We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.

[nextpage title=”Introduction”]

Athlon 64 FX-57 is the latest and fastest single-core CPU from AMD. In this review we will compare its performance with the previous model of Athlon 64 FX, FX-55, with Athlon 64 4000+, which is the fastest single-core CPU on Athlon 64 line, with other CPUs from AMD and also with some CPUs from Intel, like Pentium 4 3.2 GHz, Pentium 4 3.4 GHz, Pentium 4 Extreme Edition 3.2 GHz and Pentium 4 Extreme Edition 3.4 GHz. The only model we didn’t have available for comparison was the 3.6 GHz Pentium 4. Also, we won’t compare the new Athlon 64 FX-57 with dual-core CPUs, we will keep that for the Athlon 64 X2 and Pentium D reviews, as soon as AMD and Intel send us samples for reviewing.

Figure 1: Athlon 64 FX-57 we reviewed.

Athlon 64 FX-57 uses the socket 939 pinout, the same platform used by Athlon 64 FX-55, Athlon 64 FX-53 and newer Athlon 64 CPUs like Athlon 64 4000+ and Athlon 64 3800+. The main feature of this platform is the Dual DDR Channel support, which doubles the transfer rate of RAM memory compared to socket 754 Athlon 64 CPUs (6.400 MB/s vs. 3.200 MB/s if DDR400 is used).

Figure 2: Athlon 64 FX-57, bottom view. It uses socket 939.

The new Athlon 64 FX-57 runs at 2.8 GHz and has a 1 MB L2 memory cache. The previous model, Athlon 64 FX-55, runs at 2.6 GHz. The fastest Athlon 64, Athlon 64 4000+, runs at 2.4 GHz, the same clock rate as Athlon 64 FX-53. Actually Athlon 64 FX-53 and Athlon 64 4000+ are the same CPU.

The difference between Athlon 64 FX and Athlon 64 is the fact that Athlon 64 FX series doesn’t have its clock multiplier locked, allowing more overclocking possibilities. The complete specs of the Athlon 64 FX-57 we reviewed you can check in Figure 3. By the way, our ABIT AV8 motherboard does by default a small overclocking (HTT is set as 204 MHz). We’ve corrected this to 200 MHz in all CPUs we installed in order to have honest results in our benchmarking.

Figure 3: Specs of the Athlon 64 FX-57 we reviewed.

In the table below you can see the specs of all Athlon 64 FX CPUs released so far.

CPU	Internal Clock	L1 Cache	L2 Cache	Socket	Hyper Transport	DDR Dual Channel
Athlon 64 FX-51	2.2 GHz	128 KB	1 MB	940	800 MHz	Yes, Registered
Athlon 64 FX-53	2.4 GHz	128 KB	1 MB	940	800 MHz	Yes, Registered
Athlon 64 FX-53	2.4 GHz	128 KB	1 MB	939	1 GHz	Yes
Athlon 64 FX-55	2.6 GHz	128 KB	1 MB	939	1 GHz	Yes
Athlon 64 FX-57	2.8 GHz	128 KB	1 MB	939	1 GHz	Yes

The very first Athlon 64 FX processor lauched, FX-51, used the socket 940 pinout. This platform is the same used by the Opteron CPU, which is targeted to the server market. The problem with this platform is that it requires registered memories, which are more expensive and a little bit slower than conventional memories. From FX-53 model on Athlon 64 FX uses the socket 939 platform, which is the same platform used by newer Athlon 64 CPUs.

Talking a little bit about pricing, on https://www.pricewatch.com you can find the Athlon 64 FX-57 at USD 1138, on average. Athlon 64 FX-55 is quoted at USD 915, on average, while Athlon 64 4000+ is quoted at USD 586, also on average.

[nextpage title=”How We Tested”]

During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions the only different device was the CPU being tested and also the motherboard, since AMD processors use a different socket from Intel processors.

