AMD’s Athlon 64 FX-53 processor

Perhaps it was the competition. Intel’s Pentium 4 Extreme Edition was magically transmogrified from a Xeon into a desktop part to do battle with the Athlon 64 FX, and Intel recently cranked the P4 Extreme Edition up to 3.4GHz. The Overall Performance Lead is a very important thing to have, especially if you’re playing number two to an 800-pound gorilla like Intel. Or perhaps it was just time for AMD’s Hammer processors to make the move to 2.4GHz, finally reaching clock speeds higher than their Athlon XP predecessors.

Whatever the case, the Athlon 64 FX-53 is here, ready to challenge the P4 Extreme Edition and all comers for the Overall Performance Lead in the x86 processor world. To gauge the FX-53’s success in its quest for the Overall Performance Lead, we’ve lined up sixteen of its competitors and tested it against ’em all. Our contestants range from the exotic (the P4 Extreme Edition 3.4GHz) to the novel (the Pentium 4 Prescott) to the massively overclocked (the Athlon XP-M 2500+ dialed up to 2.4GHz.) To make things even more interesting, we’ll be exploring how the various AMD and Intel processors assembled here scale with clock speed and model number increases.

I promise, if you like performance graphs like I do, this review will satisfy your cravings. Read on to see what I mean.

Big picture of the chip
Here, folks, is a big picture of the chip.

To review, the Athlon 64 FX is much like AMD’s regular Athlon 64 processors. Like the Athlon 64, it has a 1MB L2 cache on chip and a built-in memory controller with support for DDR400 memory. Like the Athlon 64, it has support for SSE2 and extensions for 64-bit memory addressing, giving it a healthy dose of future-proofing. Unlike the Athlon 64, though, the Athlon 64 FX nestles into a 940-pin socket and supports dual channels of DDR400 memory, giving it up to 6.4GB/s of memory bandwidth.

To unlock this goodness, you’ll need a 940-pin motherboard and registered DIMMs, because the Athlon 64 FX won’t work with regular ol’ unbuffered memory. AMD has said that all Athlon 64 chips, including the FX line, will eventually be moved to a new 939-pin socket that doesn’t require registered DIMMs. However, that glorious day hasn’t come yet. For our testing, we used an Asus SK8N motherboard and a pair of Corsair CMX512RE-3200LL memory modules.

About the funny line graphs
Many of the graphs you’ll find on the following pages are colored line graphs intended to demonstrate how the various processor types we’ve tested scale with clock speed increases and other such enhancements. We’ve used such graphs before in processor and graphics reviews, but they’re a little tricky here. In this case, we have six distinct CPU types, ranging from the Athlon XP to the new Pentium 4 “Prescott” processors, and we have multiple performance grades of each type. The Pentium 4 chips are plotted according to straight clock speeds. For the AMD chips, we’ve graphed them according to their model numbers, provided they correspond roughly to Pentium 4 clock speeds.

That leaves a couple of exceptions. Most prominently, we have today’s star, the Athlon 64 FX-53. I chose to graph the Athlon 64 FX chips to correspond with the clock speeds of the regular, non-FX Athlon 64 processors. As a result, the Athlon 64 FX-53 is in a category that’s currently all its own. I believe this arrangement makes sense, all things considered. Then there’s the Athlon XP-M 2500+, which we have overclocked to 2.4GHz. Since the Athlon XP 3200+ runs at 2.2GHz, the XP-M at 2.4GHz was the next logical step in the sequence, so I’ve plotted it as a “3400.” Once we get to the line graphs, you’ll see what I mean. Again, I think this placement is eminently sensible. You’ll see for yourself shortly.

Finally, we’ve used the Pentium 4’s “C” and “E” designations on our bar graphs, where we can list individual product names one by one, while we’ve stuck with codenames on the line graphs, where a single letter designation would be inadequate to capture the lovely nuances of Intel’s product naming schemes. If all this talk of Northwoods and Prescotts baffles you, please read our Pentium 4 Prescott review to become further confused.

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:

All tests on the Pentium 4 systems were run with Hyper-Threading enabled, except where otherwise noted.

