Review

AMD Athlon™ 64 4000+ and FX-55 Review

by John Reynolds

Introduction

AMD is announcing today two new processors to their Athlon 64 series of CPUs, the Athlon 64 4000+ and the Athlon 64 FX-55. Both processors are 939-pin, micro-PGA parts and are manufactured at AMD’s Dresden, Germany, fabrication plant using a 130nm SOI process. The FX-55, with its 2.6 GHz clock speed, is a speed stepping of 200 Mhz above the FX-53 and its 2.4 GHz frequency, with both parts sharing the same amount of cache, 128 KB (64 data + 64 instruction) of L1 and 1 MB of L2 cache. The Athlon 64 4000+, however, has an identical clock speed to that of the 3800+ reviewed earlier this summer, which is a core frequency of 2.4 GHz. Interestingly enough, the 4000+’s L2 cache has been doubled in size to 1 MB to match that of the FX chips. The chart below shows the different specifications of the CPUs that will be tested in this review.

It is somewhat surprising to see AMD increasing the L2 cache of the Athlon 64 4000+ rather than its frequency because of the increase in transistors this has added to the processor. Increased die size translates into fewer chips per wafer for the manufacturing process being used, which obviously means fewer processors to sell to OEMs or the retail market. AMD’s roadmap still lists a 90nm Athlon 64 core, code named Winchester, due the second half of this year, so once available we should see clock speed increases in addition to the 1 MB L2 cache as the cores are migrated to the new process. The move of the Athlon 64 FX processors to 90nm is not expected, however, until early next year. Until then, the 4000+ is a cache rather than clock speed increase over the 3800+.

In addition to the integrated memory controller and its 6.4 GB of bandwidth the Athlon 64s boast, AMD is also continuing to heavily market EVP (Enhanced Virus Protection) and C’n’Q (Cool‘n’Quiet). Also known as NX (no execute), EVP is hardware support designed to prevent the running of code that attempts to execute in a memory area marked as a data page, which is a common tool used by many malware authors. To enable EVP requires Windows XP with Service Pack 2 to be installed. C’n’Q works by monitoring system usage (CPU cycles) and dynamically adjusting the processor’s frequency and voltage down to reduce heat during periods of lessened workload. Noise is also reduced using an on-die thermistor to detect lower temperatures, at which point the fan speed on the processor is slowed. Once the temperature rises the fan will resume its full speed. Care was taken during testing to ensure that C’n’Q was disabled to avoid inaccurate performance scores.

The Athlon 64 4000+ and FX-55’s initial pricing are slightly higher than that of AMD’s earlier releases, priced in bulk at $729 and $827 respectively. In comparison, the 3800+ was initially priced at $720 and the FX-53 $799.

The AMD Athlon™ 64 Processor

The AMD Athlon™ 64 FX Processor

Go To Page 2

AMD Athlon™ 64 4000+ and FX-55 Review

Back to Page 1

Test System Setup

• AMD Athlon 64 3800+, 4000+, and FX-55 processors
• MSI K8N Neo2 motherboard (nForce 5.10)
• 1 GB (2 x 512 MB) Corsair XMS PC3200 DDR RAM
• VisionTek 9800 Pro 128 MB (Catalyst 4.9 drivers)
• Adaptec 19160 SCSI controller
• 36 GB Seagate Cheetah 15,000 RPM HD (NTFS)
• Windows XP Professional - Service Pack 2
• DirectX 9.0c

The benchmark suite used to test the Athlon 64 3800+, 4000+, and FX-55 is listed here. All games are configured for 32-bit color and trilinear texture filtering as the baseline default. Anti-aliasing and anisotropic texture filtering are, of course, disabled throughout all tests. Windows XP is also configured to have Automatic Update, System Restore, and all unnecessary startup services disabled. Fraps v2.3.2 was used to record performance scores unless otherwise noted. However, because a faster AGP graphics board was not available for use in the test system, in-game options for each title were changed to lower, less demanding graphics settings to keep the 9800 Pro from becoming a performance bottleneck and thus preventing the tested processors from differentiating themselves. Once PCI Express chipsets for AMD systems become available here in the near future, this will no longer be an issue since SimHQ’s ‘hardware lab’ has several high-end PCIe graphics boards available for testing.

Scores for the Athlon 64 3800+ in this review are not comparable to those from SimHQ’s original review of the CPU because, as mentioned above, testing was conducted with lower in-game settings, updated Catalyst drivers from ATI were used, and a different motherboard, MSI’s K8N Neo2, which is based on NVIDIA’s nForce 3 Ultra chipset, was installed in the test system; the 3800+ review’s test system used ASUS’ A8V Deluxe mainboard, based on VIA’s K8T800 Pro chipset.

Benchmark Scores

The Comanche 4 benchmarking demo was run with texture compression enabled and hardware shaders and sound disabled.

A piece of software that has traditionally scaled very closely with CPU speed, the C4 demo displays a consistent performance increase for each resolution and processor. In fact, the FX-55 is the first CPU to break the 80+ frame rate barrier in SimHQ’s testing.

Lock On: Modern Air Combat was tested using the first three minutes of the MiG-29 Intercept demo. The in-game graphics default setting of low was applied in the hopes of keeping the 9800 Pro card from becoming a performance choke-point in the test system.

Without a doubt the graphics card is refusing to allow LOMAC to scale with processor changes.