Hardware Configuration

• Motherboard (Athlon 64 socket 754): Chaintech ZNF3-150 (nForce 3-150, October 13th, 2003 BIOS)
• Motherboard (Athlon 64 socket 939): ABIT AV8 (VIA K8T800, June 20th, 2005 2.1 BIOS)
• Motherboard (Athlon XP): Gigabyte 7NNXP (nForce 2 400 Ultra, September 12th, 2003 F14 BIOS)
• Motherboard (Pentium 4): Soyo P4I875P (Intel 875P, August 19th, 2003 BIOS)
• Motherboard (Pentium 4 “Prescott”): Intel D875PBZ (Intel 875P, P21 BIOS)
• Memory: Two 256 MB PC3200 modules from TwinMOS, in DDR Dual Channel configuration on Soyo P4I875P, Intel D875PBZ, Gigabyte 7NNXP and ABIT AV8 motherboards.
• Hard Drive: Maxtor DiamondMax 9 Plus (60 GB, ATA-133)
• Video Card: Gigabyte Radeon 9800 Pro
• Screen Resolution: 800x600x32

Software Configuration

• Windows XP Professional installed using NTFS
• Service Pack 1A
• Direct X 9.0B
• ATI video driver version: 7.95 (6.14.10.6396)
• Intel Inf Driver: 5.0.2.1003
• NVIDIA nForce driver version: 3.13
• VIA Hyperion driver version: 4.55 VP1

Used Software

• SYSmark2002
• PCMark04 Business
• 3DMark2001 SE
• 3DMark03 Pro 3.40
• Quake III Arena 1.32
• Sandra 2004 9.89

We adopted a 3% error margin; thus, differences below 3% cannot be considered relevant. In other words, products with a performance difference below 3% should be considered as having similar performance.

[nextpage title=”Processing Performance: PCMark04″]

PCMark04 measures the system performance by running a set of 10 tasks like video compression, audio conversion, file cryptography, calculations, etc, giving a numeric score. The results from this test we show on the graph below.

Athlon 64 FX-57 achieved a performance at the same level of Pentium 4 Extreme Edition 3.4 GHz on this program, being 6.38% faster than Pentium 4 3.4 GHz, 8.18% faster than Athlon 64 FX-55, 11.24% faster than Pentium 4 3.2 GHz “E” (“Prescott”), 11.53% faster than Pentium 4 3.2 GHz, 14.88% faster than Athlon 64 4000+, 16.20% faster than Athlon 64 3800+, 16.70% faster than Pentium 4 3 GHz, 35.91% faster than socket 754 Athlon 64 3200+ and 41.75% faster than Athlon XP 3200+.

[nextpage title=”Overall Performance: SYSmark2002″]

We measured the overall performance of this motherboard using SYSmark2002, which is a software that simulates the use of real-world applications. Thus, we consider this the best software to measure, in practical terms, the system performance.

The benchmarks are divided into two groups: Internet Content Creation and Office Productivity. The programs used on Internet Content Creation batch are the following: Dreamweaver 4.0, Photoshop 6.0.1, Premiere 6.0, Flash 5 and Windows Media Encoder. The programs used on Office Productivity batch are the following: Word 2002, Excel 2002, PowerPoint 2002, Outlook 2002, Access 2002, Netscape Communicator 6, NaturallySpeaking 5, VirusScan 5.13 and WinZip 8.0.

The software delivers specific results for each batch and also an overall performance result, all in a specific SYSmark202 unit.

You can check the results on the graph below.

Athlon 64 FX-57 achieved an overall performance similar to Athlon 64 FX-55 and Pentium 4 Extreme Edition 3.4 GHz. It was 6.59% faster than Pentium 4 3.4 GHz, 6,91% faster than Pentium 4 3.2 GHz, 7.23% faster than Athlon 64 4000+, 10.90% faster than Pentium 4 Extreme Edition 3.2 GHz, 12.66% faster than Pentium 4 3.2 GHz “E” (“Prescott”), 13.38% faster than Pentium 4 3 GHz, 14.10% faster than Athlon 64 3800+, 27.60% faster than socket 754 Athlon 64 3200+ and 35.88% faster than Athlon XP 3200+.

