AMD Athlon64 X2 5000+ / FX62 CPU Review :: Test Setup And Procedures

05-23-2006 · Category: Hardware - Processors

By Doc Overclock

One of the most important aspects to any system that's sitting in the lab is stability. If it's going to be here for the rest of my life, or even for the next few minutes, I have far more important things to do than fight with it on a daily basis. Now, while I'd like to say that today's installations were problem free, that just isn't going to happen, as after all, each platform had its issues. After a week of testing and retesting we need a vacation away from our test bench as every other day either we got new drivers or better memory to test with, and wanting to stay ahead we just kept testing away, all the hours of everyday.

First and foremost, Intel testing went over well, with the exception of the smoldering hot 955 Presler we used in this review. Towards the end of testing, the Intel Reference cooler was tossed, and replaced by a Gigabyte 3D Cooler and then eventually we moved to a Swiftech liquid based system. When I found temperatures to be up to 20 degrees lower while utilizing Gigabyte's solution, a conclusion had to be drawn that either there was a flaw with Intel's design, it's mounting, or in the cooling abilities of the processor itself. Apart from this hotbed of testing, Intel's portion of the testing went off without a hitch.

On AMD's side of the fence however, a pair of problems turned up to trouble me. First off, many attempts to install the N-Force 590 Drivers on ASUS' M2N32-SLI Deluxe met with miserable failure. After quite a few attempts, I did manage to get them installed, and then proceeded on with the testing. It is worth noting that on the Foxconn C51XEM2AA motherboard no problems with driver installs were encountered, so this is something to do with ASUS. The secondary issue is the result of a bug in the CPU's memory controller itself. Essentially, the memory controller is designed to operate at a reference clock of 200 MHz, and divide the frequency by the CPU's multiplier to attain the memory speed.

The issue arises here when DDR2-800 is utilized, as to get the proper memory frequency, the memory's frequency must be divided by ½ the CPU's multiplier. Unfortunately, only full multipliers are allowed, so in the case of chips such as the 5000+, who is clocked at 2.6GHz (13x200), the memory is stepped down. The stepping occurs because the CPU refuses to see .5 multipliers, and as such, rounds up to the nearest number - in this case 7. Thus, when the memory's speed is determined - 2600/6, it does not turn out to be 400. Rather, it comes up to be 371MHz, or DDR2-743, all while windows reports seeing it as DDR2-800. CPUs with even multipliers should remain unaffected by this issue. AMD's response to this issue is that memory latencies are more important than actual memory speeds, and as such, memory with similar timings at both these speeds should perform equally.

During our testing sequence we set the boards up with their best performance timing using only stock cooling and with the CPU set in its default state. All industry performance tests have been run a minimum of three times and then balanced for a more accurate score. We use an open-air test bed due to the fact we test multiple sets of parts in each motherboard we evaluate for compatibility measurements and certification.


Intel Platform Test System


AMD Athlon64 X2 5000+ / FX62 CPU Review

Motherboard:ASUS P5WD2-E i975 chipset
CPU:Intel LGA775 P4 Presler 955
Cooling:Intel Retail/Gigabyte 3D Orb
Memory:Corsair DDR2 PC8500 2GB 2X1GB Sticks
Hard Drives:2X WD 74GB Raptors RAID 0
VGA:XFX 7800GT 256MB PCIe

AMD Platform Test System


AMD Athlon64 X2 5000+ / FX62 CPU Review

Motherboard:ASUS M2N32-SLI Deluxe/ Foxconn C51XEM2AA Nforce 590 chipset
CPU:AMD X2 Athlon64 FX5000+/Athlon64 FX62
Cooling:AVC Z7U7414002
Memory:Corsair DDR2 PC8500 2GB 2X1GB Sticks
Hard Drives:2X WD 74GB Raptors RAID 0
VGA:XFX 7800GT 256MB PCIe

AMD Athlon64 X2 5000+ / FX62 CPU Review
AMD Athlon64 X2 5000+ / FX62 CPU Review


Test Software

  • Sysmark 2004
  • Sisoft Sandra 2005 Memory
  • PCMark05
  • PCMark04 Memory
  • F.E.A.R. 1024x768 max settings timedemo
  • Doom 3 1024x768 HQ 32-bit timedemo1
  • 3DMark06
  • 3DMark05
  • 3DMark03
  • Aquamark 3