has that). I always looked for the features. I have an ASRock motherboard now running a Pentium Dual-Core that is OLD. I’m sure you remember, the 4Core Dual-VSTA? That showed a lot of innovation because nobody else had one. Now I have an MSI K9A2 Platinum. If you ask some people, MSI is crap. *scratches his head* What I was getting at was that if you want to compare what CAN be had, then the AMD chips rule with pricing. Especially when you consider that sure, maybe the i3 overclocks better but with the cost savings, you can just buy the AMD CPU that natively runs the same speed as the overclocked i3. Then the OC advantage is to AMD.

    Thank you for your response. You gave a great explanation and I look forward to reading your next review! 😀

  5. Avro:

    Hey, good to hear your feedback. Let me explain myself on a couple of points, if you will.

    On the code names, the reason I wrote what I did, listing the Intel code names and then saying “Socket AM3 processors” was pretty clearly evident in the data to which I was referring. The results were separated into tiers by chip and socket type, and I wanted to explain that fact. Since Intel’s naming scheme prevents easy summarization (Core i5 and i7 overlap across chips, for instance), using the code names was the easiest shorthand available to me. Delineating the AMD processors by code name would have gone against my purpose, since all of the Socket AM3 chips were grouped together in performance.

    I had, of course, presented a table of code names and chips earlier in the article and had full command of the AMD code names if needed.

    As for the value section, one can always skew that analysis a little differently by going for cheaper motherboards from less respected brands or the like. Our goal, as stated, was to try for reasonable parity, and all of the boards came from a respected brand: Gigabyte. If you dislike our choices, as you seemed to do, we have of course shown our work and enabled you to offer alternatives, as you have done.

    I would note that your suggested alternative pricing seems to avoid cheaper H55/H57 boards available for the LGA1156 processors. Also, it seems to me that triple CrossFire is an odd feature on which to key in making such choices. I expect many of our readers would have taken us to task had we done so.

    You may be interested to see that we did an alternative analysis with cheaper motherboards and components here:

    §[<<]§ The AMD chips proved to be better values in a less expensive overall build, as one might suspect. As for your final question, I guess the answer depends on the context. In our recent system guide, we did in fact pick an Athlon II X4 630 as the processor in our Econobox build, precisely because we were putting it into a relatively inexpensive system. See here: §[<<]§ So in the right situation, yes, we would (and have) changed our choice based on value concerns. However, if you are asking me to choose a CPU for my own system, I would personally be willing to spend the extra to go with a Core i3-530 instead, for the same reasons I mentioned in the conclusion of the review: sweet, sweet overclocking headroom and vastly superior power efficiency from the i3-530. Neither of those things figures into a raw perf/dollar equation, but they matter a lot to me. Scott __ Scott Wasson The Tech Report PC Hardware Explored §[<<]§

  6. Ah well, my breath was already wasted. I just had to say something because I couldn’t believe that people didn’t see the pro-criminal Intel bias. I also want people to realize that it IS cheaper to go AMD. This whole review of these things just smacks of Intel propaganda and all anyone else can do is cheer it on like a bunch of Zombies.

  7. Logical motherboard selection has never been a strong suit of any PC site I’ve seen. I think a lot of it depends on what companies give them to test, which is never going to be the cheap ones that make sense.

    Don’t waste your breath.

  8. Just wanted to add that ASRock used to be a division of ASUS so it’s not like I chose a crap manufacturer.

  9. I’m very sorry but I’m going to have to give this review two thumbs down and here’s why:

    Your Intel bias was immediately evident with this line:

    “These results are separated neatly by CPU type. The triple-channel Bloomfield Core i7s are fastest, followed by the Lynnfields, Clarkdales, and Socket AM3 processors.”

    I’m sorry but taking the time to say Bloomfield, Lynnfield and Clarkedale but then not taking the time to say Deneb, Propus, Rana and Regor but instead calling them “Socket AM3 Processors” shows either a bias towards Intel / Dislike of AMD or it shows a level of ignorance that would have had you fired long ago. Since you’re still there and you were even able to put this together, my guess is that you’re not an ignorant person.

    The next problems are with the value section. First of all, that AMD motherboard is misprinted as a fictional AMD 770 motherboard but typos aren’t a crime. The problem I have is that you chose an overpriced board with an inferior chipset. You TRIED to make the AMD motherboard more expensive. How do I know? I’ll show you. The 790X is a dual-crossfire motherboard with no on-board video and relatively weak features. This Gigabyte board is $125, that is a fact but it’s also a fact that newegg sells the ASRock M3A785GXH/128M AM3 motherboard with TRIPLE CrossfireX and a Radeon HD 4200 IGP. The only thing that the Gigabyte board has that the ASRock board doesn’t is a mini firewire port. Thanks, but I’ll take the on-board graphics and the extra crossfire slot. At the very heart of the matter is the fact that this ASRock motherboard costs only $92.00 and it is always in stock:
    §[<<]§ Now I'm not trying to compare apples to oranges here. A Gigabyte should not be compared to an ASRock. It just so happens that ASRock also makes a triple-crossfire board in LGA1156/P55 trim. It's only $140.00 at newegg: §[<<]§ It doesn't have on-board graphics either but really, who cares when you have 3 CrossfireX slots? It has the added advantage of also supporting SLI but if I had selected any other ASRock LGA1156/P55 board, the AMD would have had the unfair advantage of having a 3rd Crossfire slot. Now I must ask you Scott... Suddenly the i3 motherboard is $48 more expensive than the Athlon II motherboard. Does that perhaps change the conclusion of your comparison?

