Last we checked in a big processor benchmark roundup, the 1.8GHz Pentium 4 was trailing close behind the 1.4GHz Athlon. Since then, AMD has laid low, holding back on releasing a new Athlon speed grade. If the Pentium 4 2GHz can catch the 1.4GHz Athlon, Intel will have pulled off a neat double play, reaching 2GHz and retaking the performance lead at once.
We’ll investigate whether Intel has accomplished this goal below. We’ll also consider in more depth what, if anything, the 2GHz milestone really means to the PC market.
The primary difference between the Pentium 4 2GHz and the previous P4s we’ve tested is that this one is packaged to fit into Intel’s new microPGA 478-pin socket. Intel says Socket 478’s extra pins for power and ground connections improve stability for higher clock speeds. The older, larger Socket 423 will be phased out over time, and Socket 478 will supplant it. For the time being, the Pentium 4 will be available in both packages at speeds up to 2GHz.
The P4 2GHz is not the upcoming P4 chip code-named Northwood. It’s still built on Intel’s 0.18-micron fab process, like all previous P4s, and it’s the same basic core design, code-named “Willamette” in a past life. If you want to buy a Pentium 4 now and upgrade to the 0.13-micron Northwood chip later, be sure to get a Socket 478 motherboard. The P4 won’t be available in 423-pin form above 2GHz. We’ve kicked around the possibility of a Socket 423-to-Socket 478 adapter, but if anyone’s planning to make one, we haven’t heard about it. It certainly seems possible, but obviously its use would be relatively limited.
The 478-pin P4s are teeny little beasts, and Socket 478 takes up what seems like a ridiculously small portion of the motherboard’s surface area. Have a look at the pictures to see what I mean.
Despite the differences in physical appearance, the P4 system we’re testing today uses Intel’s 850 chipset with RDRAM, just like past Socket 423 systems.
We’ll be comparing the new Pentium 4 to its top x86 competitors, pictured below, plus the top contenders from the x86 value market, the 900MHz Intel Celeron and 1GHz AMD Duron.
Pentium III 1.2GHz, Pentium 4 (Socket 423), Pentium 4 (Socket 478), Athlon We should point out that while the microPGA Pentium 4’s packaging is quite a bit smaller than the rest of the field, the chip itself, inside all of the packaging, is the largest of the bunch (though the same size as the Socket 423 P4 chip). Still, the miniature packaging is pretty spiffy.
Making sense of the megahertz “myth”
One of the things we have to return to again and again when evaluating the latest processors is the question of clock speedusually measured in megahertz and gigahertzand how that important variable affects overall performance. Now that we’ve hit the symbolic 2GHz milestone, it’s appropriate to stop and consider the question in more depth. We’ve had quite a bit of confusion about clock speed of late, in part thanks to the Pentium 4’s penchant for especially high clock frequencies. We’ve also had a Mac-versus-PC flare-up that led to some stern words over Apple’s somewhat deceptive attempts to make an otherwise-important point. (Apple hand-picked six Photoshop filters on which to base a comparison, then called the 866MHz G4 “58% faster” than a 1.7GHz Pentium 4.)
How MHz has mattered
Whatever you think about Apple’s marketing, the truth remains that clock speed isn’t everything. It is possible for a 1.2GHz processor to outrun (or completely crush) a 2GHz processor in real-world performance. Many things help determine a system’s overall performance, and the P4’s tendency to run at stratospheric clock speeds creates a substantial marketing problem the for likes of Apple and, more directly, AMD. The PC marketand especially its sales and marketing armshas keyed in on clock speeds for years as an indicator of overall performance. Not only performance, in fact: it’s safe to say that MHz has equated to merit. A 900MHz system is widely recognized by consumers as “better” than a 700MHz system.
