Palomino finally gallops onto the desktop
The Athlon XP is a revised version of the Athlon based on a core design code-named “Palomino.” We’ve seen Palomino previews in a number of placesin laptop CPUs, in the Athlon MP server/workstation chips, and in little brother “Morgan”but the Athlon XP is the first Pally aimed at the desktop.
The Palomino core packs in a number of enhancements designed to improve the Athlon’s performance, and scalability. Among them:
- Lower power requirements Thanks to some changes in the way the chip is made, the Palomino requires about 20% less power than corresponding Thunderbird chips. As you might expect, that means the Palomino runs quite a bit cooler than the T-bird, as well. (For this reason, the Palomino first hit the market as a mobile processor for notebook computers, in the form of the oddly named “Athlon 4” processor.)
Despite the changes, the Palomino is still made on the same 0.18-micron copper fab process as current Athlons. Intel will soon deliver a die-shrunk Pentium 4, but it will take a while for AMD to make the conversion to 0.13 microns. AMD likes to point out that even a 0.13-micron Pentium 4 has a larger die size than a 0.18-micron Athlon XP.
- An on-chip thermal diode Like Intel’s Pentium III and 4 processors, the Palomino core includes an on-chip thermal diode for temperature monitoring and better power management.
- Hardware pre-fetch Performance-wise, this may be the most important addition to the Palomino. The hardware pre-fetch logic attempts to anticipate what data will be needed from main memory next and preemptively loads this data into the processor’s L1 cache. This enhancement, which is similar to logic present in Intel’s Pentium 4 and Pentium III “Tualatin” processors, should allow the Athlon to take better advantage of the extra memory bandwidth available with advanced forms of memory like DDR SDRAM.
- Improved translation look-aside buffers (TLB) Though more esoteric than hardware pre-fetch, improved TLBs ought to complement data pre-fetch logic nicely. The Palomino’s revamped TLB structures are now larger, speculative (as one would expect, with hardware pre-fetch now in the picture), and exclusive (no longer shared) between caches. These improved TLBs should help keep the Athlon’s pipeline fed, increasing clock-for-clock performance.
- SSE compatibility AMD says the Palomino includes 52 new instructions that comprise something called “3DNow! Professional.” These instructions just happen to correspond to the instructions the original Athlon needed in order to be compatible with Intel’s SSE, or streaming SIMD extensions. These extra instructions do not provide compatibility with the Pentium 4’s new SSE2 instructions, but they should yield improved performance in applications optimized for SSE but not for AMD’s competing 3DNow! extensions.
- More transistors To accommodate these new features, the Palomino weighs in at about half a million more transistors than the Athlon Thunderbird, up from 37 million to 37.5 million.
Cumulatively, these improvements should make the Athlon XP quite a bit better than its predecessor. The T-bird had no significant performance weaknesses to speak of, except that it wasn’t entirely able to make use of the extra bandwidth provided by DDR memory. Palomino ought to remedy that weakness and add some new strengths.
Beyond the internal changes, the Athlon XP comes in a new package. It looks like so:
The Athlon XP’s new outfit is an organic pin grid array (OPGA) package. This package has a fiberglass substrate, like a printed circuit boardand quite similar to the packaging on Intel chips. AMD claims the new organic packaging is cheaper to produce and offers lower impedance than the ceramic packaging they’ve used in the past.
It’s also quite brown.
I’m hopeful the new packaging will flex and bend ever so slightly, preventing the disasters with cracked or chipped cores we’ve seen with Athlons in the past. However, the Athlon XP’s core is just as exposed and vulnerable as ever. I had hoped to see a metal cap (“Integrated Heat Spreader”) a la the Pentium 4, but no such luck.
AMD’s new math
The biggest news with the Athlon XP may not be the technology, but how AMD plans to market it. We have dwelt at some length here at TR on the issue of clock speeds and performance, from the great Mac Wars to our Pentium 4 2GHz review, where I offered a lengthy explanation of how MHz has mattered in the PC market in the past, and how the game has changed with the Pentium 4.