  10. Nice to have such in-depth review. Will really be complete once a G6950 is added. I do wish the overclocks were done first and included in all the tests. Also, better resolution choices these days would be 1280×960, 1680×1050 and 1920×1080. Would also be great to see a value graphics option included. Could cut down the number of tests and CPUs to balance the effort.

  11. The point of that particular P4 was not to have something that the most people could compare to based on what they personally owned.

    I already said this but I’ll just copy and paste my post again:

    “It’s the ultimate example of how far we’ve come from high speed single-cores.

    Most of the recent CPU articles have had the 90nm Athlon 64 X2 6400+ and we know where that stands.”

    It’s really not that helpful to have any of those in there. That one P4, in particular, just proves a significant point.

  12. I don’t think you understand. They aren’t saying taxes aren’t needed. They are saying don’t over pay. That is just an interest free loan to the gov. The best result is a tax return where you own nothing and they own you nothing.

  13. HA. You guys are retarded. dont drive on the roads or go to the hospitals (if you live in Canada like me) taxes are necessary.

  14. Maybe not the 965XE specifically, but the 9xx Pentium Ds were actually pretty popular among enthusiasts for a while, because you could get some huge overclocks on them. However, I can’t imagine anyone stuck with those things once Core 2 showed up on the scene.

  15. I miss the scatterplots in the GPU-reviews now, it’s so much easier to figure out if it’s worth spending another $20 on a component with these nifty graphs! Please, pretty please with a cherry on top, would you try and make a price-performance scatterplot the next time you guys review a graphics card? <3

  16. $200 is a STEAL for an i7-860! I paid almost that for my i5-750. Dang, you Americans get the best deals!

  17. Definitely. I couldn’t be happier with my recent build, putting an i5-750 on a Gigabyte GA-H55M-UD2H. The CPU needs no introduction, Scott has just put it near the top of the heap in terms of value and performance. But the real gem is this micro-ATX motherboard. It cost $30 less than the cheapest P55 offering from Asus / Gigabyte / MSI (where I live), but sports the ALC889 codec, HDMI, eSATA and FW.

    The only things I lose is Crossfire / SLI support and a couple of SATA ports. I won’t be missing them, though. So why pay more for features I don’t use?

  18. I was just yesterday speccing out a system for a colleage, one of the options we considered for him was a mATX H55 board with a i5 750. That’s a lot of cores, turbos and IPC for your money me thinks.

  19. If venting at the IRS today, please limit it to the internet and stay out of planes, though! 😛

  20. Seriously? I bet you are a real fun person in real life! Geez.
    Rain on a happy person’s parade why don’t you.

  21. /[<"Hooray for income tax returns!!"<]/ Or you could look at it as the government took your money and the interest on that money, but is not paying you back that interestg{<.<}g You lose out on potential investment and/or interest income.

  22. Just got one from MicroCenter – best deal around by far right now IMHO.
    Along with 6GB of Corsair XMS 1600 DDR3 and a MSI P55-CD53 for less then $500.

    I know what I’ll be messing with the next few days 🙂
    Hooray for income tax returns!!

  23. I just wanted to say that I thank you for this awesome review and hard work! Seriously, this was so in-depth that it required a second reading and so much fun to read that it was worth reading it twice!

    So, keep up the good work and well worth the wait.

  24. Did more people had the P4 670 than the Extreme Edition? If so, I think it provides more meaningful comparisons than the EE. There was only one Extreme Edition among the 21 CPUs tested anyway.

  25. Well if we are going to get technical, the absolutely highest performance netburst CPUwas the Pentium Extreme Edition 965. This dual core CPUcame with 4 MB total L2 cache and clocked at 3.73 GHz for each core. This would be a good CPU to test present day CPUs against since it represents the highest pinnacle of the netburst era.

  26. Sorta surprised when I first saw this on Slashdot submitted by you. A few of the articles from here have been on there.

    Overall good article as always. I was a bit disappointed when the P4 was not one of the first ones (around the Athlon XP time period).

  27. P4 670 was pretty much the fastest P4 you could realistically buy.

    The last of P4 XEs just had more cache and were clocked slightly lower than 670. (3.8Ghz versus 3.76Ghz). I doubt that the extra cache would have made any difference in this line-up.

    E7600s and Q6600s just beat the snot out of any Netburst chip.

  28. Ditto. The plots illustrate the performance / value proposition very well. In the grand scheme of things, though, it looks like each step up the CPU performance ladder doesn’t cost too much. Well, at least up to Core i5-750, since the Core i7-860 processors cost $84 more, and the X58 platform is even more expensive. The little steps add up, of course, but the sweet spot for many people is likely right before the price starts to climb steeply.

