Beyond that, there’s not a lot new to report. So, after spending untold hours benchmarking a wide range of comparative systems against this new AMD processor, I am currently ingesting a carefully tailored cocktail of stimulants and depressants in order to summarize my results with sparkling wit and rapier prose.
Or something like that.
Anyhow, this new processor from AMD should address some concerns we’ve had about the Athlon XP platform’s ability to keep up with the Pentium 4 in memory and bandwidth-intensive applications. The release is especially timely in light of Intel’s recent introduction of a new Pentium 4 3.0GHz chip with an 800MHz front-side bus. Although the Athlon XP’s effective bus clock speed is half that of the quad-pumped bus on the Pentium 4, both processors’ nominal bus speeds are 200MHz. The Pentium 4 tends to do an awful lot of very aggressive prefetching of data from memory into on-chip cache, so the P4 especially likes a fast bus and fast memory. The Athlon XP also has prefetch logic, but we’ve found it to perform relatively well in real-world benchmarks even when it has a pronounced set of disadvantages in clock speed, bus speed, and total memory throughput.
In other words, the Athlon’s move to a 400MHz bus may put it right back into the game against the screaming fast Pentium 4 3.0GHz with an 800MHz bus. Of course, we have a full set of benchmarks results, so you can see whether or not this puppy delivers.
Beyond the slightly higher clock speed and the faster bus, the Athlon XP 3200+ isn’t much different from previous spins of the “Barton” core. The chip has 512K of L2 cache, and is built using AMD’s 0.13-micron copper fab process.
The biggest news beyond this new CPU with a faster bus is, of course, in chipsets.
NVIDIA’s new nForce2 flavors
With support for a 400MHz bus on the Athlon XP comes a pair of new chipset products from NVIDIA under the nForce2 umbrella, the nForce2 Ultra 400 and the nForce2 400. Near as I can tell, the nForce2 Ultra 400 is the familiar nForce2 chipset with dual-channel DDR400 and official support for a 400MHz bus. If I’m right about that, I suppose the Asus A7N8X Deluxe 2.0 motherboard we’re using to test the Athlon XP 3200+ today is based on the nForce2 Ultra 400 chipset. Little did I know.
The nForce2 400, meanwhile, is a little more novel. The nForce2 400 has support for only one channel of DDR400 memory, and it doesn’t support the MCP-T south bridge chip, so it won’t have features like AC-3 audio, Firewire, or a 3Com NIC. The nForce2 400’s obvious mission in life is to take on VIA’s KT400A head to head, while establishing NVIDIA’s nForce2 Ultra 400 as a premium product.
VIA’s answer: KT600
VIA has its own chipset refresh in the works with 400MHz bus support, and they’ve gained tactical numerical superiority by calling their chipset the KT600. The KT600 isn’t a dual-channel DDR solution, but a second channel of memory isn’t usually all that helpful on the Athlon XP platform anyhow. The KT600 is different from the KT400A in two basic ways: it has a slightly optimized memory controller, and official support for a 400MHz front-side bus. Beyond that, the KT600 is essentially the same thing as the KT400A.
VIA says KT600 motherboards should begin arriving within weeks. NVIDIA’s nForce2 400 series is shipping now, so NVIDIA has a bit of a head start here. Given NVIDIA’s impressive success with the nForce2 and VIA’s tenacity, the Athlon XP 3200+ should be set for quality core-logic chipsets. The tough question may be which one to choose.
We’ll revisit this little chipset spat once a KT600 board arrives in Damage Labs, which should be fairly soon. For now, let’s get on to the CPU benchmarks.