More MHz myth madness
The long and the short of it is this: a processor’s clock speedmeasured in MHz and GHzisn’t a reliable indicator, all by itself, of performance. We’ve seen that time and time again, as Athlons at 1.2 and 1.4GHz have handed higher-clocked Pentium 4 processors their heads. On a platter. The Pentium 4’s NetBurst microarchitecture simply does less work per clock cycle than the Athlon. That fact doesn’t necessarily reflect poorly on the P4 or on the Athlon, it’s just the way things are. An Athlon Thunderbird at 1.2GHz is roughly equivalent, performance-wise, to a 1.7GHz Pentium 4. A 1.4GHz Athlon runs neck-and-neck with a 2GHz P4.
Trouble is, the Pentium 4’s super-high clock speeds tend to look mighty appealing on the features list of a brand-new PC. Given the choice between a 1.7GHz Intel and a 1.2GHz AMD at about the same price, the mythical Joe Sixpack is probably gonna opt for the 1.7GHz Intel. That uncomfortable fact threatens to become a marketing nightmare for AMD. Few companies would want to face the daunting prospect of explaining to Joe Sixpack why their 1.2GHz system is faster than the other guys’ 1.7GHz box.
Apple tried it, but it didn’t work especially well for themand people already knew Macs were weird to begin with.
AMD’s solutioncheck that, AMD’s coping mechanismis to try another spin at a well-worn tactic from the bad old days, when AMD and Cyrix CPUs played a sad second fiddle to Intel’s: the Pentium Rating. Well, it’s not exactly the Pentium Rating, but it is this: assigning a model number to a CPU based on relative performance rather than clock speed.
When we first caught wind of AMD’s plans on this front, Dissonance teed off. He wrote:
While Intel has been able to capitalize on Joe Sixpack’s love of high numbers, AMD’s naming scheme would only manipulate it. AMD isn’t lying outright, but there’s an air of deception when one defies industry standards that even the MHz-crippled Mac adheres to. If AMD’s naming scheme didn’t so closely resemble the MHz values posted by Intel, then it could be forgiven here. However, using model numbers similar in value to MHz, in an attempt to compare performance with Intel, only does the consumer a disservice by failing to disclose the processor’s actual speed.
Techies don’t tend to like marketing types and their plans anyhow, but when a hint of deception is in the air, we close in like a pack of dogs.
Truth be told, though, AMD’s plans aren’t quite as sinister as all that. We should start by acknowledging that AMD is actually in quite the bind here, and there’s not a simple, easy, obvious answer to the dilemma. Much as we’d like to think folks could be educated to understand why clock speeds aren’t a reliable indicator of performance, the fact is that they’ve been taught to believe in MHz as such an indicator for the past 20 years. AMD’s success on this front could only be partial at best.
Given the constraints, AMD’s solution is actually fairly savvy, even if it is imperfect. And there’s a little more to it than the initial rumors suggested.
AMD will indeed be offering the Athlon XP in a range of models, including the current flagship, the Athlon XP 1800+. These model numbers will be assigned to different frequency processors like so:
The “plus” after each number supposedly denotes that the Athlon XP is actually faster than competing chips at 1.8GHz. But don’t take my word for it. For the best example of how AMD’s new marketing scheme works, take a long, hard gander at this sample retail PC price tag:
This price tag tells it all. Note a couple of things. First, the CPU model number is prominently displayed, but the clock speed isn’t entirely hidden, either. Instead, there’s an asterisk next to “QuantiSpeed architecture” that points to the fine print, where the clock speed is listed. This scheme is actually quite similar to what some of our readers suggested in the great debate over this issue. It’s a small step, but an important one, to disclose the actual clock rate of the processor. Doing so avoids the appearance of deception.