    Considering that the jump from a HD5770 to a HD5850 can cost $100, it makes little sense trying to save a few bucks by going with a lesser CPU unless you are on a tight budget. The Core i5-750 looks especially good, and “average” gamers (mild overclock, single graphics card) can save on the motherboard by going with a H55 micro-ATX one instead of a P55 ATX one.

  29. Damn fine review, very in-depth and well put together, but I do hold the complaint that there’s nothing in the Pentium-D era in there.
    Those netburst-based steaming piles of silicon were pretty common between that P4-670 and the e6500, and the difference two netburst cores over one makes interests me.
    Hell, for that matter, how about that Pentium Extreme(or emergency) Edition that took the place of the enthusiast chip of the era, with it’s unlocked multi, 3.46/3.75Ghz clock speeds AND hyperthreading?
    Netburst-based they all may have been, but I still see enough of those machines about, including my emergency XE box, I know at least four other people using Pentium D systems. Be nice to know where they stand among the ranks of the e7600 chips and q6600 chips that everyone seems to have these days (Your machine sucks <—–>this much now!).

  30. I guess I lost this round to bdwilcox, because my wife laughed hardest at 113 too. I was teasing her this morning and she told me to kiss it. I said, “It’s hard to believe, but you’re the second person to tell me that in 24 hours,” and showed this thread.

    Wilcox, we’re lucky Scott settled us down first (and thanks in advance for not making a joke about being in the backseat with her).

  31. The memory bandwidth makes me think its time to upgrade from my Q6600 to i920 for After Effects work.

  32. q[<]§ §[<<]§ It probably doesn't make a huge difference to the value proposition, but since you bothered to overclock the Core i3 -- it might have been worthwhile using ACC or NVCC to unlock one or more additional cores on your Phenom II X2. (I own one and one of the two disabled cores is now fully functional with NVCC at stock clocks with a mere 0.025 Vcore boost, being Orthos and Intel Burn Test stable. A three core processor with 6 MiB L3 cache at 3.1 GHz and potentially a little higher isn't half bad for a day's worth of testing.)

  33. totally agree.

    is there any reason why, to placate those who are complaining about the costs of the systems configuration in comparison to the cost of the cpu, that you cant just ‘move the dot’ on the scatter plot to the left or right in order to match up how much you would pay for the system? how much would performance really change…?

  34. That’s right and the linky goodness has brought our little corner of the tech web to its knees. I think the engadget link among others is the cause of the sluggish performance of TR today (Engadget had to yank the link out of courtesy I think).

    Any publicity is good publicity I suppose even if you don’t have the bandwidth.

    Time to buy more servers at Damage Labs! 🙂

  35. Not sure anyone mentioned this, but the article got a mention on Engadget. Very nice write-up, Scott.

  36. There’s a need for an upgrade when the amount of additional performance you want is worth more to you than the amount of money you’re losing. It’s nothing about absolute statistical benchmark score differences.

    If the Q6600 is fast enough overclocked, then it’s worth keeping. If not, a 4.5GHz i5-661 could be well worth the money.

  37. I’m sure you’ll figure it out. Something tells me you’re an expert when it comes to men’s ass-cheeks.

  38. A standard user will go to Best Buy and say, “I need a computer.” And rely on that sales associate’s expertise/ineptitude to get them a good system with nary a thought as to third-party benchmarks.

    A user who does a variety of things on the desktop will value the general price/performance comparison a bit more than a more specialized comparison.

    The professional who is more concerned about specific applications will actually just look at that page of the article with the corresponding graphs.

  39. Sure, you’re a busy guy. I already feel blessed for all the time you’ve given with helpful and constructive contribution like your second point in this thread’s OP.

  40. Lord almighty! I see what you’re talking about now.

    Nehalems normally drop to about 0.8v at idle. Since it not only wasn’t doing that when they overclocked it, but was also running drastically higher voltage, naturally, the power use skyrocketed.

    I don’t think it’s really supposed to work that way. It looks like the motherboard automatically disabled voltage regulation whenever it increased it. Lame on a stick.

  41. I think the only place price/performance graphs work is when comparing cpus in the same family (So the rest of the system is identical). Even then there are going to be people who want everything overclocked and compared.

    There are just too many options. The first time an individual sees a combination of parts that he/she doesn’t agree with there is trouble.

    Where would Dell’s $249 Zino Hd show up on the graph?

    There is always going to be a guy, “I can buy an ecs mb for $37, get a processor off ebay for $19 and steal 4gigs of ram off by brothers computer, why isn’t that on the graph?”

    Somewhere in a dark corner, Steve Jobs is folding his hands, doing his Montgomery Burns impression.

  42. Sorry, Scott doesn’t do this work for charity and neither do I. If you’d like to offer up some funds for my work or convince Scott to give me a percentage of his advertising revenue for this article, I’d be more than happy to rework the numbers for you.

  43. Gotcha, so I assume these were still in play for the overclocked i5? I just figured it wouldn’t be that bad. The power draw bump is considerable. I’ve never been a watt-pincher on my desktop systems–just a curiosity for me.

    I haven’t done power draw tests on my C2D, but the speedstep CnQ stuff definitely affect temperatures even with my largish S1283 cooler, so I have always assumed it is doing a decent job of keeping power draw reasonable at idle. I may be giving it too much credit in my head.

