AMD Athlon 64 FX-53 Review March 18, 2004

Fresh off the release of the Athlon 64 3400+, AMD is back again with an even faster chip, the Athlon 64 FX-53. The FX-53 boasts a 2.4GHz clock speed which promises more performance, but how does it compare to previous Athlon 64 processors and Intel's 3.4GHz Extreme Edition. Step right on up for the benchmarks inside!

Introduction

Page:: ( 1 / 16 )

It’s an interesting time to be shopping for flagship processors, isn’t it? Intel has three different 3.4GHz Pentium 4 chips currently listed for sale. Of those, only the “Northwood”-based variant is widely available, with its 512KB L2 cache. And purportedly, that’s the way it’s going to be for a while longer.

AMD has a contender of its own rated for 3400+ levels of speed. At $400, it isn’t exactly affordable, but it does come in just a tad cheaper than Intel’s 3.4GHz Northwood. Then, if you’re willing to step up to the plate, AMD’s Athlon 64 FX-51, a processor with no given indication of performance other than its FX designation, beckons. True to its pedigree, the Athlon 64 FX offers the utmost in performance from AMD at a premium price. Moreover, there’s only one of them, meaning AMD will only offer one version of the FX at a time.

And it seems as the FX-51’s time is up. A 2.4GHz variant, the FX-53 is upon us, bringing with it the same feature set as its predecessor and the same price, too. It’s nearly undistinguishable from the FX before it, save the 53 in its model string. What’s it competing against? Well, Intel’s 3.2GHz Pentium 4 Extreme Edition is readily available, but even the FX-51 was faster than that. That 3.4GHz Northwood previously mentioned is also out in quantity, though it isn’t able to keep pace with the Athlon 64 3400+, much less this powerful new FX-53. So, at least for the time being, AMD’s Athlon 64 FX-53 has the spotlight to itself. Perhaps when Intel ramps up production of the 3.4GHz Extreme Edition, which is currently limited to OEM boxes, we’ll see a fairer fight. Fortunately, we’ll know in just a couple of pages, when we pit the two heavyweights against each other.

[image]

Athlon 64 FX-53

The principle characteristics of the Athlon 64 FX remain intact with today’s FX-53 launch. That is, AMD’s flagship retains its 940-pin ceramic micro-PGA packaging (making it the heaviest processor we’ve ever handled). Manufactured on a 130nm SOI process in Dresden, Germany, it’s comprised of nearly 106 million transistors and measures a massive 193 square millimeters under than heat spreader.


SIDEBAR: The Athlon 64 FX-53 features AMD’s enhanced virus protection security technology, which helps prevent buffer underrun errors in hardware.

More Athlon 64 FX-53 and Overclocking

Page:: ( 2 / 16 )

Athlon 64 FX-53

There are a number of reasons that the Athlon 64 FX is such a large chip. Bigger caches are one; the FX-53 sports 64KB of both L1 data and instruction caches, in addition to a full 1MB of L2 memory onboard. Moreover, the processor incorporates its own memory controller with support for everything from PC1600 to PC3200 registered DDR memory modules. And because it’s a 128-bit controller, it requires that modules be installed in pairs for maximum performance.

One direct result of the integrated controller is that the Athlon 64 FX realizes much more real-world throughput than the Pentium 4, which also employs a dual-channel memory bus through Intel’s 875P chipset. The Sandra 2004 bandwidth test demonstrates this advantage.

The Athlon 64 execution pipeline is laid out much differently than that of Intel’s new Prescott core and its extremely deep, 31-stage setup. The FX’s integer pipeline is 12 stages and the floating-point pipeline is 17, making the chip less flexible in achieving higher frequencies, but much more efficient. Thus, it is very responsive to clock speed increases, such as the 200MHz bump seen moving from Athlon 64 FX-51 to FX-53. Again, the benchmarks should indicate the new 2.4GHz chip’s advantage.

It should also be noted that, for its size and composition, the Athlon 64 FX-53 has much better thermal characteristics than Intel’s Prescott. At 2.4GHz, AMD rates the chip’s maximum thermal power at 89W, whereas Intel’s thermal design power (not necessarily the maximum thermal power) sits at 103W for its 3.4GHz Prescott. On the other hand, Intel’s 3.4GHz Northwood sports a much lower 89W TDP.

Overclocking

Because the Athlon 64 FX sports an unlocked multipler, AMD’s Damon Muzny claims that it cannot support the Cool’n’Quiet feature offered by standard Athlon 64 chips. That’s of little consequence, though. Power users interested in a high-end part like the Athlon 64 FX generally aren’t interested in clock throttling and thermal conservation, anyway.