In Office Productivity test, Athlon 64 FX-57 achieved a performance similar to Athlon 64 FX-55 and Athlon 64 4000+, being 5.74% faster than Pentium 4 Extreme Edition 3.4 GHz, 8.40% faster than Pentium 4 3.2 GHz, 11.21% faster than Pentium 4 3.4 GHz and Athlon 64 3800+, 15.70% faster than Pentium 4 3.2 GHz “E” (“Prescott”) and Pentium 4 3 GHz, 17.27% faster than Pentium 4 Extreme Edition 3.2 GHz, 24,04% faster than socket 754 Athlon 64 3200+ and 25.24% faster than Athlon XP 3200+

Analyzing Internet Content Creation results alone, Athlon 64 FX-57 achieved a performance similar to Pentium 4 3.4 GHz, being 4.48% faster than Pentium 4 Extreme Edition 3.2 GHz, 5.15% faster than Pentium 4 3.2 GHz, 7.46% faster than Athlon 64 FX-55, 9.62% faster than Pentium 4 3.2 GHz “E” (“Prescott”), 10.86% faster than Pentium 4 3 GHz, 14.75% faster than Athlon 64 4000+, 16.95% faster than Athlon 64 3800+, 30.67% faster than socket 754 Athlon 64 3200+ and 46.71% faster than Athlon XP 3200+. In this test Pentium 4 Extreme Edition 3.4 GHz was 4.29% faster than Athlon 64 FX-57.

[nextpage title=”3D Performance: 3DMark2001 SE”]

3Dmark2001 SE measures video card performance simulating DirectX 8.1 games. It is very effective software for evaluating the performance from previous-generation games, programmed using DirectX 8. On the graph below you see our benchmarking results.

Athlon 64 FX-57 achieved a performance similar to Athlon 64 FX-55, which was 4.55% faster than Athlon 64 4000+, 6.77% faster than Athlon 64 3800+, 8.86% faster than Pentium 4 E.E. 3.4 GHz, 9.84% faster than Pentium 4 E.E. 3.2 GHz, 11.68% faster than socket 754 Athlon 64 3200+, 15.54% faster than Pentium 4 3.4 GHz, 17.75% faster than Pentium 4 3.2 GHz, 19.34% faster than Pentium 4 3.2 GHz “E” (“Prescott”), 22.56% faster than Pentium 4 3 GHz and 27.87% faster than Athlon XP 3200+.

[nextpage title=”3D Performance: 3DMark03″]

3Dmark03 measures performance by simulating games written to DirectX 9, which are contemporary games. On the graph below you see our benchmarking results.

On 3DMark03 Athlon 64 FX-57 achieved a performance similar to the majority of high-end processors we’ve tested. There were meaningful performance differences only when compared to Pentium 4 3.2 GHz (3.10%), socket 754 Athlon 64 3200+ (3.50%), Pentium 4 3 GHz (4.05%) and Athlon XP 3200+ (6.48%).

[nextpage title=”3D Performance: Quake III”]

Even though Quake III is an old game, it continues to be an excellent system benchmarking program, since this game is very sensitive to any memory and memory cache change.
 
We used the demo four available on version 1.32 of Quake III to make our benchmarking with this game. We ran this demo three times at the game’s default configuration and we picked the middle value for our comparisons, i.e., we discarded the highest and the lowest values.

The results, in frames per second, you can see on the graph below.