That perception is a bit naive, but its foundation is solid. Generally, clock speed has served as a marker for a PC’s place in the grand scheme of things. Before you scoff, consider that in an Intel-dominated PC market, MHz is a surprisingly effective marker. It can tell you all sorts of things about a PC in an instant. For instance, say you have two retail desktop PC systems, one 333MHz and the other 350MHz. Judging by the clock speeds, they’re probably Pentium IIs. One system has a 66MHz front-side bus (FSB), while the other has a 100MHz FSB. The 333MHz box talks to memory at 66MHz, while the 350MHz box uses PC100 SDRAM.
Then compare that 350MHz box to a 1.2GHz system. We’re probably looking at AGP 2X versus 4X, ATA/33 versus ATA/100, a system without USB ports versus one with four, the list goes ona wide disparity in hard drive performance, graphics chip horsepower, RAM, and more. The standard amount of RAM on the 350MHz box probably matches the amount of RAM on the 1.2GHz box’s graphics card.
This sort of feature escalation is standard practice for the Dells and Gateways of the world, and it’s one reason why the “MHz myth,” as it’s been called, has been an effective means of communicating a very complex reality to consumers, whether those consumers be clueless first-time PC buyers, aloof IT managers, or even in-the-know PC enthusiasts. (I hate to do it, but I feel bound to note that in the case of the G4, 866MHz is a pretty good marker. The Mac system probably has a 133MHz FSB, PC133 memory, and an ATA/66 storage interface, much like an older, 866MHz Pentium III system. Not that that’s the whole story.)
Needless to say, if you’re selling an Athlon-based PC that runs at 1.4GHz and your competitor is selling a 2GHz Pentium 4-based system for a similar price, you’re going to be fighting a nasty uphill battleeven if your product actually offers more bang for the buck. If you have to count on the wannabe-geeks in blue knit shirts at Best Buy to explain “the megahertz myth” to consumers, you are in deep, deep trouble.
Where MHz come from
That said, let’s look at how these clock speed numbers get determined. The abiding reality of the PC market for the past twenty years has been the tendencywrapped up in Moore’s Law and executed with precision by Moore’s company, Intelfor processor power and clock speeds to ratchet upward with regularity. Speaking simply, a processor’s clock speed is determined by several things. Among them:
- Manufacturing technique and efficiency I’m listing this one first because it is, in some respects, the most important variable here. Most regular clock speed increases come from improvements in manufacturing techniques or efficiencies. When Intel, AMD, or whoever produces a wafer full of chips, the quality of the chips on that wafer determines how fast each chip will run. Although the Athlon 700MHz and the Athlon 1.4GHz were made using the same basic manufacturing process (with a 0.18 micron feature size and copper interconnects), AMD has gotten better at producing these chips over time, so clock speeds have risen. Often, these minor refinements in manufacturing efficiency come from tweaks to the process used to fabricate chips, or from minor changes in the design of the chips themselves.
Of course, CPU makers try to control things by timing the release of new chips and offering pricing points up and down the supply and demand curves. But they’re simply copingquite cleverly, it must be saidwith the realities of chip fabrication. (We crazy overclocking types try to catch CPU makers selling chips rated to run at much lower speeds than they’re capable. Occasionally, we find a real gem, like the Celeron 300A, that will happily run over 50% faster than its rated speed. We thenalso quite cleverlybuy ’em cheap and run ’em at higher speeds.)
Every so often, CPU makers transition to a newer, more advanced manufacturing process. Recently, Intel has been making just such a move from its 0.18-micron, aluminum-based process to a copper-based, 0.13-micron process that uses low-capacitance dielectrics. Chips made on this newer process are smaller, consume less power, run cooler, and are able to run at higher frequencies. The 1.2GHz Pentium III, which we reviewed not long ago, is fabbed on Intel’s new process. These chips have been reported to run just fine at upwards of 1.4GHz, while previous, 0.18-micron Pentium IIIs haven’t been good for much over 1GHz. This sort of transition to a new manufacturing process, known as a die shrink, usually brings with it headroom for ever-higher clock speeds. The Pentium 4 has yet to undergo a die shrink on Intel’s new process, but it should soon.