  44. Pretty much. A standard user looks at performance in a range array of application because their system will be used for a wide range duties. But a professional cares about how their specific industrial level software will perform pretty much to the exclusion of everything else. Do you really think someone buying a workstation to run 3D Studio Max cares how well Crysis runs on it or how well it multi-tasks Office programs? But a person buying a general use computer will care how well it games as well as multi-tasks Office as well as how it handles lightweight video editing and light photo editing.

  45. I would think even a GTX260 is more power than anyone is likely to pair with some of the lower end CPUs, but not by much. I could easily see someone throwing a 5770 or 5750 at it. Overshooting a bit seems ok, but it seems useless to put a 5970 in a system which it will likely never be paired.

    Since the video card really only relates to game performance it seems reasonable to keep in in check.

    I’m certainly not hard line on the issue, it is arguable, but for a gamer wanting to spend money and make a specific purchase decision I think picking a more likely video card pairing gives the article far more context.

  46. That was the point, silly goose. It’s the ultimate example of how far we’ve come from high speed single-cores.

    Most of the recent CPU articles have had the 90nm Athlon 64 X2 6400+ and we know where that stands.

  47. There are all sorts of ways to rationalize it….but look at the graphs.

    I don’t like guessing.

  48. AMD should have had hyper threading for Opterons.

    But for PC CPUs, I seriously doubt they’re really losing out for not having it. Anyone who buys an Intel CPU with hyper threading probably would have bought the same thing even if Intel disabled hyper threading on the entire Core iX series.

    There’s not exactly much common application for it.

  49. damn. Intel is killing it with hyperthreading.

    AMD should have had hyperthreading ages ago.

    I have an athlon II x4, but the only reason I do is that it only cost me $50.

  50. Always love reading these; I’m curious to see how Westmere’s AES will affect TrueCrypt. A niggling typo from the Windows Live Movie Maker graph blurb:

    “The justification for a high-end processor is pretty clear: you’ll save over four and a half minutes by going with a r[

  51. Higher cost motherboards don’t boast better performance, they boast features: Overclocking fine-tuning, WIFI, more USB/USB3, SLI, quality of supercaps, etc.

    There hasn’t been a significant performance difference in motherboards among the same generation in many yearsg{<.<}g

  52. Here’s a benchmark I’d like to see, just to put things into perspective.

    Take a D510MO, and compare it to the Pentium 4 in applications that don’t use the GPU.

  53. people have a bad case of benchmark fever sometimes and get caught up in some differences that may be completely benign in 99% of cases. were not all tryign to run crazy setups. some of us wanna play games, like to support the underdog, and dont mind a 3 frame drop. ive owned both intel and amd systems, the only difference (as an ex hardcore gamer and now part time “between things” gamer) i see (notice) are the fanboys. different generations have offered different advantages to each side and most of them have been more than capable of playing modern decent games. except prescott.


  54. So the high-end user will look at specific applications but the general user won’t look at the specific application types they use?

  55. Yes, but the model I’ve proposed is one that’s built on common sense. Someone who is going to be looking at generalized performance ratings like “price to performance” is not going to be running high end apps, so base the performance rating on the apps they’re going to be using. The apps they’re going to be using are email, browser, Office, Quicken/Quickbooks/TurboTax, lots of games and maybe some light photo and video editing. Why skew performance numbers with high-end rendering apps and hard-core scientific modeling tests? They’re not going to be using them and are meaningless in that light.

    The person who’s going to be doing particle motion analysis or molecular modeling isn’t going to be looking at generalized “price to performance” ratios. They’re going to go straight to the chart that tests their field and base their decision on that. And in these cases, price to performance usually becomes nebulous since, as you say, “time is money” and slight variances in cost amount to nothing, especially when it can be written off. Conversely, the home or general user doesn’t make money from their computers, so cost is a basic “loss” and retains much more importance.

  56. Great choice on finally ditching the X48 S775 test platform.

    The high power consumption of the X48 board – 15 watts additional over G43 – really muddied the waters on power efficiency in every previous article.

  57. This and other posts miss and important point – these articles are to be interpreted on an individual basis. Personally, I consider it brilliant for a gamer to take an i5-whatever and overclock the snot out of it and match it with lots of RAM and a fast video card.

  58. Look at the variations in minimum FPS across the lower resolution settings in Borderlands.

    Then look at the 1600×1200 graph. What would 1920×1200 look like?

    Rather than argue opinions of the uses of different tests to no end, that’s all I’m going to say.

  59. agreed, and to boot it was a motherboard in line with the processors being tested, as opposed to test entry level and budget processors on high-end 790FX/GX motherboards that just add tons of extras (and the increased power load of said extras) to the tests.

    overall I liked very much the article, specially the part where these cheap quad (or pseudo-quad) processors really are a great value for the money overall.

  60. 2 crossfire cards needs a powerful CPU and yes Q6600 at 3.4GHZ is defiantly adequate.
    BTW, mine is Oced till 3.7GHZ and i am sure it’s more than adequate for today’s games… 🙂

  61. ATX cases are essentially free. If you haven’t found a free one you haven’t been looking. Add $25 for a decent power supply, and maybe $8 to swap out the fan in that for something quieter. Dremel and spray paint work optional.