Second, there’s this whole “QuantiSpeed” name. This name is the latest in a time-honored tradition of questionable names assigned to technologies by marketing departments. Such names usually signify a technology or a collection of features that may or may not be related to one anotheror to reality. These names are often chosen to evoke something specific that a marketer wants to communicate about a product. In this case, QuantiSpeed is the name AMD has given to the Palomino architecture, and it’s clearly intended to evoke, in a sense, that its time is more valuablethat it delivers more work per clock.
Before the Intel fanboys start hooting, they should pause to consider Intel’s own use of loopy marketing terms. NetBurst, anyone? That one is chosen to correspond with Intel’s whole “makes the Internet go faster” line, which is a load of bunk. Other Pentium 4 marketing terms? How about these: Advanced transfer cache, Hyper pipelined technology, Rapid execution engine, and Advanced dynamic execution.
No serious chip geek is going to look under the hood and think, “Oh, it’s got a Rapid Execution Engine,” but the terms do serve an important marketing purposemuch as I hate to admit it.
That sample price tag speaks volumes about where AMD is going with this thing. I think the few simple twists they’ve added are quite clever. And there’s one more twist.
Where the numbers come from
There is the question of how AMD arrives at a number to assign to its new processors that will accurately reflect how the Athlon XP stacks up to the competition. If that estimate is too high, as so many of the old Pentium Rating numbers employed by Cyrix were, it will leave folks with a bad taste in their mouths. If the estimate is too low, AMD risks losing any competitive edge they’d hoped to gain with this numbering scheme. Overshoot or undershoot and AMD’s customers and partners will be unhappy.
To avoid comparing themselves directly to the Pentium 4, AMD decided to make the model numbers “convey performance relative to other AMD Athlon processors.” In other words, they’re using Athlon Thunderbirds as a basis for comparison, not Pentium 4s. This move neatly sidesteps the “Pentium Rating” stigma. It also makes for a relatively conservative performance estimate for each model number, as you’ll see in our test results below.
But how does AMD arrive at these numbers? For the future, the company has big plans. AMD is promising “an industry-wide initiative to develop a new and more complete measure of performance that end-users can trust.” This new performance metric is due for rollout in 2002. When we asked, AMD had no word yet on whether a sanctions body would be formed to sanction the new metric. They also couldn’t tell us much about the likely scope of the metricwhether it would take into account performance across multiple operating systems, or whether the metric could be used to compare Pentiums and Athlons with non-x86 chips. Obviously, this effort is still in the early formative stages.
Until that new standard arrives, AMD has cooked up its own set of benchmarks to use. AMD calls the results from this suite of tests a “bridge metric.” It is intentionally and overtly temporary (at least for now; these things tend to grow roots). This performance metric is comprised of 14 different benchmarks which make use code from 34 separate applications. Most of the component tests will be familiar to TR readers; they’re tests like SysMark, Business Winstone, Content Creation Winstone, 3DMark, Expendable, and Quake III Arena. AMD has divided these tests into three categoriesvisual computing, gaming, and office productivityeach of which is weighted equally. Inside each category, each sub-test is weighted equally, as well.
The list is notable for its lack of tests using Linux or other non-Windows operating systems. We here at TR don’t get the chance to test in Linux as much as we’d like simply because of limits on our time and resources. AMD’s slight is a little more significant because they’re an x86 CPU manufacturer, and because of the way this metric will be put to use. I’m sure Linux fans will have a thing or two to say about the snub, but then again, Linux users are more likely to understand the disjunction between clock speeds and overall performance.
Model numbers in action
To get a sense for how the model number scheme will work in practice, have a look at the BIOS boot screen pictures below. As you can see, the BIOS prominently displays the model number without showing the clock speed. I decided to have some fun with this setup, so I overclocked the Athlon XP to 1610MHz on a 140MHz bus (yep, it was stable there at the default voltage). At boot time, the model name came up as “1900+”. So I decided to really turn the crank, up to 1645MHz. At that speed, the BIOS just threw up its hands and reported a clock speed.