- Microprocessor design The other big variable in the clock speed equation is processor design. Older designs don’t generally take well to higher clock frequencies, which is one reason why you don’t see any 1.5GHz 486s selling in “value” PCs. Newer designs employ deeper pipelineswhere less work gets done at each stage of the gamein order to better tolerate higher clock frequencies. Thus, some processor designs are better suited to higher clock speeds than others.
The contrast between the Pentium III and Pentium 4 is a case in point. The PIII made it up to 1.13GHz when manufactured on Intel’s 0.18-micron fab process. The Pentium 4, with its radical, 20-stage pipeline, debuted at 1.4 and 1.5GHz, and has now hit 2GHzall on the same 0.18-micron process. The chips were made in essentially the same fashion, but the P4 design takes to higher frequencies better. (For more on the Pentium 4 and how it’s optimized for high clock speeds, see our original Pentium 4 review.)
Of course, there are tradeoffs here. The P4 typically gets less work done per clock cycle than the PIII, so a 1.4GHz P4 isn’t necessarily a better performer than a 1GHz PIII. But, as you’ll see, the P4 at 2GHz makes the PIII look like the sad, old man that it is.
- Power and cooling It’s possible to wring some extra speed out of a processor after supplying it with additional power and better cooling. Extreme cooling solutions like those from Kryotech can allow for big jumps in clock speed, though they’re not always cost effective. Most overclockers use minor voltage increases and beefier versions of conventional air-cooling equipment to help bring stability when they crank up the clock.
Power and cooling requirements are limitations, too. Laptops generally run at lower clock rates than desktops to avoid excess heat and power draw. As clock speeds have risen, standard-issue desktop PCs have gone from small, passive heatsinks in the 386 and 486 eras to the massive, ducted, fan-driven active cooling solutions of today.
The question of how clock speed translates into performance is a complex one, but we can make a few generalizations. One of the most useful concepts in understanding processor performance is instructions per clock, or IPC. IPC describes the amount of work a processor does in a clock cycle.
Modern processors play all sorts of tricks that make the concept of IPC a little bit slippery. For example, techniques like branch prediction and speculative execution can make some instructions appear to execute in “zero” clock cycles. Likewise, SSE instructions can complete the equivalent of a whole series of conventional instructions in just a few clock cycles. With deep pipelines that keep a raft of instructions “in flight” at once, and with different types of instructions that take different amounts of time to execute, pinning down an exact number of instructions per clock on modern CPUs is pretty much impossible. The concept of IPC has survived because it’s a generalization, a useful conceptual term. It’s safe to say that the Pentium 4’s IPC is usually lower than the PIII’s or Athlon’s, even though that’s not always the case.
Put simply, processor performance is determined by the intersection of IPC and clock speed. We’ve demonstrated in the past that the Pentium 4 1.7GHz performs roughly the same as a 1.2GHz Athlon. Taken together, the respective IPCs and clock speeds of these two processors produce an approximate match.
I should point out here that in the world of processors, neither IPC nor clock speeds require a value judgment. The Pentium 4 doesn’t “suck” because it has a relatively low IPC, and the G4 isn’t a screamer simply because it has a relatively high IPC. By the same token, the P4’s 2GHz clock speed doesn’t automatically make it the greatest CPU ever, and the G4’s relatively pokey top speed of 866MHz isn’t a sure sign of a loser. Only the combination of IPC and clock speed determines performance; how you arrive there is interesting, but it’s a secondary question.
All sorts of things affect the overall performance of a computer system, of coursethe amount and type of RAM, hard disk transfer rates and access times, the front-side bus speed, and truly esoteric considerations involving overall system tuning. A relatively slow memory subsystem tied to lower-speed bus can starve a processor’s pipelines, effectively reducing its IPC. Putting the same processor at the same clock speed into a system with a faster bus and memory could bring its IPC back up. And CPU performance itself is only part of the game. You can have the fastest processor on the block, but if your graphics card’s a dog, your system ain’t gonna run 3D games well.