When we received our Athlon 64 FX-51, we were told that it was an engineering sample, equal to a production part sans the unlocked multiplier. Thus, our production Athlon 64 FX-53 is the first sample we’ve tested without that pesky multiplier locked in place. Now, it should be noted that proper support for multiplier manipulation isn’t universal yet. We had problems setting different speeds with our ASUS SK8V, but our SK8N worked without incident. In fact, a recent BIOS revision adds half-steps, enabling 100MHz jumps.

Unfortunately, the FX-53 didn’t prove to be an eager overclocker, at least with the stock cooling solution in place. Using a combination of voltages up to 1.7V, we were able to hit 2.6GHz and run a couple of benchmarks, but stability was fleeting, at best. The more stable setting was an even 2.5GHz, easily stable throughout testing.

In 3DMark03, the overclock yielded half a percent of additional performance in the 3D test and nearly three percent in the processor test. It bumped the PCMark04 test up an extra three percent, and knocked the Comanche 4 demo score up three percent at 800x600. Unreal Tournament 2004 benefited a percent and a half at 800x600 and Quake III jumped up a meager percent at the same resolution. The extra 100MHz didn’t make an appreciable difference in any of the high-res gaming tests, unfortunately.


SIDEBAR: In order for the enhanced virus protection security technology to work, you’ll need the upcoming Service Pack 2 for Windows XP.

System Setup

Page:: ( 3 / 16 )

System Setup

AMD Athlon 64 FX-53
AMD Athlon 64 FX-51
AMD Athlon 64 3400+
Intel Pentium 4 3.4GHz Extreme Edition (800MHz bus)

ASUS SK8N Socket 940 nForce3 150 Motherboard
Gigabyte K8VNXP Socket 754 K8T800 Motherboard
ASUS P4C800-E Deluxe Socket 478 875P Motherboard

1GB Mushkin Registered ECC 2-3-2 PC3200 DDR Memory (2x512MB)
1GB Corsair Pro Series 3-4-4 PC4000 DDR Memory (2x512MB)

GeForce FX 5950 Ultra 256MB
Detonator 56.56

34GB Western Digital Raptor (10,000RPM, 8MB cache)

Windows XP Professional with Service Pack 1

DirectX 9.0b

Desktop resolution 1024x768, 32-bit color, 85Hz refresh

All power saving options were turned off, as were the Automatic Update and System Restore services. Graphics options under the ‘Performance’ tab were all disabled for maximum performance.

Benchmarks

PC Magazine Business Winstone 2004
PC Magazine Content Creation Winstone 2004
SiSoft Sandra 2004
Futuremark PCMark04
Futuremark 3D Mark03 v.340
Novalogic Comanche 4
Square Enix Final Fantasy XI Benchmark 2
Epic Unreal Tournament 2003 Demo
Epic Unreal Tournament 2004 Demo
Ubisoft IL2 Sturmovik: Forgotten Battles
id Software Quake III v.1.32
Yeti Studios Gun Metal Demo

Lab Notes

You’ll notice that we used the Detonator 55 56.56 driver set, which was the most recent at the time of our testing. Since then, NVIDIA has released the 56.64 driver, which boasts the same general feature set, but is reported to fix a bug NVIDIA discovered in the Unreal Tournament 2004 Demo. At this point, we’ve still not encountered the bug in question, and all Unreal Tournament 2004 testing should be representative of game play using the new driver.


SIDEBAR: Expect to see lots of press releases, as today is the first day of CeBIT.

PC Magazine Benchmarks

Page:: ( 4 / 16 )

Content Creation Winstone 2004 and Business Winstone 2004

Notes

When we first explored the Prescott core last month, we were disappointed to find performance, for the most part, comparable to or slightly behind the older Northwood. In Business Winstone 2004, it didn’t matter much as all of Intel’s offerings were easily bested by AMD’s Athlon 64 FX-51. This time around, the FX-53 furthers that advantage; even the “mainstream” Athlon 64 3400+ is able to outperform the 3.4GHz Extreme Edition.

Another test that comes with the Business Winstone package is a specialized Multitasking Test, which runs the same applications as Business Winstone 2004, only it runs some of them in the background. We had originally expected Intel’s Pentium 4 to dominate here, compliments of its Hyper-Threading technology, but the Athlon 64 FX-53 pulled a surprising victory over even the fastest Extreme Edition.

Content creation is one of the Pentium 4’s purported fortes, and it does quite a bit better in PC Magazine’s Content Creation 2004 metric. Nevertheless, both Athlon 64 FX processors are able to fly right past Intel’s flagship, the FX-53 by more than nine percent.


SIDEBAR: Intel is rumored to be adopting a model number scheme for its upcoming Pentium 4 processors.