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:
|Athlon XP 3200+||Athlon XP 2500-3000+||Pentium 4 2.53-3.06GHz||Pentium 4 3.0GHz|
|Processor||Athlon XP ‘Barton’ 3200+ 2.2GHz||Athlon XP ‘Thoroughbred’ 2600+ 2.083GHz
Athlon XP ‘Barton’ 2500+ 1.83GHz
Athlon XP ‘Barton’ 2800+ 2.083GHz
Athlon XP ‘Barton’ 3000+ 2.166GHz
|Pentium 4 2.53GHz
Pentium 4 2.8GHz
Pentium 4 3.06GHz
|Pentium 4 3.0GHz|
|Front-side bus||400MHz (200MHz DDR)||333MHz (166MHz DDR)||533MHz (133MHz quad-pumped)||800MHz (200MHz quad-pumped)|
|Motherboard||Asus A7N8X Deluxe v2.0||Asus A7N8X Deluxe v2.0||Aopen AX4R Plus||Intel D875PBZ|
|North bridge||nForce2 SPP||nForce2 SPP||82845G MCH||82875P MCH|
|South bridge||nForce2 MCP-T||nForce2 MCP-T||82801DB ICH4||82801DB ICH4|
|Chipset drivers||2.03||2.03||Intel Application Accelerator 2.3||Intel Application Accelerator for RAID 3.0|
|Memory size||512MB (2 DIMMs)||512MB (2 DIMMs)||512MB (2 DIMMs)||512MB (2 DIMMs)|
|Memory type||Corsair TwinX XMS3200LL DDR SDRAM at 400MHz||Corsair TwinX XMS3200LL DDR SDRAM at 333MHz||Corsair TwinX XMS3200LL DDR SDRAM at 266MHz||Kingmax DDR-400 SDRAM at 400MHz|
|Hard drive||Seagate Barracuda V 120GB ATA/100||Seagate Barracuda V 120GB ATA/100||Seagate Barracuda V 120GB ATA/100||Seagate Barracuda V 120GB SATA 150|
|Graphics||ATI Radeon 9700 Pro 128MB (7.84 drivers)|
|Sound||Creative SoundBlaster Live!|
|OS||Microsoft Windows XP Professional|
|OS updates||Service Pack 1, DirectX 9|
All tests on the Pentium 4 3.0GHz and 3.06GHz systems were run with Hyper-Threading enabled. The lower speed grades of Pentium 4 chips tested here don’t support Hyper-Threading.
Thanks to Corsair for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, Corsair’s RAM is definitely worth considering.
The test systems’ Windows desktops were set at 1024×768 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- Cachemem 2.65MMX
- SiSoft Sandra Standard 2003 (2003.3.9.44)
- Compiled binary of C Linpack port from Ace’s Hardware
- ZD Media Business Winstone 2002 1.0.1
- ZD Media Content Creation Winstone 2002 1.0.1
- Cinebench 2003
- Sphinx 3.3
- LAME 3.92
- Xmpeg 4.5 with DivX Video 5.03
- FutureMark 3DMark03
- Unreal Tournament 2003 demo benchmark
- Comanche 4 demo benchmark
- Quake III Arena v1.31
- Serious Sam SE v1.07
- SPECviewperf 7.1
- picCOLOR NT 4.0 demo build 415
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.
Although the nForce2 has technically been capable of running with DDR400 memory for some time now, we’ve found that it performs best when the memory and front-side bus clock speeds are in sync. With the Athlon XP 3200+’s move to a 400MHz bus, we’ve raised our test platform’s memory clock speed to 400MHz, as well. The synchronous 400MHz bus and memory should work well together. All told, the 400MHz bus offers a max of 3.2GB/s of bandwidth, and the nForce2 Ultra 400’s dual-channel DDR400 memory subsystem offers 6.4GB/s of peak throughput.
The Pentium 4 3.0GHz, of course, features 6.4GB/s of memory bandwidth from the CPU into memory across its 800MHz bus. We’ll use some synthetic memory benchmarks to see how these things stack up in actual systems.
The Athlon XP 3200+ pushes through about 300MB/s more data than the 333MHz bus Athlons. As expected, it’s not able to keep up with the Pentium 4 chips in peak throughput.
Linpack shows us visually how the memory subsystems on the Athlon XP and Pentium 4 differ. The Pentium 4 system with Intel’s 875P chipset and an 800MHz bus performs exceptionally well with large matrix sizesa result of its high-bandwidth bus and memory subsystem. The Athlon XP 3200+ is only marginally faster than the Athlon XP 3000+ here, which is a little bit disappointing.
Our bandwidth results with cachemem mirror our Sandra results above, for the most part. The Athlon XP 3200+ is just a notch above the other Athlon XP chips here.
With quality dual-channel DDR chipsets from Intel and NVIDIA, all of the systems we’re testing have nice, low latency numbers in cachemem. The Athlon XP 3200+ gains a bit of an edge on the older Athlon XP chips thanks to its faster bus. Still, the Pentium 4 systems are even faster.
On the Pentium 4 side, we have results from a dual-channel DDR266 system with a 533MHz bus (the P4 3.06GHz) and a dual-channel DDR400 system with an 800MHz bus (the P4 3.0GHz).