As AMD releases new processor speeds, BIOS updates will probably be needed in order to keep the model numbers up to date.
Once in Windows, utilities like WCPUID show the processor clock speed just fine, and the BIOS passes along the name string to the utility, as well. Like so:
Overclocked, the story is the same as in the BIOS boot screen:
Hiding the processor’s actual clock speed in the boot screen seems a little heavy handed, but I’m not sure how else AMD can avoid confusion. Joe Sixpack is going to want to see that 1800+ number once he’s shelled out the cash for an 1800+ system.
Anyhow, that’s about it for AMD’s new model numbering methods. Love it or hate it, that’s how it works. Let’s move on to the benchmarks see how the Athlon XP really performs.
Our testing methods
We decided to give our tried-and-true test suite one more run before putting it out to pasture in favor of Windows XP, new drivers, and newer applications. AMD shipped its eval systems with Windows XP installed and encouraged testing in XP, but the OS isn’t widely available yet, so we decided to wait a little longer before switching. We abandoned Win98 testing quite a while ago, and Win2K should produce comparative benchmark results very similar to Windows XP
Going back to the well one more time also allowed us to include comparative data for a broad range of systems, from a lowly Celeron 900MHz to the fastest Pentium 4.
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.
Our test systems were configured like so:
|Socket A (DDR)||Socket A (PC133)||Socket 423||Socket 478||Socket 370|
|Processor||AMD Athlon 1.2GHz
AMD Athlon MP 1.2GHz
AMD Athlon 1.4GHz
AMD Athlon XP 1800+ 1.53GHz
|AMD Athlon 1GHz
AMD Duron 1GHz
AMD Duron 1.1GHz
|Intel Pentium 4 1.4GHz
Intel Pentium 4 1.6GHz
Intel Pentium 4 1.8GHz
|Intel Pentium 4 2GHz||Intel Celeron 900MHz
Intel Celeron 1.1GHz
Intel Pentium III 1.2GHz
|Front-side bus||133MHz (266MHz DDR)||100MHz (200MHz DDR)||100MHz (400MHz quad-pumped)||100MHz (400MHz quad-pumped)||100MHz (Celeron)
|Motherboard||Gigabyte GA-7DX rev. 4.0||Asus A7VI-VM||Intel D850GB||Intel D850MD||Intel D815EEA2|
|Chipset||AMD 760/VIA hybrid||VIA KM133||Intel 850||Intel 850||Intel 815EP|
|North bridge||AMD 761||VIA VT8365||82850 MCH||82850 MCH||82815 MCH|
|South bridge||VIA VT82C686B||VIA VT8231||82801BA ICH2||82801BA ICH2||82801BA ICH2|
|Memory size||256MB (1 DIMM)||256MB (1 DIMM)||256MB (2 RIMMs)||256MB (2 RIMMs)||256MB (1 DIMM)|
|Memory type||Micron PC2100 DDR SDRAM CAS 2||Infineon PC133 SDRAM CAS 2||Samsung PC800 Rambus DRAM||Samsung PC800 Rambus DRAM||Infineon PC133 SDRAM CAS 2|
|Graphics||NVIDIA GeForce3 64MB (12.41 video drivers)|
|Sound||Creative SoundBlaster Live!|
|Storage||IBM 75GXP 30.5GB 7200RPM ATA/100 hard drive|
|OS||Microsoft Windows 2000 Professional|
|OS updates||Windows 2000 Service Pack 2, Direct X 8.0a|
The test systems’ Windows desktops were set at 1024×768 in 32-bit color at a 75Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- SiSoft Sandra Standard 2001.3.7.50
- Compiled binary of C Linpack port from Ace’s Hardware
- ZD Media Business Winstone 2001 1.0.1
- ZD Media Content Creation Winstone 2001 1.0.1
- POV-Ray for Windows version 3.1g (multiple compiles)
- Sphinx 3.3
- ScienceMark 1.0
- SPECviewperf 6.1.2
- MadOnion 3DMark 2001 Build 200
- Quake III Arena 1.17
- Serious Sam v1.02
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
What to watch for in our test results
There are a number of things we’ll want to watch out for when looking through the test results. Naturally, the most important question is: “Which processor is fastest?” Last we tested, the Pentium 4 2GHz just barely came out king of the hill. The Athlon XP has a good shot to regain the top spot for AMD.