  62. This is a bit of a guess but for the ‘system build price’ I am betting that Damage chose a card priced at about the best part of the price/performance curve. Anyone enthusiast enough to buy a higher-end card will know if wuch a card has more value for them in terms of fps.

  63. Hey thanks for the reply. I’m surprised someone saw it before I edited it, and reposted in the System Builders forum.

  64. This article was trying to measure CPU-performance not GPU performance. 260GTX was more than sufficient for the given task.

  65. I can’t see spending more than $80 on a motherboard. There is certainly no extra performance to be found in them. My old $50 G31 motherboard can run a Wolfdale at 4+ GHz just fine. I did splurge a little on a couple of 785G motherboards for my AMD systems–mainly to use the built-in video, since <$50 video cards are worthless again (long live the 9600GSO!)

  66. EDIT: Oops, sorry, I totally misread your post and said something that made no sense, but the rest of what I said stands:

    I wish they’d used a 5870, even though I’d never buy anything close to that. A GTX 260 isn’t necessarily enough to tell you which CPUs run games better at 1600×1200, and 1920×1200 is a bit higher yet, but still common place.

    I don’t think a 5970 is necessary, but lots of people will end up with a card somewhat comparable to a 5800 sooner or later, but using one of these existing CPUs.

  67. All Core iX CPUs drop their multiplier, voltage, and then some. Any Nehalem can shut off voltage, rather than just lower it, to individual cores, and I believe all 32nm Westmeres can do that to basically any part of the chip, on top of that.

    Desktop Core 2s don’t do anything but drop the multiplier to 6x. It’s amazing they do so well, but I imagine a lot of it has to do with using a rather low frequency/power front side bus, compared to the potentially excessive QPI and DMI links, and more powerful memory controllers, in Core iX CPUs.

    It’s the same deal with AMD. I think Athlon 64s actually used less power at idle, likely for those reasons.

  68. I completely agree. Additionally, I appreciate the use of a common video card rather than the “super ultra mega video card that not even enthusiasts actually buy” that is the standard for cpu reviews.

    The more I think about it the more I feel this article is the best thing since sliced bread.

  69. Scott – when viewing the comments there is no ‘quick link’ to the article in the article lead-in blurb.

    For all the people talking about this application vs that application (or application set) and various uses just put in a little effort. There is a chart for CPU ‘value’ alone, or if you are only interested in lower requirement programs just look at the benchmarks for those, or realize that for lower uses just about any modern dual core CPU is great.

  70. I agree that the “system price to performance” ratio doesn’t totally make sense in that people won’t pair a budget cpu with a $180 graphics card but there really isn’t a better option so you should simply look at the data differently. “processor price to performance” overvalues cheap processors and a “system price to performance” that has a unrealistically expensive system price undervalues cheap processors. These should be viewed as bounding scenarios.

    The only alternative I see is to run benchmarks comparing systems instead of individual components. For instance, running an abbreviated series of benchmarks on each system in the next TR system guide would be very interesting, but not at all the point of this article.

  71. Great review! One question, is there no dynamic multiplier drop on idle for the i5? I have my E8500 @ 4ghz (420×9.5), and while the locked voltage I still let the multiplier set itself down, meaning it will run at only 2.5ghz or so at idle (IIRC it can set the mulplier down to 6x). I forget the exact feature, C2something maybe? I know there has been some lack of ability to dynamically clock the i5/i7 line, but I’ve always appreciated it on my C2D.

    … And one critique:
    _[<"I'm growing tired of watching the GPU bottleneck take over, but ..."<]_ I really wish you didn't feel this way. I think your perception of your readers' views are perhaps a bit off. I found the DiRT2 benchmarks extremely insightful even if the GTX260 was showing some *[

  72. Thank you for using a 785G board, and using largely the same motherboards for compatible CPUs.

    It drives me absolutely nuts that sites run benchmarks on 790FXs or whatever nonsense that nobody I’ve ever known buys, and 40 different motherboards to directly compare CPUs.

  73. Like some of the other comments I think the low end value assumptions are wrong. I doubt many people are going to take a low end cpu and match it with a $160 graphics card. I would also suspect 4Gb of RAM would be suspect as well. You are likely looking at embedded graphics and 2GB of RAM on these systems. Also: if I am building something cheap I would skip on the case a little. $75 or less.

    I know this is more Art than science but the low end is web surfing, music and Word docs. Maybe a little photo and video editing. What I really want to know at that price point is can I do video editing without wanting to kill myself, not really how much faster it takes to render than the $2000 box.

  74. You know, as a hardware enthusiast I really do wish AMD had something with more “oomph” at this time.

    But as a computer science major, I gotta say that it’s incredible that AMD has gotten so much mileage out of refreshing the K8 microarchitecture.

  75. Although a detailed review, there are some points that don’t ‘rationalize’. The example by bdwilcox of the $60 processor going with the $180 GPU is a great point. As well as the high end industry specific Apps being tested on any Athalon or Core-i3 isn’t realistic. Just as the reverse is hard to comprehend about Grandma’s recipe program being run on Core-i7 Extreme! A more ‘focused’ CPU-to-Industry task is something that would be welcomed. Such as a graphics designer must get 3 professional Apps and a PC (Software & Monitor as well) to run them on. And it all has to be done for under $4k.