SiSoft Sandra 2004

Page:: ( 5 / 16 )

SiSoft Sandra 2004

Notes

The arithmetic tests in Sandra 2004 have always seemed to favor Intel’s NetBurst architecture. The Gallatin-based Extreme Edition responds very well to the first of three standard Sandra 2004 benchmarks, as AMD’s entire processor lineup lags behind significantly.

The difference between Intel’s NetBurst and the AMD64 architecture is much more pronounced in the multimedia test, where the Pentium 4 processor proves its SSE2 alacrity.

Sandra’s memory bandwidth test naturally favors the Athlon 64 FX-53, which benefits from its integrated memory controller and dual-channel memory bus. The Pentium 4 processor follows by about one gigabyte per second, while the Socket 754 Athlon 64 3400+ comprises the rear, limited by its single-channel bus.


SIDEBAR: You can find more about Sandra 2004 here.

PCMark04

Page:: ( 6 / 16 )

Futuremark PCMark04

Notes

Unlike PC Magazine’s Winstone tests, which are derived from real-world applications controlled by a script, Futuremark’s PCMark04 is a synthetic test that, in its default form, measures overall system performance. As with SiSoft Sandra 2004, the Pentium 4 Extreme Edition does very well here, leading the new Athlon 64 FX-53 by more than 12 percent. We’ll see if that synthetic differential carries over into real-world applications, though.


SIDEBAR: With the release of PCMark 04, FutureMark no longer supports PCMark 02.

3DMark03

Page:: ( 7 / 16 )

Futuremark 3Dmark03 v.340

Notes

3DMark03, another synthetic Futuremark application, is similarly unkind to AMD’s Athlon 64 architecture. The 3.4GHz Pentium 4 takes a notable seven percent lead in the 3D portion of the test, while the slower FX-51 and Athlon 64 3400+ fall in place further down the line.

The CPU test is a different story, as AMD’s Athlon 64 FX-53 claims a first place finish. The FX-51 comes in second, followed by Intel’s 3.4GHz Pentium 4 Extreme Edition, and finally by the Athlon 64 3400+.


SIDEBAR: FutureMark is expected to released 3DMark 04 soon.

Comanche 4

Page:: ( 8 / 16 )

Novalogic Comanche 4 Demo

Notes

We’re doing something a little different today – rather than run exclusively low-resolution tests, we wanted to mix in a more realistic gaming resolution. At 800x600, it’s easy to see the profound impact of processor performance. Comanche 4 is notorious for demonstrating particular sensitivity to CPUs, and it does so once again here. AMD’s Athlon 64 FX-53 bags another win, followed by the FX-51, 3.4GHz Extreme Edition, and 3400+, in that order.

At 1600x1200 the difference is a bit more nebulous. Intel’s Pentium 4 Extreme Edition still takes third place, but it does so by a much narrower margin. Undoubtedly, enabling anti-aliasing and anisotropic filtering would balance the field even further, transferring dependence wholly onto the graphics card.


SIDEBAR: The US Army recently cancelled the Comanche, guess we’ll be seeing an Apache simulator from them next.

Unreal Tournament 2004 Demo

Page:: ( 9 / 16 )

Epic Unreal Tournament 2004 Demo

Notes

Once again, there’s a measurable performance difference between all four platforms at 800x600 using the Colossus botmatch test included with Epic’s Unreal Tournament 2004 Demo. The Athlon 64 FX-53 performs particularly well, and is followed by the FX-51, Athlon 64 3400+, and finally, Intel’s 3.4GHz Pentium 4 Extreme Edition. At 1600x1200, however, the delta between AMD Athlon 64 FX-53 and Intel’s flagship is down to six percent, though – not nearly as large as the 15 percent difference that separated the two at 800x600.


SIDEBAR: Gamespot has already given UT2004 a 5 out of 5. Sheesh.

Final Fantasy XI

Page:: ( 10 / 16 )

Square Enix Final Fantasy XI Benchmark 2

Notes

The Final Fantasy test runs in a preset time frame, outputting the number of frames successfully rendered. The results here are odd, yet reproducible. For some reason, the Athlon 64 3400+ takes a first place finish, almost two percent faster than the Athlon 64 FX-53 and an impressive 12 percent in front of the Pentium 4 Extreme Edition at 3.4GHz.


SIDEBAR: Opteron recently got another design win with IBM.