Beyond that, we’ve changed the playing field a little bit by giving our Athlon testbed CAS 2 memory instead of the CAS 2.5 DIMM we used before. It’s possible this minor change, all by itself, could shift the balance of power. The T-bird Athlon 1.4GHz could potentially overtake the Pentium 4 2GHz now, leaving the 2GHz P4 in the number-three spot.
We’ll also want to see how the Palomino core compares to the T-bird. To make that comparison possible at the same clock speed, we’ve tested a T-bird Athlon and an Athlon MP processor (Palomino), both at 1.2GHz. The Thunderbird versus Palomino question is especially interesting because AMD’s new ratings system is supposedly based on how the Athlon XP compares to previous Athlon processors, as we mentioned above.
Similarly, we’ll want to watch how the Athlon XP 1800+ stacks up against AMD’s real target, the P4 1.8GHz. AMD claims the “plus” denotes performance superior to a competing 1.8GHz processor, and by that they mean the Pentium 4. Is AMD’s estimate too optimistic? Too conservative? Or did they somehow manage to land at just the right balance?
As always, these memory tests are for exhibition only; they don’t denote application-level performance, just the real-world ability to move memory efficiently. On that front, we’re expecting the Athlon XP to be quite a bit more capable than the Athlon Thunderbird, thanks to the XP’s hardware pre-fetch. Let’s see if our expectations are met.
Both of the Palomino processorsthe Athlon XP and Athlon MPare faster than the T-birds, particularly in the integer tests. That fact bodes well for the Athlon XP; the processor can likely put its DDR memory to good use, as predicted.
Nevertheless, the Pentium 4 with RDRAM rules this test, as usual. Let’s use Linpack to dig under the surface a little bit.
Linpack graphs aren’t the easiest to read. Take a second to look at the axis labels, though, and you’ll get it. We’re measuring processing throughput here, in megaflops, for data matrices of different sizes. The small matrices, on the left half of the graph, fit into the processors’ L1 and L2 caches, so processing throughput is high. On the right half of the graph, where the matrices are too large to fit into the caches, performance drops. It’s there, at the larger data sizes, where we get a better sense of main system memory bandwidth. Overall, the shape of the graph gives us a nice visual picture of how a system’s tiered memory architecture performs.
The Pentium 4 reaches its peak at about 192K, when everything fits into its L1 and L2 caches, and the numbers are crunching. Its peak performance is just under the 1.4GHz Athlon’s. When it has to go to main memory, once the matrices get bigger than about 256K (in the right half of the graph), the P4 doesn’t drop off nearly as much as the others. That’s because the P4’s very fast accessing its main memory, and its dual RDRAM channels deliver gobs of bandwidth. All in all, it’s an impressive performance.
Unlike the Pentium 4, the Athlon XP’s L2 cache doesn’t duplicate data stored in its L1 data cache. As a result, the Athlon’s effective total cache size is larger. The Athlon XP doesn’t really drop off going to main memory until we hit about 320K. However, the Athlon XP’s L2 cache is a tad slower than the P4’s; you can see, at about 64K, where the Athlon XP has to start moving beyond its L1 data cache into it slower L2 cache, and the MFLOPs drop off.
When it comes time to go out to main memory, the Athlon XP easily outruns the T-bird. That’s hardware pre-fetch in action.
Business Winstone 2001
Business Winstone tests performance in general office applications, like word processors, spreadsheets, and web browsers. This test is also usually a good indicator of overall system performance, especially where “light use” patterns are concerned (in other words, not gaming, heavy computation, or multimedia).