    Bench the AMD system and the Intel system that can be configured for this $4k budget…

    Does one get SSD & SAS? Does the other get SSD and SATA? Does one get 16GB of DDR3-1600 and the other 6GB DDR3-1333? Does one get an ATI 5830 Radeon? The other gets a FirePro V8750? How about a workstation class HBA/Raid Controller? The other may already be at the $4k limit and cant factor in that performance increase in the overall system.

    Things get more *real* when a budget is introduced. Bench the end result of the configurations.

    Also, just a note of one of the justifications for Scott to go for the Core i-3 system was the overclockability of the Core i-3. However that gives up the ‘low power’ consumption title by 30w if overclocked at 4.4ghz to the Athalon’s overclock of 3.4Ghz.

  76. I have a Q6600 oc’ed to 3.43ghz and it powers 2 5870’s in crossfire on and x48 board quite comfortably. It would be interesting to see how it fares with the rest of the bunch while overclocked. Overall i think its more than adequate!

  77. Have an E8500, and I’m lazy. Rebuilds seem more of a hassle than something to get excited about. Will wait another year I thinkg{<.<}g

  78. Yeah I have a Q6600 that sits at 3.2 GHz on stock voltage (1.25v). I also have a Phenom II X4 3.0 GHz around and it’s performance is effectively the same, if slower in some apps. It doesn’t overclock very well as 3.2 GHz is all it can do on stock voltage (1.3v).

    I think turbo boost is great stuff considering how most apps barely use even dual cores yet so I hope that continues to evolve and that AMD dupes it.

  79. (bdwilcox beats head against wall)
    Let me clarify again. The people who care about a general measurement like a “price to performance ratio” usually don’t care about performance in professional apps, so don’t average them in (as they will skew the results). Average in the performance numbers for what they would care about.

  80. There are dozens of ways you could build the graphs to ‘fit’ a certain buyer/user profile. No amount of ‘Liberal arts guilt’ is going to get you all of them!

    Great article, Scott.

  81. I could run some benchmarks on my two AMD systems, but I know they run pretty quick considering the other hardware installed with them.

    From a reliability/stability standpoint, they’re all the same. AMD and Intel servers at work only go down for patching, and my AMD systems at home are the same way.

    I agree, if you’re having stability issues with any system, you need to look at the specific motherboard model you have and drivers and whatnot. The specific chipset you have shouldn’t be the issue.

  82. I agree with bdwilcox that the system price-performance metric was skewed- especially by the selection of LGA755 mobo. Good LGA755 motherboards are available for around $50 these days, and getting one which was *more* expensive than the i5 motherboard is messed up. Even if you insist on putting fancy Gigabyte USB3 motherboards on all these machines for comparison, go with the GA-EP43T-USB3 at >$40 less than the mobo you listed.

  83. The point I was trying to make is that including high-end and professional benchmarks in a general “price to performance” ratio skews the results. If
    you’re looking at general price to performance ratings, you’re probably not going to be running pro-level apps and, conversely, if you’re going to be running pro-level apps, you’re not going to care so much about general performance. So for a general “price to performance” ratio, factor in the apps that matter to the typical user and leave out the hardcore pro apps.

  84. Excellent article, I know a lot of us have been asking for the inclusion of older processors and I was happy to see the P4. It’s nice to know what kind of performance boost I’ll see when upgrading my parents old computer. I was also glad to see the performance per total system price as this makes more sense than just including the processor price.

    I was actually a bit surprised that the P4 performed as well as it did. Though I guess 3.8ghz and HT help mitigate netburst’s other inadequacies.

    Also I would recommend you include the “…and everything else” subtitle as part of the main title that shows up on the front page. When looking through the archive relatively nondescript titles like “Core i3 takes on Athlon II” don’t scream *[

  85. I would also like to see some comparisons on these cpus, although the 910e should be exactly the same as the “normal” 910.

  86. Actually I think the AMD system used for testing (page 3) is a pretty good overall system for the price points being used, it makes much more sense to have that system than the “value” one (page 17).
    As for the 775 system I would use the same as the value or a G45 based one.
    The rest of the components on the value system, a part from the 5770 which I would probably swap for the 5750 because, like it was noticed, if you just want to play at 1680×1050 any midrange card will give you enough fps to be happy.
    And I have to agree with #40, although USB3 makes more sense than SATA3, I wouldn’t pay more for any of those right now (or in the current year) there isn’t much of a need for USB3 a part from some hdd cases and I already have e-sata on my 780G motherboard, as for SATA3, that even makes less sense seeing the kind of throughout up we have on devices these days.

  87. I’d really enjoy seeing some proof of this claim, as many people in my Lan Club at college absolutely love their AMD chipsets and see equal performance to any intel variant, especially with SB750

  88. One thing that I actually got out of this test was a bit of a refreshed view of the Phenom II 965, it appears to do better in this slew of tests compared to Intel than when mentioned in its own article.

  89. Yep this pretty much applies to any quad core from the last generation and a half from intel. I know my overclocked Q9400 is about the same as derfunks Phenom II overclocked.