Athlon 64 FX-57 performance on this game was really impressive: 9.15% faster than Athlon 64 4000+, 11.53% faster than Athlon 64 FX-55, 15.44% faster than Pentium 4 E.E. 3.4 GHz, 17.21% faster than Pentium 4 E.E. 3.2 GHz, 27.66% faster than Athlon 64 3800+, 32.88% faster than socket 754 Athlon 64 3200+, 35.76% faster than Pentium 4 3.4 GHz, 41.29% faster than Pentium 4 3.2 GHz, 48.45% faster than Pentium 4 3.2 GHz “E” (“Prescott”), 55.13% faster than Pentium 4 3 GHz and 110.55% faster than Athlon XP 3200+.

[nextpage title=”Memory Performance”]

Since on Athlon 64 who controls memory access is the CPU and not the chipset, it is very interesting to compare the memory performance between AMD and Intel CPUs.

This test was done with the aid of Sandra 9.89 program and the results, shown on the graph below, are in MB/s. Since we used two DDR400/PC3200 modules, the maximum theoretical performance of this memory was 3,200 MB/s in single channel and 6,400 MB/s in dual channel. We included these rates on the chart below for comparison.

The embedded memory controller on Athlon 64 processors are far better than memory controllers located on the chipset, allowing the memory transfer rate finally go near the theoretical 6,400 MB/s DDR Dual Channel transfer rate. Athlon 64 FX-57 achieved a 5,728 MB/s transfer rate, using 89.50% of the available bandwidth.

To understand what this number means, the Intel CPU that reached the best memory access performance was Pentium 4 3.2 GHz “E” (“Prescott”), with a 4,889 MB/s transfer rate, a 76.39% use of the available bandwith (Athlon 64 FX-57 acessed RAM memory 17.16% faster than this processor).

Funny enough, the memory performance of Athlon 64 FX-57 was lower than all other socket 939 processors we’ve tested.

[nextpage title=”Overclocking”]

Our overclocking tests were made with the CPU using its original cooler and standard voltage (1.425 V).

As we mentioned, the main difference between Athlon 64 and Athlon 64 FX is that FX has its clock multiplier unlocked, allowing you to make more aggressive overclocking. Actually Athlon 64 clock multiplier is unlocked, but only for configuring it below its standard ratio. On Athlon 64 FX you can go both ways: up and down.

First we increased its HTT clock using its standard clock multiplier (14x). We could achieve only 209 MHz, making the CPU to run at 2,926 MHz internally, a 4.5% increase on the CPU internal clock. Over 209 MHz the system was unstable.

Then we increased the clock multiplier to 14.5x, making the CPU run at 2.9 GHz, a 3.57% increase in its internal clock. We couldn’t increase HTT, the system was always unstable.

Then we played descreasing the clock multiplier to 13.5x. At this configuration we could put the HTT clock at 216 MHz, making the CPU run at 2,916 MHz internally, a 4.14% increase on the CPU internal clock). Over that the system was unstable.

We didn’t play with voltage or with the cooler. You will probably achieve better results than us with time and patience.

[nextpage title=”Conclusions”]

We were really impressed by the Athlon FX-57 performance. It is a very good high-end CPU. On CPU-intensive games, Athlon 64 FX-57 is the best single-core CPU available on the market today. But we have to keep in mind that we haven’t got a 3.6 GHz Pentium 4 to include on our benchmarking.

For overclockers, this CPU is a dream, since it has its clock multiplier unlocked.

The only problem with Athlon 64 FX-57, however, is its price. Costing USD 1,138 in the US, on average, it is not a lot more expensive than Athlon 64 FX-55 (around USD 915), but comparing it to Athlon 64 4000+, which costs almost half the price (USD 586, on average), makes us think who is going to buy a CPU like this.

For the hardcore gamer, Athlon 64 4000+ will deliver the best cost/benefit ratio, but if you want to build the ultimate gaming machine and money isn’t an issue, you should go for an Athlon 64 FX-57 with a GeForce 7800 GTX.

We also have to mention that AMD has done a terrific job launching socket 939; that was the ace up AMD’s sleeve needed to compete with Intel on the high-end computer market.