The Athlon XP 1800+ shows its stuff here, taking a commanding lead in Business Winstone. The Athlon 1.4GHz pulls ahead of the Pentium 4 2GHz, too, thanks to faster CAS 2 memory this time around.
Content Creation Winstone 2001
Content Creation Winstone is arguably more important than Business Winstone, since it tests more performance-sensitive apps, like image and audio processing suites, desktop publishing, web layout programs, and the like. As with Business Winstone, the test runs scripts using code from real applications, not just generic simulations.
In this performance-sensitive test, the Athlon XP 1800+ delivers a pretty serious whuppin’ to the competitionover six points ahead of the 2GHz Pentium 4. That 1800+ model number is already starting to look mighty modest.
POV-Ray 3D rendering
POV-Ray is a freeware software ray-tracing program that creates high-quality 3D scenes. It’s also a very useful measure of a processor’s performance, particularly on floating-point math. Our POV-Ray tests use the original release of POV-Ray 3.1, plus Steve Schmitt’s recompiled versions, just to see what difference the various compilers and compiler settings can make.
The recompiled POV-Ray comes in two flavors: “PIII” and “P4”. Both were produced with Intel C v. 5.0. The “PIII” version doesn’t use any instructions proprietary to Intel processors or to the PIII; it runs just fine on the Athlon and the P4. The “P4” version uses a small bit of SSE2 code, but it doesn’t take advantage of the P4’s SIMD capabilities. I’ve indicated which version of POV-Ray was used in the graphs below next to the processor/speed labels, so it should be easy to track.
Also, because of time constraints, I’ve omitted results for the value processors here.
The Athlon XP continues the Athlon’s tradition of handling older, legacy code well. Here, the original compiled version of POV-Ray is markedly slower, but the Athlon XP 1800+ manages to render the scene about 24 seconds faster than the Pentium 4 does. With the newer, recompiled versions, the Pentium 4 puts in a respectable showing, but the Athlon XP still comes out on top.
The chess2 scene is quite a bit more intensive than our first POV-Ray scene. Objects in chess2 reflect light realistically through ray tracing, and as a result, the scene takes much longer to render, even at the same resolution. And here the Athlon XP does even better, as all the faster Athlons bunch up near the top of the chart. The Athlon’s superscalar, fully pipelined floating point unit is nearly a force of nature.
LAME MP3 encoding
LAME is the encoder of choice around Damage Labs for high-quality output, so this test holds some interest for me. More speed for MP3 encoding is always good.
You were expecting something else? 🙂
Quake III Arena
Now we’re on the Pentium 4’s home turf. The P4 just loves Quake III Arena. Can the Palomino enhancements make the Athlon competitive again here?
For once, the 1800+ model name doesn’t look like an underestimate. Still, the Athlon XP has moved much closer to the Pentium 4 in Quake III performance. With a little more memory bandwidth than the 760 chipset can provide, the Athlon XP 1800+ might be running neck-and-neck with the Pentium 4 2GHz.
Notice, also, how the 1.2GHz Athlon MP manages a 10 frame per second lead over the 1.2GHz T-bird. The Palomino’s enhancements definitely help out here.
We generally graph results from this OpenGL first-person shooter using a nifty time scale, like so. However, that sort of graph is almost impossible to read with so many processors and speeds, so we’ve had to resort to averages.
The Athlon XP 1800+ is nearly 20% faster than the 2GHz P4 in this test.
3DMark’s DirectX 8-based tests stress a system in some different ways. Most scenes are absolutely loaded with triangles, and advanced graphics features, including AGP performance, are on prominent display.
The field of contenders in 3DMark 2001 lines up about like it does in Quake III Arena. In only the second test so far, the Athlon XP 1800+ comes out slower than the 1.8GHz Pentium 4.
SPECviewperf workstation graphics
Viewperf measures a different brand of graphics performance: professional OpenGL applications like CAD and 3D modeling. These workstation-class apps often stress a system in different ways than the gaming tests.