Notice how the difference isn’t nearly as dramatic when the Athlon XP moves from a 333MHz bus to a 400MHz bus as it is when the P4 bus goes from 533 to 800MHz. This suggests perhapsjust maybethe Athlon XP isn’t running into a significant memory latency bottleneck on a 333MHz bus. Only at step sizes of 4,096 bytes do we see a real difference between the Athlon XP 3000+ and 3200+.
Athlon XP chips crunch through this test of everyday applications like web surfing and office productivity applications with aplomb. The 3200+ is only a hair faster than its predecessor, but that’s probably par for the course in this test, where bus speed isn’t likely a big bottleneck.
Content Creation Winstone
AMD supplied us with a patch for Content Creation Winstone because they’ve identified a problem with the test. With Windows XP Service Pack 1 installed, certain components (related to media encoding) don’t recognize the Athlon XP’s ability to execute SSE instructions, putting a damper on Athlon XP scores. This is actually a familiar issue that we’ve seen before, but apparently the latest WinXP Service Pack put the kybosh on a Microsoft-implemented patch for this problem.
Rather than argue over the merits of using the patch and the ethics of benchmark(et)ing, I’ve elected to run the test on the Athlon XP 3200+ both with and without the patch. Choose whichever score you’d like to believe and ignore the other one at your leisure.
Either way, the Athlon XP 3200+ comes out smelling like a rose in Content Creation Winstone. To understand why we’re using the 2002 edition of CC Winstone instead of the 2003 version, please go read the CC Winstone section here. It’s all very complicated.
LAME MP3 encoding
We used LAME 3.92 to encode a 101MB 16-bit, 44KHz audio file into a very high-quality MP3. The exact command-line options we used were:
lame –alt-preset extreme file.wav file.mp3
Here are the results…
The 3200+ shaves 1.5 seconds off the 3000+ chip’s encode time, but it’s not quite enough to catch the Pentium 4 chips at 3.0 and 3.06GHz.
DivX video encoding
Our video encoding test seems a likely candidate to benefit from the new 400MHz front-side bus, and it does benefit a little bit. With Hyper-Threading and a faster bus, though, the Pentium 4 can encode our test video clip nearly a minute faster than the Athlon XP 3200+.
Sphinx is a high-quality speech recognition routine that needs the latest computer hardware to run at speeds close to real-time processing. We use two different versions, built with two different compilers, in an attempt to ensure we’re getting the best possible performance.
There are two goals with Sphinx. The first is to run it faster than real time, so real-time speech recognition is possible. The second, more ambitious goal is to run it at about 0.8 times real time, where additional CPU overhead is available for other sorts of processing, enabling Sphinx-driven real-time applications.
The Athlon XP 3200+ comes in second to only the 800MHz-bus Pentium 4.
Cinebench 2003 lighting and rendering
Cinebench is based on Maxon’s Cinema 4D modeling, rendering, and animation app. This new revision of Cinebench measures performance in a number of ways, including 3D rendering, software shading, and OpenGL shading with and without hardware acceleration.
Cinebench 2003 is multithreaded, so it takes advantage of Hyper-Threading. For the P4 3.06 and 3.0GHz systems, I’ve reported the multithreaded rendering test result, which was always better than the single-threaded result. For the Athlon XP and the other P4 chips, I’ve reported the single-CPU result.
The Pentium 4’s combination of SSE2 and Hyper-Threading proves especially potent in Cinebench’s rendering test. The Athlon XP 3200+ shows only a slight gain over the 3000+ model.
The Athlon XP 3200+ puts in a better showing in the two software shading tests, where it just barely fails to outrun the Pentium 4 3.06GHz. In the OpenGL hardware shading test, however, the Athlon XP chips excel.
Kribibench 3D rendering
Kribibench uses an SSE-aware, software-only rendering engine to generate some very nice looking 3D images. Kribibench is multithreaded, too. We’ve tested with a couple of different 3D scenes, just in case there are any big differences in terms of rendering workload between them.
Since rendering is easily parallelizable, SIMD extensions like SSE and multithreaded execution via Hyper-Threading both benefit performance greatly. Kribibench runs very well on the high-end Pentium 4 chips. The Athlon XP 3200+ is slightly faster than the 3000+, but not enough to matter much.