To pile on just one more layer of complexity, there’s also the question of the software you’re running. Really well optimized code can make brand-new processors hum, but moldy, old executables can make a Pentium 4 perform more like a Pentium II. We considered this issue at some length in this article, when we compared programs compiled with different compilers or compiler options. We’ll revisit this question briefly in our POV-Ray tests below.
Now, let’s move on to the benchmarks, where we can put all this theory into action.
Our testing methods
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 1.4GHz
|AMD Athlon 1GHz
AMD Duron 1GHz
|Intel Pentium 4 1.4GHz
Intel Pentium 4 1.6GHz
Intel Pentium 4 1.8GHz
|Intel Pentium 4 2GHz||Intel Celeron 900MHz
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||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 2.5||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
- LAME 3.70
- SPECviewperf 6.1.2
- POV-Ray for Windows version 3.1g (multiple compiles)
- 3DMark 2001 Build 200
- Quake III Arena 1.17
- Serious Sam v1.02
- ScienceMark 1.0
- Sphinx 3.3
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.
We’ll kick it off, as usual, with a look at memory performance. The Pentium 4 has traditionally excelled at memory bandwidth performance, and the 2GHz model is no exception.
The P4’s impressive memory bandwidth has often been attributed to its dual channels of Rambus DRAM, but recent previews of the Pentium 4 using DDR SDRAM, like the 1.2 and 1.4GHz Athlons do in the test above, have shown impressive memory performance, as well. Our preliminary internal testing with a VIA P4X266 reference motherboard and PC2100 DDR SDRAM has shown Sandra memory scores in excess of 1000MB/swell above Athlons using the same RAM. The long and short of it is that the Pentium 4 makes great use of a fast memory subsystem.
To get a little more detailed picture of how these processors access memory, we’ll look at Linpack results. To keep things simple, I’ve only included one result per CPU type. Otherwise, the graph’s darn near impossible to read.
Of course, this graph isn’t terribly easy to read even now without some explanation. What you’re seeing is the performance, in MFLOPS, of a floating-point calculation being performed on data matrices of varying sizes. A smaller data matrix will fit into the CPU’s L1 an/or L2 caches, allowing for some very fast calculations. As the matrix size grows, the data must be accessed in main memory, slowing things down. Let’s break it down:
- The orange line is the Pentium 4. It 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.
- The green line represents the Athlon. Unlike the Pentium 4, the Athlon’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 doesn’t really drop off going to main memory until we hit about 320K. However, the Athlon’s L2 cache is a tad slower than the P4’s; you can see, at about 64K, where the Athlon has to start moving beyond its L1 data cache into it slower L2 cache, and the MFLOPs drop off. Once the Athlon gets to main memory, it’s quite a bit slower than the P4, but faster than all the PC133 SDRAM-based systems.
- The two value processors here, the Duron and Celeron, each have an effective cache size of 128K. The Duron whups the Celeron, though, delivering a much higher peak and outrunning all the other PC133-based systems when accessing main memory thanks to its hardware prefetch logic. The new Athlon Palomino, due to replace the current desktop Athlon chips before long, will include this same prefetch mechanism, and it should take better advantage of its available memory bandwidth as a result.
Keep in mind that memory performance alone doesn’t determine overall performance, as we’ll see below. However, it does give us some important clues about why these different chips perform like they do.
Business Winstone 2001
Now for our first indication of how these processors really stack up. Business Winstone tests performance in everyday office apps like spreadsheets, word processors, and web browsers.
For the first time ever, the fastest Pentium 4 beats out the fastest Athlon in the Business Winstone test. That’s new territory for Intel, since AMD’s Athlon has dominated the top ranks of performance for nearly a year now.
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.
Once again, the 2GHz Pentium 4 just edges out the 1.4GHz Athlon, bolstering Intel’s possible challenge to the throne.
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 the graphs were getting big enough already, I’ve again omitted results for some of the processors in our test.
