Athlon 64 FX-62 Dual-Core Socket AM2 Review

[nextpage title=”Introduction”]

Athlon 64 FX-62 is right now the most high-end CPU from AMD, incorporating dual-core technology, socket AM2 (i.e., DDR2 memory support), 2.8 GHz internal clock and 1 MB L2 memory cache for each core. We had the chance of reviewing this beast and compare it to other dual-core CPUs from AMD: Athlon 64 X2 5000+ (socket AM2, 2.6 GHz, 512 KB L2 memory cache), Athlon 64 X2 4600+ (socket 939, 2.4 GHz, 512 KB L2 memory cache) and Athlon FX-60 (socket 939, 2.6 GHz, 1 MB L2 memory cache). Let’s see how this new CPU performs.

Figure 1: Athlon 64 FX-62.

Since it is based on a new pinout, you will need to replace your motherboard by a socket AM2 one if you want to use DDR2 memories with Athlon 64.

The memory controller integrated on AM2 CPUs can support DDR2-533, DDR2-667 and DDR2-800 memories. The problem, however, is how the memory bus clock is achieved. Instead of being generated through the CPU base clock (HTT clock, which is of 200 MHz), it divides the CPU internal clock. The value of this divider is half the value of the CPU multiplier.

For example, an AMD64 CPU with a clock multiplier of 12x will have a memory bus divider of 6. So this CPU will work at 2.4 GHz (200 MHz x 12) and its memories will work at 400 MHz (DDR2-800, 2,400 MHz / 6). Keep in mind that DDR and DDR2 memories are rated with double their real clock rate.

The problem is when the CPU clock multiplier is an odd number. For an AM2 CPU with a clock multiplier of 13x, theoretically its memory bus divider would be of 6.5. Since the AMD64 memory bus doesn’t work with “broken” dividers, it is rounded up to the next rounded number, seven in this case. So while this CPU will work at 2.6 GHz (200 MHz x 13), its memory bus will work at 371 MHz (742 MHz DDR) and not at 400 MHz (800 MHz DDR), making the CPU to not achieve the maximum bandwidth the DDR2 memory can provide.

This is the case of Athlon 64 X2 5000+. As it works internally at 2.6 GHz multiplying its HTT clock by 13x, it can access memory only up to 742 MHz and not 800 MHz. Athlon 64 FX-62, on the other hand, works internally at 2.8 GHz multiplying its HTT clock by 14x, allowing it to access the memory truly at 800 MHz. We will talk more about this later, as we thought it was important to take a closer look at this issue.
 
Here are some examples:

CPU Internal Clock	CPU Multiplier	Memory Divider	Memory Bus
2.8 GHz	14x	7	800 MHz
2.6 GHz	13x	7	742 MHz
2.4 GHz	12x	6	800 MHz
2.2 GHz	11x	6	733 MHz
2 GHz	10x	5	800 MHz
1.8 GHz	9x	5	720 MHz
1.6 GHz	8x	4	800 MHz

[nextpage title=”Specs”]

As mentioned, Athlon 64 FX-62 is a dual-core CPU running at 2.8 GHz, achieved by multiplying its 200 MHz base clock by 14. It has two 1 MB L2 memory caches, one for each core. The other dual-core AMD CPUs we reviewed recently and included in this benchmark were Athlon 64 X2 5000+, which runs at 2.6 GHz, has two 512 LB L2 memory cache and is also based on socket AM2 platform, Athlon 64 X2 4600+, which runs at 2.4 GHz, has two 512 KB L2 memory caches and is based on socket 939 platform, and Athlon 64 FX-60, which runs at 2.6 GHz and has two 1 MB caches and is also based on socket 939 platform. We also included in our benchmark single-core Athlon 64 3800+, which runs at 2.4 GHz, has a 512 KB memory cache and is based on socket 939 platform.