POV-Ray 3D rendering
POV-Ray has been part of our test suite forever, so we’ll include it here for the sake of completeness. Even the relatively new 3.5 release of POV-Ray isn’t multithreaded and doesn’t take much advantage of SIMD extensions, so this test leans on good the ol’ x87-style FPU for rendering power.
Here’s a rendering test the Athlon XP can win. Oddly enough, though, it’s the 3000+ that comes out on top. The difference is easily within our margin of error, though, so no real surprises here.
Quake III Arena
The Athlon XP 3200+ makes a nice showing in Quake III, but the Pentium 4 has long considered this test its home turf. The move from a 333MHz bus to 400MHz is good for nearly 10 frames per second in our Quake III test, for what it’s worth.
3DMark03’s CPU tests show a real advantage for the Athlon XP 3200+ over the 3000+, but the clear winner here is the P4 3.0GHz/875P chipset combo.
Serious Sam SE
If the P4 considers Quake III its home turf, Serious Sam SE is the Athlon XP’s home field. However, the gap between the top three contenders here, including that pesky P4 3.0GHz, is only about four frames per second.
Pentium 4 = winner.
Unreal Tournament 2003
UT2003 is a fun one. The Athlon XP 3200+ juuuuust edges out the Pentium 4 3.06GHz, only to be beaten by the P4 3.0GHz with the 800MHz bus. Story of its life, apparently.
SPECviewperf workstation graphics
SPECviewperf simulates the graphics loads generated by various professional design, modeling, and engineering applications.
The lead here oscillates between the Pentium 4 3.0GHz and the Athlon XP 3200+.
picCOLOR image processing
picCOLOR came to us via Dr. Reinert Mueller, who wanted us to test his image processing program on a dual-processor Athlon system for him. picCOLOR does a nice job testing common image-processing-related functions with multiple threads, including with Hyper-Threading. To save on graphs, we’ve got an overall score for each CPU, then we have detailed results for four key CPU types.
Dr. Mueller’s reference system, which represents a score of 1.0 on each test, is a 1GHz Pentium III.
The Athlon XP chips prove fastest here, but there’s little net advantage to the Athlon XP 3200+ versus the 3000+. Let’s look at scores in the individual tests.
Although the Athlon XP is faster overall, like so many other applications, which processor is faster in picCOLOR will depend on what you’re doing.
If you’re like me, you’re probably wondering several things about the Athlon XP 3200+. The first question in my mind is: Why only 2.2GHz? The bus speed boost is a nice move, but why not crank up the clock just a tad more from the Athlon XP 3000+? Heck, the fastest clock rate we’ve ever seen on an Athlon XP was in the limited-run T-bred 2800+, which we reviewed at 2.25GHz way back in October of last year. As it is, the Athlon XP 3200+ generally isn’t much faster than the 3000+. Had AMD offered a higher CPU clock speed in addition to the faster bus speed, we’d be… well, maybe not overwhelmed, but at least a little bit whelmed. Instead, we’re slightly underwhelmed.
Next, I hate to say it, but AMD’s credibility on this model-number rating system is showing some strain with the 3200+. Clearly, the Pentium 4 3.0GHz processor combined with the Intel 875P chipset is faster than the Athlon XP 3200+ in a slight majority of our tests, and sometimes by a notable margin. Yes, I know AMD’s rating system is based not on a competing Intel processor but on the performance of a mythical Athlon “Thunderbird” processor running at the clock speed corresponding to the rating. But we all know how the ratings get used in practice, and in practice, the 3200+ model number is a bit of a reach.
I wouldn’t have much of a problem with that if the Athlon XP 3200+ offered a friendly price-performance ratio, but as of today, that’s not quite the case. The Athlon XP 3200+ will sell for $464 in lots of 1000 units. To compare, the Pentium 4 3.0GHz currently lists for $417. AMD may well push street prices for the 3200+ below those of the P4 3.0GHz, but the actual list price for the 3200+ looks mighty steep right about now. For that kind of money, I’d rather have a fast Pentium 4 with an 800MHz bus.
Finally, you may be wondering why the 66MHz bus speed boost didn’t deliver more performance benefits. I believe our memory bandwidth and latency benchmarks more or less predicted this outcome. In truth, the Athlon XP probably didn’t need a faster front-side bus as much as one might have thought. Still, I’m always one to appreciate faster bus speeds, and the Athlon XP’s move to a 400MHz bus is nothing but welcome. Now that the plumbing is up to date, we may yet see another clock speed boost or two for the Athlon XP before the Athlon 64 arrives in September.