Thoughts on performance and overall usage

Where I would like to say I installed this ram, plugged in a few settings in bios, and all was fine and dandy. Life is rarely so simple and agreeable. Like children, our computers can give us immense pleasure and then endless frustration in the blink of an eye.

My initial attempts at using/validating/benchmarking this ram went awry with high HTT/MHz instabilities. This actually would prove beneficial to this reviewer and hopefully many readers as well, which I will elaborate on shortly. As you may remember, I recently reviewed AMD's 0.09u 3200+ Socket 939 Athlon64 CPU. While this CPU was a very fine performing piece, whenever I scaled HTT and consequently memory speeds above 250 MHz I experienced many memory errors that were immediately apparent in Memtest86+. My instant reaction was that of a memory failure at high speeds and possibly that of an overzealous manufacturer in their rating system.

I contacted Kevin Wu, the U.S. Marketing Director regarding my initial problems, at which time he suggested a beta bios known to help increase stable HTT bus speeds on the MSI board I was using, and also to LOWER memory voltage from 2.85 to 2.65v. To be honest, I am a bit embarrassed to state on record that I really hadn’t considered lowering VDIMM as a viable option to increasing stability. While I have in the past experienced diminished returns on my CPU overclock by adding more voltage, I had become accustomed to setting the highest available VDIMM selectable in bios to insure stability in my previous overclocking. Often times I would risk damage to my motherboards, and voiding manufacturer’s warranty by performing a hardware voltage modification on my motherboards to be able to provide even MORE voltage to my ram modules. However, after flashing to the beta bios and lowering VDIMM to a suggested level of 2.65v, I was able to run 265 MHz. Better, but still short of the rated specifications for this memory I was beginning to get frustrated.

In time, as I began to troubleshoot the situation I decided to try my trusty Athlon FX-53 CPU in an effort to duplicate my problems across multiple CPUs. One might ask why the CPU might be an issue! Well as we know, the Athlon64 uses an on die memory controller that has proven to have excellent performance characteristics. Inherent to the design is the possibility that as we overclock the CPU, we are at the same time overclocking the memory controller. When a CPU is overclocked, there comes a point at which some portion of the core is a limiting factor in further clocking and instability intercedes. It is therefore possible that the on die memory controller could become a limiting factor in overclocking, and therefore the decision to try my FX-53 seemed logical.

Further discussions ensued with Kevin regarding my continuing problems and suspicion of the CPU as a possible culprit versus the ram. He stated that he has heard of other people with 3000+ and early 3200+ 0.09u Socket 939 Athlon64 CPU s that had trouble running very high HTT at 1:1 ram speeds at 1T command rate. I did attempt to try 2T command with my 3200+ CPU but performance was abysmal in comparison to 1T and in my opinion not a viable option.

After installing the Athlon FX-53 CPU, I was rewarded with the ability to pass Memtest86+ at 275MHz with timings of CAS 2.5, 3:3:7. 3DMark2001SE looped 12+ hours without issue and all benchmarks ran w/out errors. While this was a welcome verification of the memory’s ability to run at its rated specifications, I was concerned that a more mainstream CPU would be a better choice for this review and therefore ordered a new 3500+ 0.09u Athlon64 CPU from Monarch Computer.

After the 3500+ arrived, I installed the new CPU and the system maintained Memtest86+, 3DMark2001SE, and Benchmark stability as with the FX-53. I was confident at this point that my initial attempts at running these ram modules with my 3200+ were limited by the CPU and not the memory. While this is mildly disappointing that the 3200+ was unable to scale higher than 265 MHz HTT/Memory speeds at 1T command rate, there is no denying the value and performance of that CPU.

In hindsight, I believe it was probably a good thing these problems were identified for the purposes of this review. I feel it is my responsibility as a hardware reviewer to identify potential issues that end users might experience with products I review. Certainly, I was finally able to get the ram running at its rated speeds and timings. I would however extend caution to persons considering ANY high performance computer parts and in this case extreme performance memory, do not assume that just because a memory is rated at PC4400 specifications as these G.Skill modules are, that you will automatically achieve these results UNLESS all other parts in your system are up to running these speeds as well. This might include your motherboard, CPU, power supply, and in some cases active memory cooling (not an issue encountered with G.Skill or any TCCD based memory I have used).