The difference between dual-core Athlon 64 FX and Athlon 64 X2 is not only the amount of memory cache, but also the fact that all Athlon 64 FX CPUs don’t have their clock multiplier locked, enabling you to increase the CPU clock multiplier in order to achieve higher clock rates. On other CPUs you can only overclock them by increasing the CPU HTT clock (sometimes decreasing the clock multiplier at the same time, as this configuration is possible on all AMD64 CPUs).

In Figure 2, you can see the complete specs from the CPU we reviewed.

Figure 2: Athlon 64 FX-62 specs.

We didn’t compare these CPUs with Intel counterparts (namely Pentium D and Pentium Extreme Edition) because unfortunately Intel didn’t provide us samples for reviewing. A pity.

[nextpage title=”How We Tested”]

During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions the only variable was the processor being tested and the motherboard, as socket AM2 CPUs require a new motherboard.
 
Hardware Configuration

• Socket 939 Motherboard: MSI K8N Diamond Plus (BIOS 1.12, December 22nd, 2005).
• Socket AM2 Motherboard: ASUS M2N32-SLI De Luxe
• Memory: Two Corsair CMX1024-3500LLPRO modules with 1 GB each, installed on DDR Dual Channel configuration (2-3-2-6 timings) with socket 939 CPUs.
• Memory: Four Corsair CM2X512-8500 modules with 512 MB each, installed on DDR2 Dual Channel configuration (using 4-4-4-12 timings) with socket AM2 CPUs.
• Hard Drive: Maxtor DiamondMax 9 Plus (7,200 rpm, 40 GB, ATA-133).
• Video Card: XFX GeForce 7800 GTX.
• Video resolution: [email protected]
• Power Supply: OCZ ModStream 520 W.

Software Configuration

• Windows XP Professional installed using NTFS
• Service Pack 2
• DirectX 9.0c

Driver Versions

• NVIDIA video driver version: 84.21
• NVIDIA nForce 4 SLI X16 driver version: 7.15
• All motherboard drivers were installed

Used Software

• SYSmark2004 – Patch 2
• PCMark05 Professional 1.1.0
• Cinebench 9.5
• 3DMark06 Professional 1.0.2
• Quake 4 – Patch 1.2
• Sandra Lite 2007a 10.98.6.x

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=”Overall Performance: SYSmark2004″]

We measured the overall performance of the CPUs included in this review using SYSmark2004, which is a program 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: Simulates the authoring of a website containing text, images, videos and animations. The following programs are used: Adobe After Effects 5.5, Adobe Photoshop 7.01, Adobe Premiere 6.5, Discreet 3ds Max 5.1, Macromedia Dreamweaver MX, Macromedia Flash MX, Microsoft Windows Media Encoder 9, McAfee VirusScan 7.0 and Winzip 8.1.
• Office Productivity: Simulates the use of an office suite, i.e., simulates sending e-mails, word processing, spreadsheets, presentations, etc. The following programs are used: Adobe Acrobat 5.05, Microsoft Office XP SP2, Internet Explorer 6.0 SP1, NaturallySpeaking 6, McAfee VirusScan 7.0 and Winzip 8.1.

This software delivers several results, all of them using a specific SYSmark2004 unit. First we have a SYSmark2004 overall score. Then we have a group result for each batch listed above. And for each batch, we have specific results: 3D Creation, 2D Creation and Web Publication for Internet Content Creation and Communication, Document Creation and Data Analysis for Office Productivity.

We were eager to know what performance gain an average user would have by using a dual-core Athlon 64 instead of a single-core one. The results you can see on the chart below.

The overall SYSmark2004 score for Athlon 64 FX-62 (2.8 GHz, socket AM2) was 5.68% higher than the one achieved by Athlon 64 FX-60 (2.6 GHz, socket 939), 7.32% higher than the one achieved by Athlon 64 X2 5000+ (2.6 GHz, socket AM2), 19.74% higher than the one achieved by Athlon 64 X2 4600+ (2.4 GHz, socket 939) and 43.08% higher than the one achieved by the single-core Athlon 64 3800+ (2.4 GHz, socket 939).