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 1152×864 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 2004 (9.89)
• Compiled binary of C Linpack port from Ace’s Hardware
• Discreet 3ds max 5.1 SP1
• NewTek Lightwave 7.5
• Cinebench 2003
• POV-Ray for Windows v3.5
• PICCOLOR v4.0 build 472
• SPECviewperf 7.1.1
• ScienceMark 2.0 beta (23SEP03 build)
• Sphinx 3.3
• LAME 3.95.1 (build from mitiok.cjb.net)
• Xmpeg 5.0.3 with DivX Video 5.11
• FutureMark 3DMark03 build 340
• Comanche 4 demo
• Quake III Arena v1.31
• Serious Sam SE v1.07
• Splinter Cell v1.2
• Unreal Tournament 2003 demo v.2206
• Wolfenstein: Enemy Territory v2.55

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

Memory performance
Before we get to those line graphs I keep talking about, we’ll start off with out usual set of synthetic memory benchmarks.

The FX-53 stomps everything in Sandra’s memory bandwidth test, thanks to its integrated memory controller and twin channels of DDR400 RAM. Interestingly enough, the 200MHz clock speed increase from the FX-51 to the FX-53 improves Sandra scores a fair bit.

Cachemem isn’t as aggressive at wringing the highest possible bandwidth numbers out of each CPU as Sandra is, and here the Pentium 4 Prescott, with its 1MB L2 cache and aggressive data prefetching logic, scores highest.

Linpack shows the FX-53’s memory hierarchy in action, from the L1 cache through the L2 cache and into main memory. In every stage, the FX-53 looks very fast, although the P4 Extreme Edition’s gargantuan 2MB L3 cache makes the FX-53 look tame by comparison.

The Athlon 64’s onboard memory controller pays big benefits for memory access latencies, but the FX series is slower than the regular Athlon 64s, in part because of the latency penalty that comes with registered DIMMs. Nevertheless, the FX-53 is just a tad quicker than the FX-51, probably because its built-in memory controller is running 200MHz faster.

Let’s examine this latency stuff with an additional dimension…

Memory performance (continued)
Not only are our 3D graphs indulgent, but they’re useful, too. I’ve arranged them manually in a very rough order from worst to best, for what it’s worth. I’ve also colored the data series according to how they correspond to different parts of the memory subsystem. Yellow is L1 cache, light orange is L2 cache, and orange is main memory. The red series on the Extreme Edition graph represents L3 cache. Of course, caches sometimes overlap, so the colors are just an interesting visual guide.

All in all, memory access latency is definitely one of the FX-53’s strong suits. The Pentium 4 processors, by contrast, take nearly twice as long to access main memory in some cases, although the P4 Extreme Edition does mitigate this effect somewhat with its L3 cache.

Unreal Tournament 2003

The FX-53 obliterates the competition in the first of our real-world performance tests, slotting into the top spot with ease. Since the tasty new UT2004 uses the same basic game engine, the FX-53 is a very good bet for Onslaught domination.

All of the CPU types scale fairly well with clock speed and model number increases, but the Athlon 64 and FX trajectories look especially promising.

Quake III Arena

The FX-53 can’t quite catch its nemesis, the P4 Extreme Edition 3.4GHz, in Quake III. The Extreme Edition chips seem able to load up vast swaths of Quake III into their L3 caches, giving them a big boost over the stock Pentium 4s.

Wolfenstein: Enemy Territory

The tide turns in Wolf: ET, as the FX-53 again comes out on top.

Tom Clancy’s Splinter Cell

Spliter Cell’s really no contest, with Athlon 64s taking the top four spots, followed by the P4 Extreme Edition processors.

Comanche 4

The P4 Extreme Edition 3.4GHz just barely falls to the Athlon 64 FX-53 in Comanche 4, by a margin of under two frames per second.

Serious Sam SE

Serious Sam has long been friendly territory for Athlons, and this newest Athlon 64 FX is no exception. The Pentium 4 shows little sign of catching up any time soon.

3DMark03

Overall 3DMark performance is obviously constrained by the graphics card, but the FX-53 manages to outrun everything but the P4 EE chips, which barely top it.