Next is Microsoft’s Flight Simulator 2004, the ruling king of civilian aviation simulation. Testing consisted again of SimHQ’s dusk flight over the city of Hong Kong, with an external camera view positioned behind the aircraft. The game’s display settings were placed at medium low and all box options were unchecked.

Unfortunately, in the effort to prevent the slower graphics card from bottlenecking the test systems, FS2004 appears to have capped itself at 100 fps. The in-game frame rate option was set to Unlimited and the display's refresh rate was not set at 100 Hz. Further experimentation with in-game graphics settings, such as placing each sub-section at high, still saw the frame rate pegged at 100; in fact, not until the hardware options were placed at near-maximum settings did the frame rate begin to drop.

FS2004 also scales fairly well with faster processors once the graphics settings are dialed down a few notches as noted above. The game appears to benefit more from the FX-55’s higher frequency than the 4000+’s increased L2 cache, showing at 800x600 a 12 frame improvement for the FX-55 over the 4000+ compared to the latter’s 8 fps gain over the 3800+. This pattern also continues at 1024x768.

Go To Page 3

AMD Athlon™ 64 4000+ and FX-55 Review

Back to Page 2

IL-2: Sturmovik Forgotten Battles - Aces Expansion Pack was tested in OpenGL and with all video options set to medium (normal for Objects detail) using the Black Death track.

For each processor change, IL-2 shows roughly a 5% performance gain for the tested resolutions, with AMD’s latest CPUs hitting a 3-digit frame rate at 640x480.

Falcon 4 performance testing again used the FreeFalcon 3 upgrade mode with SimHQ’s in-house test of a low level, air-to-ground flight that consists of two Falcons using Mk20s and Mavericks. Graphics options, however, were left at their highest settings since the title can hardly be considered as stressful to the test rig’s graphics sub-system.

With the FF3 mode installed, F4 essentially refuses to scale for both resolution change and for the 4000+. For the FX-55’s faster core frequency, however, the game showed a surprising 15% frame rate improvement at 640x480, and settled into roughly 8% for the higher resolutions.

Far Cry is undoubtedly the most graphically advanced title in SimHQ’s benchmark suite. As such, all in-game advanced video options were set at medium. Testing consisted of repeated run-throughs of the Research map in God mode since the level includes an excellent combination of the beach, jungle, and interior settings found throughout the game.

Far Cry displays a performance delta between the tested CPUs of roughly 3-4%, indicating that even with medium graphics settings the title’s frame rate is still possibly limited by the test system’s 9800 Pro, though the scores are certainly much higher compared to when higher in-game settings are enabled.

Developed using id Software’s five-year-old Quake 3 engine, Call of Duty (v1.4) is the second title SimHQ uses testing OpenGL rather than the D3D API. Because the game is based on such an aged engine, graphics settings were left unchanged. Scores were derived from the Dawnville demo using the in-game timedemo utility to capture performance. The “com_maxfps” console command was also used to lift the default frame rate cap of 85.

As usual, Call of Duty scales extremely well with hardware changes. Worth noting is the game appears to benefit more from the FX-55’s clock speed compared to the 4000+’s larger cache.

Go To Page 4

AMD Athlon™ 64 4000+ and FX-55 Review

Back to Page 3

Last is NASCAR Racing 2003 Season, tested using a crowded Daytona track and a camera view set inside of Earnhardt’s cockpit. Graphics settings were also configured to medium options and shadows disabled.

At the two lower resolutions, the FX-55 shows more improvement over the 4000+ than the latter does over the 3800+, though the difference in gains is marginal at 640x480. Like most other titles not bottlenecked by the graphics card, NASCAR displays close to 5% difference between CPUs.

Gallery

IL-2: Sturmovik - Forgotten Battles - AEP (250 kb jpg)	Lock On: Enemy Air Combat (2.25MB bmp)

Conclusion

SimHQ’s benchmark suite displayed a fairly uniform 5% performance gain across the board for AMD’s new processors over their slower parts, with the FX-55’s clock speed increase allowing the CPU to scale itself slightly better than the 4000+. The Athlon 64 family of processors has proven that they can dominate in a 32-bit gaming environment, though the performance they offer is certainly not as inexpensive as past Athlons have been alongside comparable Intel parts. It will be interesting to see if the migration to the 90nm process and its smaller die sizes will have any impact on Athlon 64 pricing; and with Winchester still due this year and San Diego (the 90nm FX core) sometime next, it will also be interesting to see if AMD releases any additional 130nm processors before moving the entire Athlon 64 family to the new process. And of even more excitement to hardware enthusiasts, there is of course the expectations for dual-core processors that both AMD and Intel are currently striving to bring first to the market late next year.

The MSI K8N Neo2, using the nForce 3 Ultra chipset, proved itself to be as stable as the A8V Deluxe motherboard used earlier this year. Though the unfortunate inclusion of a 9800 Pro in the test system clearly held the Athlon 64 4000+ and FX-55 from reaching higher performance levels in several of the simulations tested, the two processors only strengthen AMD’s performance lead over the competition. And with their release, the new CPUs should help trickle the pricing on previously introduced Athlon 64s down. Yet for the simulation gaming fan who feels compelled to build the absolute fastest rig available today, they need look no further than the Athlon 64 FX-55.

Download a pdf of this article here (x kb).

The "How SimHQ Tests" page is here.
The page includes our test criteria and links to the new Benchmark Suite.

Click here to go to top of this page.

Copyright 2008, SimHQ.com. All Rights Reserved. Contact the webmaster.