On Internet Content Creation batch Athlon 64 FX-62 was 6.32% faster than Athlon 64 FX-60, 8.50% faster than Athlon 64 X2 5000+, 18.21% faster than Athlon 64 X2 4600+ and 66.67% faster than Athlon 64 3800+.

On Office Productivity Athlon 64 FX-62 was 5.00% faster than Athlon 64 FX-60, 5.53% faster than Athlon 64 5000+, 21.39% faster than Athlon 64 X2 4600+ and 22.09% faster than Athlon 64 3800+.

The highest performance difference between Athlon 64 FX-62 and Athlon 64 FX-60 was on 3D Creation, 7.21%, and the highest performance difference between the reviewed CPU and Athlon 64 X2 5000+ was 11.55%, on Document Creation.

[nextpage title=”Processing Performance: PCMark05 Professional”]

PCMark05 Professional measures the system performance by running several tests. We selected two batches for our comparisons, System and CPU.

The System batch performs the following tests: HDD XP Startup, Physics and 3D, 2D Transparent Window, 3D Pixel Shader, Web Page Rendering, File Decryption, 2D Graphics Memory – 64 lines, HDD General Usage and three multithreading tests.

The CPU batch performs the following tests: File Compression, File Decompression, File Encryption, File Decryption, Image Decompression, Audio Compression and two multithreading tests.

The results are given in a PCMark05 specific unit.

On System batch Athlon 64 FX-62 was 3.86% faster than Athlon 64 FX-60, 5.66% faster than Athlon 64 X2 5000+, 9.02% faster than Athlon 64 X2 4600+ and 43.66% faster than Athlon 64 3800+.

On CPU batch Athlon 64 FX-62 was 7.94% faster than Athlon 64 FX-60, 8.45% faster than Athlon 64 X2 5000+, 17.59% faster than Athlon 64 X2 4600+ and 65.34% faster than Athlon 64 3800+.

[nextpage title=”Rendering Performance: Cinebench 9.5″]

Cinebench 9.5 is based on the 3D software, Cinema 4d. It is very useful to measure the performance gain given by having more than one CPU installed on the system when rendering heavy 3D images.

This software provides five results, Rendering 1 CPU, which measures the rendering performance using just one CPU, Rendering x CPUs, which measures the rendering performance using all CPUs available on the system, Cinema 4D shading, OpenGL Software Lighting and OpenGL Hardware Lighting. Since we were interested in measuring the rendering performance, we are going to compare the “Rendering x CPUs” results from all CPUs (“Rendering 1 CPU” in the case of Athlon 64 3800+).

Here Athlon 64 FX-62 was 7.87% faster than Athlon 64 FX-60 and Athlon 64 X2 5000+, 16.74% faster than Athlon 64 X2 4600+ and 115.75% faster than Athlon 64 3800+.

[nextpage title=”3D Performance: 3DMark06 Professional”]

3DMark06 is the latest version of 3DMark franchise, measuring Shader 3.0 (i.e., DirectX 9.0c) performance. We run this software on its default configuration (1280×1024 resolution with no image quality settings enabled), checking the CPU batch results for comparison.

To be honest, 3D performance nowadays depends much more on the video card used than on the system CPU. This is so true that Athlon 64 FX-62, Athlon 64 FX-60 and Athlon 64 X2 5000+ achieved the same performance level on this test, while Athlon 64 FX-62 was 17.41% faster than Athlon 64 3800+.

When comparing the results for the CPU tests alone, Athlon 64 FX-62 was 8.39% faster than Athlon 64 FX-60, 10.64% faster than Athlon 64 X2 5000+, 16.76% faster than Athlon 64 X2 4600+ and 124.66% faster than Athlon 64 3800+. Once again, this shows how the improvement on CPU performance won’t probably reflect on a higher 3D performance (however, as we’ve seen from the overall score, dual-core will reflect on a higher 3D performance compared to single-core technology).