The Athlon wins this one by a half-second margin, but it’s extremely close. The different compiled versions of this program show what a dramatic difference a compiler can make. The Pentium 4 at 2GHz ties the Pentium III at 1.2GHz using the original code, but the P4-optimized version vaults the same CPU to the second spot on the charts. That’s a 37-second difference between the highest and lowest scores for the 2GHz Pentium 4. By contrast, the gap between scores for the Athlon 1.4GHz is only 20 seconds. Clearly, both CPUs benefit from compiler optimizations, but the Athlon does a better job slogging through legacy code.
The Athlon’s lead grows when rendering the more complex, ray-traced chess2 scene. For pure computational power, it’s hard to beat the Athlon’s killer FPU.
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. However, to keep it fair, we’ve avoided the newer builds of LAME that incorporate support for the Athlon’s 3DNow! instructions.
The seesaw battle we’ve seen in the benchmarks above comes to rest here in a dead heat. The fastest Athlon and the fastest P4 will encode your music in exactly the same time using this version of LAME. It may seem kind of sad that the P4 “needs” 2GHz to match the 1.4GHz Athlon’s performance, but in truth, it’s a sign that the Pentium 4 has arrived. To match the 1.4GHz Athlon in raw computational performance, no matter how you do it, is no mean feat.
Quake III Arena
Now for some 3D gaming benchmarks, where those memory bandwidth numbers we looked at above are more likely to come into play. The P4 has always loved Quake III Arena, and the 2GHz version should be no exception.
Yep. The P4 just clobbers the Athlon in Quake III. Notice that this game is an exception to the rule: the P4 isn’t only faster in absolute terms here; it’s faster clock-for-clock than the Athlon. For whatever reason, the Pentium 4’s IPC is higher than the Athlon’s in Quake III. It doesn’t happen often, but it does happen.
Let’s try another OpenGL-based first person shooter for good measure. Serious Sam allows us to plot performance over time, so we can see how the different processors handle different portions of the game demo we’re timing. In this case, we’ve used five-second intervals. The graph does get a bit crowded, but I think it’s still comprehensible. The end result looks like so:
This one is close. If we were reporting only an average frame rate here, the top Athlon and the top P4 would be essentially tied. However, our graph shows us that the P4’s peaks are higher, and lows lower. The edge in playability would have to go to the Athlon, since it handles the worst-case scenarios better, ensuring smoother overall gameplay.
And oh yeah: don’t buy a Celeron.
The battle for 3DMark supremacy has been a back-and-forth affair, with the release of NVIDIA’s 12.41 graphics drivers pushing things back in Intel’s favor. Here’s how our contestants stack up:
Here’s another 3D graphics test where the P4 darn near matches the Athlon in clock-for-clock performance. Once the P4 reaches 2GHz, the margin of victory is substantial. When it comes to 3D graphics performance, especially with NVIDIA’s highly optimized drivers, the Pentium 4 is hard to beat.
SPECviewperf workstation graphics
Viewperf measures a different brand of graphics performance: professional OpenGL applications like CAD and 3D modeling.
In the Awadvs, DX, and Light tests, the 2GHz P4 system shows some performance anomalies, turning in much lower scores than one would expect. I checked and double-checked this one. Vsync was disabled, the screen resolution and bit depth were set right, the proper drivers and patches were installed, the AGP aperture size was the same (256MB) as the other systemseverything seemed right. We checked with Intel, who in turn checked with NVIDIA. NVIDIA confirmed there is an issue with the 2GHz Pentium 4 and the current drivers. It apparently relates specifically to high graphics bandwidth benchmarks, and like us, they’ve only seen the problem in viewperf.
Newer (currently unreleased) drivers from NVIDIA, labeled version 14.61, resolved the problem for us. As you can see, the red bar on the graph (which represents results at 2GHz with the 14.61 drivers) shows the P4 2GHz performing more line with expectations.
The Sphinx speech recognition tests, a relatively new addition to our test suite, came to us via Ricky Houghton, who works in the 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. You can find source for Sphinx at SourceForge.org.