The A64 FX-53 performs even better in the 3DMark CPU tests, where the GPU isn’t such a limitation, taking the top spot in test 1 and tying for the lead in test 2.

I should note that in our gaming tests, not only were the Athlon 64s generally faster than the Pentium 4s overall, but the A64 performance trend is definitely up. If AMD can navigate the transition to higher clock speeds without significantly reducing the A64’s clock for clock performance, things look very good for future Athlon 64s.

Sphinx speech recognition
Ricky Houghton first brought us the Sphinx benchmark through his association with speech recognition efforts at Carnegie Mellon University. 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 64 scales very well in Sphinx, as the FX-53 takes the second spot overall. Were it not for the fact the P4 Prescott is an absolute screamer here, AMD would have taken the top spot.

LAME MP3 encoding
We used LAME 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

Pentium 4s have generally excelled in media encoding type tasks, so this result is no surprise. The FX-53 ties with the P4 3.2GHz and the Athlon XP-M at 2.4GHz.

DivX video encoding
This new version of XMPEG includes a benchmark feature, so we’re reporting scores in frames per second now. This is the first app we’ve looked at so far that makes good use of Hyper-Threading, so keep an eye on the HT and non-HT results.

DivX encoding is very much the Pentium 4’s domain. The FX-53 is just middle of the pack here, while its arch-rival takes top honors.

3ds max rendering
We begin our 3D rendering tests with Discreet’s 3ds max, one of the best known 3D animation tools around. 3ds max is both multithreaded and optimized for SSE2. We rendered a couple of different scenes at 1024×751 resolution, including the Island scene shown below. Our testing techniques were very similar to those described in this article by Greg Hess. In all cases, the “Enable SSE” box was checked in the application’s render dialog.

With support for Hyper-Threading and SSE2, 3ds max runs very well on the Pentium 4, but the A64 FX-53 manages to hold its own. Interestingly enough, the P4 Prescott scales pretty well here, threatening to overtake the P4 Northwood at 3.4GHz.

Lightwave rendering
NewTek’s Lightwave is another popular 3D animation package that includes support for multiple processors and is highly optimized for SSE2. Lightwave can render very complex scenes with realism, as you can see from the sample scene, “A5 Concept,” below.

We’ve tested the Hyper-Threaded processors with one, two, and four rendering threads. For non-Hyper-Threaded processors, we just tested with one and two threads. For the line graphs, we’ve tried to pick results from the optimal number of threads to represent each processor.

The FX-53 comes impressively close to matching the Pentium 4 in Lightwave. However, the P4 Prescott shows signs of life again in the Radiosity_ReflectiveThings scene, surpassing the Northwood in clock-for-clock performance.

POV-Ray rendering
POV-Ray is the granddaddy of PC ray-tracing renderers, and it’s not multithreaded in the least. Don’t ask me why—seems crazy to me. POV-Ray also relies more heavily on x87 FPU instructions to do its work, because it contains only minor SIMD optimizations.

The Athlon 64 continues to scale linearly with clock speed in POV-Ray, and the FX-53 reaches a new peak in x87 FPU performance. The Pentium 4 chips just can’t keep up. The Extreme Edition’s extra cache makes no difference at all here.

Cinebench 2003 rendering and shading
Cinebench is based on Maxon’s Cinema 4D modeling, rendering, and animation app. This revision of Cinebench measures performance in a number of ways, including 3D rendering, software shading, and OpenGL shading with and without hardware acceleration. Cinema 4D’s renderer is multithreaded, so it takes advantage of Hyper-Threading, as you can see in the results.

There’s no beating the Pentium 4 in Cinema 4D rendering. Hyper-Threading gives the Pentium 4 a big boost over the Athlons, but even without it, the P4s would be faster.

The Athlon 64 FX-53 leads solidly in the rest of the Cinebench shading tests, reclaiming some of the luster it lost in the rendering results.

SPECviewperf workstation graphics
SPECviewperf simulates the graphics loads generated by various professional design, modeling, and engineering applications.

For workstation graphics, the FX-53 is tough to beat.