    Intel would be some amount more ahead with todays higher end chips with an overclock but it really just isn’t enough to matter.

  90. I don’t understand the selection of USB3+SATA3 motherboards for everything except the LGA1156 P55 platform. Why not go with the GA-P55A-UD3 or P7P55D-E for this socket?

  91. Agreed. The Core i7-860 should be a better value than the i7-870. I also believe that the AMD Phenom II X4 955 Black Edition offers a slightly better value than the X4 965 included in the review.

  92. I’m pretty sure it’s not a ‘feature’ nor is it particular to Asus but rather just what happens when you leave voltages on Auto and set a higher clockspeed.

  93. I kind of agree on your first point, however you have to set some baseline to be able to compare platform costs at all. If you’re going to budget the rest of the parts to match the ‘grade’ of the CPU, you can just go ahead and base your decision on the first price/performance scatterplot, assuming you’ll spec the other components to be relatively as cheap/expensive as the CPU is.

    As for your second point, sure TR could slice the data up another way and add even more graphs, but aren’t there enough already? It would be awesome to be able to use their data and mix/match your own particular brand of coloured bars and points, but that would take considerable effort on TRs part.

  94. what is this feature on the ASUS board that is adjusting voltages?

    and does it still increase the voltages if they are inadequate even after you have manually set the voltage higher already?

  95. Only if you run the upgrade at stock speeds. If you overclock a Core i5 750 to 3.5+ GHz, I think you’d notice the difference.

  96. And again. All in all a very nice comparison, if there wasn’t the one things that spoils it for me (= for me personally): All XXXe CPUs from AMD are left out beside the power draw results for the new PII 910e.
    Question here is: why do this? The 910e (similar to its predecessor the 905e) are the most expensive desktop CPUs AMD is selling at the moment.

    The Athlon 600e and 605e are very interesting HTCP-CPUs with their 45 Watt TDP, as are the PII 9XXe with their 65 Watt TDP (the latter suited for silent workstations as well).

    I would have loved to see how the 910e fares against a Athlon 605e.

  97. That’s because Nehalem is in many ways a Core 2 with an IMC and other additional goodies like hyperthreading and Turboboost. I’m talking about architectural things such as # instructions per clock (I don’t mean effective IPC there) and whatnot. In much the same way Phenom I/II are A64s on steroids, or maybe there’s a better comparison for them. The next ‘revolution’ from Intel will be Sandy Bridge, for AMD it will be Bulldozer.

  98. Good CPU’s for sure, but it’s more evolutionary rather than revolutionary which c2d was.

    An o/c e6600 still works well with most programs/games.

  99. The price premium for the c2d 8600 over the 8400 was very steep. An OCed 8400 was the real value.

  100. I have to agree with you, my Q6600 at 3.7GHZ would compete very well with today’s CPU… Today i am sure my decision to buy Q6600 was very wise…

  101. A couple things that bothered me:

    First, the “system price to performance ratio” is a bit biased. If you’re going with a value processor, you’re most likely going to go with a value motherboard, video card, and case. I don’t know anyone that would match a $120 motherboard, a $130 case, and a $180 video card with a $60 processor. I just built a cheapo system with a $12 Rosewill case (Newegg Shell Shocker), a $50 video card, a $55 motherboard and a $60 CPU (Newegg combo). That’s more like what you’d see with people buying Athlon X2s or Intel Pentiums.

    Second, I think factoring all the benchmarks into the “price to performance” ratio is a bit misleading, especially the inclusion of benchmarks of pro-level applications. A general “price to performance” measure is more important to home users than businesses since most businesses won’t be pinching pennies like a home user would. Plus, businesses would look at the benchmark that affects their area of expertise (rendering, video editing, photo editing, etc.) and make a decision on that benchmark. Thus, it’s probably more practical to include only those benchmarks that home users would face most often: Office, browser, ZIP, games, maybe some Photoshop, maybe some video editing. But why include scientific benchmarks, folding and high-level professional apps like hard-core 3D rendering? The pros will look at those benchmarks and make a decision independent of the other benchmarks whereas the home user wants to see what will cover all their general activities.

  102. lol at the poor P4 in the games you did test. besides that abomination, all the other cpus are pretty good so of course the results are about what should be expected for gaming. its not like you were using some old A64 X2 cpus or Pentium D cpus as those would certainly struggle a bit in many newer games.

    of course there are also games you didnt use like Far Cry 2, GTA 4, Prototype, ARMA 2, Red Faction Guerrilla, Ghostbusters(dont laugh) and a few others that would really have the faster cpus, especially quads, showing their worth.

  103. I know and read a number of users who are quite happy with current batch of AMD chipsets, despite the known issues with southbridge.

    Current batch of chipsets from every vendor are all robust and stable. Problems at that level now reside in motherboard implementation.

  104. That’s what they all say., I’m still not buying into AMD chipsets are just as good as Intel. That’s what they told me last time and guess what? it’s bogus. Nvidia chipsets? No thanks!

    Anywho, Scott I would really like it if you put some quad+ optimized games in the mix instead of left for dead. That engine does not need anymore benchmarking from modern CPU’s.