Normally, we’d throw all the viewperf results into one large graph, but this time around, we had too many different processors and clock speeds to do so comfortably. Individual graphs are the order of the day. I’ve omitted results for the ProCDRS test. Every single CPU scored between 16.03 and 16.07 on ProCDRS.
The Awadvs test is limited by our GeForce3 video card in this test. Once the processors reach a certain level of performance, the results all come out about the same.
There’s a wide gap here between the Athlon 1.2GHz and the Athlon MP 1.2GHz. I’m not sure if it’s SSE instructions or hardware pre-fetch or what, but the performance difference between the two CPUs is striking.
With the exception of Light and Awadvs, the tests are a back-and-forth battle between the Athlon XP 1800+ and the Pentium 4 2GHz for the top spot. These two CPUs are extremely well matched for workstation 3D graphics apps.
The Sphinx speech recognition tests came to us via Ricky Houghton, who works in a speech recognition effort at Carnegie Mellon University. They’re based on Sphinx 3.3, which is an advanced system that promises greater accuracy in speech recognition. However, Sphinx 3.3 still can’t quite run fast enough on a standard PC to handle tasks in real time, as you can see below.
The Palomino enhancements put the Athlon XP in the running in Sphinx 3.3. Sphinx really likes hardware pre-fetch logic. All of the top seven slots are occupied by CPUs with pre-fetch capabilities, and there’s a noticeable speed gap between the processors with and without pre-fetch.
None of the processors here is quite able to run Sphinx in real time, but the time is drawing near. If either the Athlon XP or Pentium 4 gets a faster memory subsystem soon, Sphinx will finally be running faster than real time.
On to Tim Wilkens’ computational benchmark, ScienceMark. This suite of tests measures number-crunching ability by running some computationally intensive scientific equations. Like 3DMark, ScienceMark then spits out a composite number denoting a system’s overall score in the suite.
Here’s how our contenders fared:
If you plan to be talking smack in the physics lab, you’d darn well better have an Athlon XP system under that lab coat. The Athlon XP takes this one in a walk. When you bust it out into some of the individual tests, the results are like so:
The Athlon XP shows broad strength, while the Pentium 4 demonstrates a special aptitude for Primordia.
Well, there’s not much I can say for the Athlon XP that the preceding test results haven’t already said for it. It’s the fastest x86-compatible processor on the planet, and by more than just a hair. AMD has easily recaptured the performance lead from Intel, courtesy of an old-fashioned whoopin’. At 1.53GHz, AMD’s Palomino core delivers serious speed.
Whether AMD’s new marketing plan will delivers similar success in the sales department remains to be seen. It’s clear that AMD has been very conservative with their model number ratings for the Athlon XP, which is probably for the best. There’s little reason why some know-it-all ought to be able to complain that the Athlon XP 1800+’s performance isn’t really on par with a 1.8GHz Pentium 4. I’m curious to see how this model numbering scheme works out for AMD.
The Athlon XP arrives with some very competitive prices, although they aren’t quite as crazy cheap as T-birds have been of late. Liberated from matching Intel at price-per-MHz, AMD has established a much more sane pricing strategy with the Athlon XP. Initial prices for the different speed grades line up like so:
AMD says the Thunderbird Athlon chips will continue to be available into 2002, and these chips will not receive the model number treatment. On the SMP front, AMD will continue to release Athlon MP processors as separate products; the company says Athlon XP processors will not be tested or validated for use in dual-processor systems.
All in all, the Athlon XP looks very good. However, I should mention something. Last time out, when the P4 2GHz captured the performance lead from AMD, I said it was only a matter of time before AMD recaptured the lead. Turns out I was right. This time around, the roles are reversed. Intel is reportedly readying its Northwood Pentium 4 chip for release. When Northwood hits the streets with its 0.13-micron die shrink and 512K of L2 cache, it will give the Athlon XP the fight of its life. Until then, though, the Athlon XP is the undisputed heavyweight champion.