ScienceMark
I’d like to thank Alex Goodrich for his help working through a few bugs the 2.0 beta version of ScienceMark. Thanks to his diligent work, I was able to complete testing with this impressive new benchmark, which is optimized for SSE, SSE2, 3DNow! and is multithreaded, as well. Unfortunately, we don’t yet have a version of ScienceMark capable of taking advantage of SSE3’s new complex arithmetic instructions.

In the interest of full disclosure, I should mention that Tim Wilkens, one of the originators of ScienceMark, now works at AMD. However, Tim has sought to keep ScienceMark independent by diversifying the development team and by publishing much of the source code for the benchmarks at the ScienceMark website. We are sufficiently satisfied with his efforts, and impressed with the enhancements to the 2.0 beta revision of the application, to continue using ScienceMark in our testing.

The molecular dynamics simulation models “the thermodynamic behaviour of materials using their forces, velocities, and positions”, according to the ScienceMark documentation.

For some reason, the Athlon 64 FX chips are beaten out by the single-channel Athlon 64s in the molecular dynamics simulation. Perhaps memory latency comes into play heavily here.

Primordia “calculates the Quantum Mechanical Hartree-Fock Orbitals for each electron in any element of the periodic table.” In our case, we used the default element, Argon.

The FX-53 is untouchable in Primordia and the AES cipher test.

These last two tests, SGEMM and DGEMM, measure matrix math performance using several different codepaths optimized with several instruction set extensions, including SSE, SSE2, and 3DNow!

The Pentium 4 still excels at matrix math given the use of SSE2 and vectorized data. All the entries scale well with clock speed in these tests.

picCOLOR image analysis
We thank Dr. Reinert Muller with the FIBUS Institute for pointing us toward his picCOLOR benchmark. This image analysis and processing tool is partially multithreaded, and it shows us the results of a number of simple image manipulation calculations.

An extra 200MHz helps the FX-53 take home the crown in picCOLOR, but the P4 Prescott is coming on strong.

Performance varies greatly between the individual tests, in part because some of them are multithreaded and thus able to take advantage of Hyper-Threading.

Overclocking
One thing that separates the Athlon 64 FX series from the Opteron is the fact that the FX chips are not multiplier locked. AMD apparently figures that if you shell out big bucks for its fastest processor, you derserve a shot at hittting whatever clock speeds you can coax from it. Of course, one shouldn’t expect much in the way of overclocking headroom on a manufacturer’s top-end products.

Using stock air cooling, our FX-53 review sample happily booted and ran at 2.5GHz, but became wobbly at 2.6GHz, even with extra voltage applied. Still, another 100MHz from a CPU this fast isn’t bad for free. Here’s how it ran UT2003.

Your mileage may vary, but our review unit was quite stable at 2.5GHz, and if we were to build a system for ourselves based on it, we’d take the free extra 100MHz and run with it.

Conclusions
The Athlon 64 FX-53 obliterated the competition in our gaming tests and more than held its own everywhere else, making a very strong argument for AMD’s ownership of the Overall Performance Lead. Intel’s Pentium 4 Extreme Edition 3.4GHz gave the FX-53 tough competition in much of our test suite, especially in commercial 3D rendering apps, and the P4 Prescott 3.2GHz unexpectedly pushed the FX-53 in Sphinx speech recognition, some of the SPECviewperf component tests, and picCOLOR image processing. Still, if you held a link gun to my head and forced me to pick one Overall Performance Leader, I suppose the Athlon 64 FX-53 would have to be it—by a hair, depending on the application, blah blah blah. Then again, the Pentium 4 Extreme Edition lists at $999, so the FX-53 is a veritable bargain, if you do the math right.

Because the FX-53 will replace the FX-51 at the same $733 price point, I doubt the introduction of this new product will have much of an effect of the rest of AMD’s lineup. However, the Opteron 148 is essentially the same thing as the Athlon 64 FX-51, and it costs $733, as well. Something’s gotta give there eventually.

For most of us, though, the verdict on the Athlon 64 FX-53 is simple. If you shell out a cool $733 on an exotic new processor from AMD, you’re now going to be getting an even faster CPU, one with more distance between itself and the Athlon 64 3400+ or the Pentium 4 Extreme Edition 3.4GHz. You know what that means: bragging rights. And isn’t that what these Athlon 64 FX chips are all about?