Sphinx seems to rely heavily on two things: high memory bandwidth, and Intel’s SSE instructions, so it’s well suited to the Pentium 4. What we’re after is a computer that can run the Sphinx 3.3 faster than real time, so that this new version of Sphinx is practical for everyday applications. (Running at about 20% faster than real time would be ideal.) We’ve come close in the past, but we haven’t quite made it. Can the 2GHz P4 finally put us over the hump?
Not yet, but we’re getting very close, especially with the Microsoft compiler. It won’t be much longer now. It may take a new memory architecture to get us to 0.8 times real time, which is the ultimate goal.
On to Tim Wilkens’ (now Dr. Timothy Wilkens, thank you very much) computational benchmark, ScienceMark. This suite of tests measures computational ability by running some well-known (in the right circles) 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:
In this very pure test of raw computational power, the 2GHz Pentium 4 shows it’s again a match for the Athlon. This kind of number crunching may not be its strong suit (certainly not compared to graphics), but the P4 can definitely hold its own. These processors are fast in very different ways, however, as some of the individual test results show.
The P4 dominates in Primordia, while the Athlon does especially well in the QMC and Liquid Argon tests.
All in all, our benchmarks have made it clear: by just a smidgen, the Pentium 2GHz edges out the 1.4GHz Athlon to retake the x86 performance crown for Intel. The P4 isn’t faster in every way. It’s very close to the Athlon in intensive computational tests, and the P4 can be quite a bit faster than the Athlon in gaming, graphics, and other bandwidth-intensive tasks. The Pentium 4 still shows some weakness when running legacy code, but it’s amazing how extra MHz can help soothe those concerns. The P4 has ramped up quite nicely, and it’s finally hitting its stride. If you must have the absolute fastest desktop PC processor at any price, the Pentium 4 2GHz is your ticket.
We expect the Pentium 4 to keep its performance lead just as long as AMD wants it to. When the time comes, AMD will release a 1.5GHz Athlon based on the new Palomino core, which is even faster clock for clock, and recapture the performance lead. I wouldn’t be surprised to see these chips accompanied by a “MHz isn’t everything” consumer education marketing blitz from AMD. (Did I just use the terms “AMD” and “marketing blitz” in the same sentence? Check my temperature.)
For the P4, 2GHz is a nice, round number, but the big product transition is still ahead, when the die shrink comes in the form of the chip code-named Northwood. Rumor has it Northwood may include 512K of L2 cache and possibly other enhancements to improve its clock-for-clock performance. How well (and how quickly) the Pentium 4 will make the transition to Intel’s new, 0.13-micron fabrication process is still an open question. Judging by what we’ve seen so far from both the P4 and Intel’s 0.13-micron fab process, the two things ought to go together very well.
For now, Intel has decided to get aggressive with pricing and marketing going into the Windows XP launch and the holiday buying season. To that end, Pentium 4 prices will be cut substantially across the line, though the premium for the 2GHz model remains hefty. As of today, P4 prices will look like so:
The matching prices for the bottom three speed grades mean the 1.3 and 1.4GHz P4s will be effectively killed off soon, which is good, because they’re not very compelling options.
For enthusiasts building their own computers, these price cuts are a step in the right direction, but the Athlon remains a much better value. Retail boxed 1.4GHz Athlons are selling online now for about $130. The competing P4 is still over four times the price. And with the P4, you’re stuck with expensive Rambus DRAM. Better P4 memory options are just on the horizon, but they’re not quite here yet.
For consumers looking to buy PCs from system vendors like HP or Compaq, the P4 value proposition should be be quite a bit better. Intel offers big OEMs deep discounts on processors, and we expect to see Pentium 4-based PCs selling at very competitive prices over the next few months. Unfortunately, many of those systems may be saddled with relatively slow PC133 memory if Intel’s plans unfold as they’re apparently planning. Consumers will want to watch the system specs carefully. We’d recommend seeking out and pricing at least one Athlon-based option before making a purchase decision. Unfortunately, Athlons seem to be getting harder to find on retail and from bigger OEMs.