[nextpage title=”3D Performance: Quake 4″]

We ran Quake 4 multiplayer demo id_demo001 on 1024x768x32 with no image quality settings enabled. We run it four times and the results shown on the chart is an arithmetic average of the collected data. The results are in frames per second. For more information on how to use Quake 4 to benchmark a PC, read our tutorial on this subject.

On Quake 4 Athlon 64 FX-62 was 11.01% faster than Athlon 64 X2 5000+, 11.47% faster than Athlon 64 FX-60, 20.57% faster than Athlon 64 X2 4600+ and 29.59% faster than Athlon 64 3800+.

[nextpage title=”Memory Bandwidth: Sandra Lite 2007″]

We decided to include a memory bandwidth benchmark on this review, since depending on the CPU model the memory bus won’t be running at full speed, as we explained on the first page of this review. As Athlon 64 FX-62 has its memory bus truly running at 800 MHz (400 MHz x 2) and Athlon 64 X2 5000+ has its memory bus running at 742 MHz (371 MHz x 2), we were very curious to compare the memory bandwidth achieved by these two CPUs. On the graph below we also include the maximum theoretical performance for DDR2-800 memories running at single channel (6,400 MB/s) and at dual channel (12,800 MB/s). Since our memories were running at dual channel, in theory we had to achieve something near 12,800 MB/s. Let’s take a look.

Athlon 64 FX-62 achieved a memory transfer rate 10.20% greater than the one achieved by Athlon 64 X2 5000+ and 36.27% greater than the DDR2-800 standard transfer rate. However, it was able to use only 68.13% of the available bandwidth, as we were running it at dual channel mode. It seems that AMD has work to be done on the embedded memory controller from AM2 CPUs.

[nextpage title=”Overclocking”]

One of the reasons on buying an Athlon 64 FX is its clock multiplier feature, which enables you to configure it as if it were a faster CPU. In our case we increased the clock multiplier of our Athlon 64 FX-62 from 14x to 15x, making it run internally at 3 GHz, a 7.14% increase in the CPU internal clock. At this configuration, the performance measured by PCMark05 increased only 3.37%.

We put the CPU clock multiplier configuration back to 14x and tried to increase its 200 MHz base clock. We could only go up to 218 MHz, a 9% increase, the same result we achieved with Athlon 64 FX-60 and a little bit less than what we achieved with Athlon 64 X2 5000+ (211 MHz). To achieve this result we have to increase the CPU voltage to 1.425 V.

We tried to increase the CPU base clock when it was configured at 15x, however we couldn’t go over 204 MHz with stability. This shows us that Athlon 64 FX-62 is working on the edge, with very little room for overclocking it.

[nextpage title=”Conclusions”]

It is true that dual-core Athlon 64 FX-62 is the fastest CPU from AMD, but is its price worth buying it? Its official price for distributors in the USA is quoted at USD 1,031 while Athlon 64 FX-60’s is quoted at USD 827. And, as we saw in our review, the performance difference between the two isn’t big, on the 7~8% range, on average. And Athlon 64 FX-60 is based on socket 939 supporting standard DDR memories.

We were very disappointed with the overclocking capability of this CPU. Even though we could set it up running internally at 3 GHz by increasing its clock multiplier from 14x to 15x – as it has its clock multiplier unlocked – we weren’t able to configure its base clock over 218 MHz, a result worse than Athlon 64 X2 5000+’s. Funny enough that is the exact same value we could achieve with our Athlon 64 FX-60. Being a very expensive CPU targeted to the enthusiast market (i.e., overclockers), we expected more from it.

If you want to have one of the most high-end CPU based on the new AM2 socket, we think Athlon 64 X2 5000+ is a better buy, as it is quoted at USD 696 for distributors in the USA.

But the final choice is yours, of course.