Unreal Tournament 2003

Page:: ( 11 / 16 )

Epic Unreal Tournament 2003 Demo

Notes

In similar fashion to the Unreal Tournament 2004 Demo, we see the Athlon 64 FX-53 dominating the benchmarks. Most interesting, though, is that its advantage is most pronounced in the Flyby test at 800x600, a metric that isn’t representative of actual game play since it lacks other players. When we move to 1600x1200, the differences between AMD’s $400 Athlon 64 3400+ and Intel’s $1,000+ 3.4GHz Pentium 4 Extreme Edition (that isn’t yet available) shrinks to two percent – in favor of the Athlon 64 3400+.


SIDEBAR: Now that UT 2004 is out, hopefully Epic can release their 64-bit version of UT for Windows 64-bit soon.

IL2 Sturmovik

Page:: ( 12 / 16 )

Ubisoft IL2 Sturmovik: Forgotten Battles

Notes

Like Comanche 4, IL2 is a flight simulator with a penchant for performance processors. And as we saw with Comanche 4, the story here is short and sweet. Mainly, if you enjoy running games on the fastest hardware at low resolutions, you may see a difference between the Athlon 64 FX-53 running at 100fps and Intel’s Pentium 4 3.4GHz Extreme Edition (once it becomes available) at 93.

That sort of firepower just begs to be taxed, though. At the very least, you’ll want to run at 1600x1200, perhaps with anti-aliasing to smooth the jagged edges that mar horizon lines. Once the graphics pipeline becomes the bottleneck, these heavyweight processors cease outperforming each other by double digit percentages.


SIDEBAR: Ubisoft recently released the IL-2 Sturmovik: Forgotten Battles Ace Expansion Pack, which adds 20 new flyable aircraft including the P-51 Mustang.

Quake III

Page:: ( 13 / 16 )

id Software Quake III v.1.32 demo ‘four’

Notes

Of all the games in our repertoire, Quake III is the oldest by far. Those 400+ frame rates should be plenty indicative of that. At 1600x1200, though, less then four frames per second separate the four processors in question. That is, there is about one percent separating the top and bottom extremes.


SIDEBAR: Quake 3 was tested with the standard demo.

Gun Metal

Page:: ( 14 / 16 )

Yeti Studios Gun Metal Demo Benchmark 1

Notes

Though it isn’t an actual game just yet, the Gunmetal benchmark illustrates the graphical constraints of an application developed with stunning eye candy (it doesn’t hurt that the benchmark mode employs 2x anti-aliasing by default). Even at 800x600, there is minimal margin between the competing processors. At 1600x1200 they all score virtually the same, despite the GeForce FX 5950 Ultra used in the test system.


SIDEBAR: March has definitely been the month of the shooter.

Ballistics Report: Athlon 64 FX-53

Page:: ( 15 / 16 )

Pros:

Speed: There’s simply no way around it; AMD’s Athlon 64 FX-53 is the fastest processor you can buy. It may cost in excess of $700 and it may require that you buy a 940-pin motherboard and registered memory, but it puts enough distance between itself, Intel’s 3.4GHz Pentium 4 Extreme Edition and AMD’s own Athlon 64 3400+ to definitively land at the top. Keep in mind that is doesn’t distinguish itself so much in high-res gaming environments as it would in content creation, such as video editing.

Unlocked multiplier: Understandably, an unlocked multiplier is a boon to the zealous overclocking enthusiast, especially when Athlon 64 FX motherboards aren’t known to exhibit much system bus flexibility. The clear caveat to that is you’ll need a beefier cooling solution if you’re really looking to push the processor. It runs relatively hot in stock form, and it just doesn’t want to cooperate above 2.5GHz or so.

Cons:

Price: According to AMD, the Athlon 64 FX-53 is actually displacing the FX-51 at $733. In essence, the FX-51 will cease to be, outside of the supply still on the market, which should sell at a lower price. Even if it’s faster than Intel’s 3.4GHz Extreme Edition at $1,000+, which it is, the Athlon 64 FX-53 will still set you back quite a bit.

Platform requirements: If you already own a Socket 940 motherboard and a pair of registered DDR memory modules, then you probably have an Athlon 64 FX-51 plugged in and running, in which case you may not be ready to part ways with another $700 dollars. In the event that you don’t have the necessary components, you’ll need to sink another $400-500.

The Future: Let’s say, for the sake of argument, that you’re willing to spend the money on an Athlon 64 FX-53. You’re also ready to throw down for its accompanying platform. Consider this final thought: it’s no secret that within a couple of months, AMD will unveil its Socket 939 interface. It won’t require registered memory modules, it may perform a bit faster than today’s Socket 940 systems, and it will support a long line of Socket 939 processors in AMD’s roadmap.


SIDEBAR: AMD sponsored the Ghostrider robot in the recent DARPA challenge. Ghostrider is a self navigating robot motorcycle.

Final Verdict

Page:: ( 16 / 16 )

Print Article! | Close Window ]