    Say GTA4 and Dragon Age? Both quad+ optimized. From what I’ve seen though i3 does well in both. Dragon age and GTA4 dips so hard with my 2 thread 4ghz overclocked CPU. Mostly around the 20’s.

    I’m pretty much sold on getting an i3 for now i7 860 later option. I just need to get some money.

  105. Look at the numbers and tests again.

    Most mainstream applications do not take advantage of extra power that Bloomfields offer. You only see that power in applications where time is $$$$$. The more affordable quad-core/dual-core options are still able to deliver a smooth experience.

    Even that said, the jump from Core 2 and Bloomfield wasn’t as large as the jump from Pentium 4 to Core 2.

  106. Damage – great review, much thanks for the inclusion of the TrueCrypt test. I am glad you’re going to revisit it once the code has been changed for the AES instructions. We are seriously looking at that for work applications (single purpose processors being a waste of space).

    The only suggestion I would make is finding a good engineering graduate school to make up for that liberal arts degree! You’ve certainly got the tinkering / testing to exhaustion part down cold.

  107. That is a very nice written article. I loved the new incremental performance over system price chart, it put across the value very accurately.

    Like obarthelemy, can we have some multi-tasking test, a lot of us are recommending systems to people who just use the PC for regular usage. But a lot of that regular usage involves 5-6 browser windows, a couple of office applications open, as well as the music player playing in the background. Is it possible to simulate the test? I know it would be in-human for the entire range of CPU’s but you can show a comparative between the mainstream competition (X4 635 vs i3 530 for eg)

    In addition the power draw test will hit the bulls-eye if you can convert it into $’s saved per month/week etc. Once you get the metrics right, it will just be an excel sheet extension. You can ask the readers to map their CPU usage per week (there are a lot off apps which have charts, Rivatuner comes to mind. Based on an average you can determine the “work” watt levels (to determine how long is a system idel or at loads), multiply it with the average electricity rates in the US, and you would get a cost of running a system per week.
    However do keep in mind, energy costs goes beyond the generation costs. You do have a replacement costs (the cost of cleaning up the environment while generating electricity), but you get the hint…

  108. That overall performance vs price graph makes me even more excited about getting my i7-860 for the price of the i5-750 at Micro Center. xD

    After some quick and dirty math based only on the 4.4% difference in clock speed of the 860 vs 870, it comes out to ~.49 haha.

  109. What’s wrong AMD chipsets?

    The only problem with them is AHCI and USB performance is a little inferior to Intel southbridges.

    Otherwise, they are pretty much the same.

    SIS isn’t that bad either, if you want something that just works with no-frills.

    The days of shoddy chipsets are long gone. I still shudder at early Via and ALI chipsets.

    I forgot to mention Nvidia chipsets. They usually work, but always had some sort of stupid, stupid issue.

  110. ?

    AMD makes their own chipsets. Intel makes their own chipsets. And NVidia makes chipsets for both of them.

  111. I blame the PCI bus on older systems.

    There are a lot of more devices tied to it and they had to share ~133MBs of bandwidth. You can easily see why things can get congested.

    Since, systems and integrated devices have been moving away from PCI bus to PCIe lanes. There has been considerable improves in overall I/O performance. It helps when each lane has more bandwidth than a PCI bus (250MBs) and that bandwidth is dedicated to that lane.

  112. That was an excellent article. I loved seeing the P4, and the value calculations were great.

  113. Thanks for a very nice article, it’s a delight to see a wide range procs, and meaningful info on perf/$ and power/$. This article is very useful in recommending and buying systems.

    2 suggestions:
    – a multi-tasking+I/O test. I don’t know if it’s still the case, but a few years back things used to get very jery when I was doing file I/O + USB I/O + netwrok I/O at the same time
    – interactive recap charts that let us change the weight of the different parameters: gaming system, productivity system, content creation system.

    But that would be icing on the cake.

  114. Thanks for the hard work Damage. You deserve a long rest. 😉

    It is insane that you can have a stupid amount of performance for an affordable price tag. Quad-cores and fast dual-cores for the masses.

    Netburst was a such an inefficient architecture. Even the mightiest example of it gets completely destroyed by modern chips. FYI, P4 670 had the “highest” stock clock speed in this round-up. Although, the turbo-clocked Clarkdales and Lynnfield likely surpass it in single-threaded loads.

  115. Q6600 managed to hold itself up despite its age.

    IMO, there hasn’t been improvement in CPU arena since Core 2 introduction.

    Any of current CPUs only make sense if you had skipped the Core 2 bandwagon or time is $$$$.

    Its funny that it wasn’t that ago long when quad-cores were considered to be ultra-high end along with the price tag. They are quite affordable now.

  116. How good is/was the Q6600?

    If you accept the proposition that general users really need a 50% performance increase to notice the difference, then you’d have to go to an i860, i920 to notice the upgrade from the Q6600. But given where the Q6600 appears in the performance part of that graph, I’d say the use of multi-threading in general software has come a long way from when the Q6600 was released. And if you’ve overclocked the beasty from 2.4Ghz to 3.0/3.2 GHz, then only the application of serious $$$ is going to get you a noticeable upgrade.

  117. Is it just me or do your projected points per day for fah not make any real worlds sense? i7920’s routinely get around 12,000 points per day. Am i reading that wrong?