G.Skill PC3200-PC4400 Memory tested on A64 S939

Memory by JNav89GT @ 2004-11-08

The results and benchmarks speak volumes for the potential performance to be had with high HTT and memory speeds on a Socket 939 Athlon64 setup. I have no hesitation in recommending this memory to those users seeking the utmost in performance potential from their high end computing hardware!

  • next

Introduction

Introduction:


It had been over 1yr since Winbond BH-5 memory had ceased production and about 6-9 months since supplies of BH-5 based memory became scarcer and consequently more expensive. While BH-5 was gone (in production) but not forgotten with the computer hardware enthusiast sector, we were left with modules that offered compromises in performance. Higher MHz modules were in production that offered some initial hope. However, quite lax timings of CAS 3 4:4:8(CAS/tRCD/tRPD/tRAS), often meant that performance suffered despite the higher overall speeds. Many predicted the end of low latency memory and resigned themselves to using such higher latency modules to promote overall system overclocking potential. For the purposes of this article, as memory timings are listed, they will correspond to CAS/tRCD/tRPD/tRAS in this order. For explanation of memory timings in greater detail, please refer to this excellent article at CoolTechZone

A few months ago, several major memory manufacturers began to release memory modules based on Samsung's new TCCD IC (integrated circuit). Specifically, revision "F" for the TCCD Samsung chips seem to offer some truly impressive performance in comparison with then current production memory IC's. Initially seen in the form of Corsair's 3200XL, and followed shortly thereafter by OCZ, Kingston, and Mushkin based variants; Samsung TCCD based ram modules offered low latency at PC3200 speeds. Even more impressive was the ability to scale higher MHz, often up to PC4000 specifications with moderate timings of CAS 2.5 3:3:7. While these timings are less ideal than CAS 2 2:2:5, performance was better at 250 MHz CAS 2.5 3:3:7 versus 200 MHz at Cas2 2:2:5 timings. Flexibility in utilizing these modules at either low latency near PC3200 speeds' or moderate timings at near PC4000 speeds, offered the user allot of "tuning" that he/she could perform to optimize their computer systems and prospective overclock.

What became really exciting was the number of reports that began to surface on various overclocking forums. Various individuals were experiencing that newer production modules based on Samsung TCCD, such as OCZ's EL PC3200 Rev 2 modules were often getting 260-270+MHz at Cas 2.5 3:3:7 timings. What was particularly of importance to many, was the relatively low voltage the ram needed to operate at these levels. It appeared that 2.85v was the max voltage these modules preferred. The importance of which was not lost on many overclockers, as at this point voltage modifications to our precious motherboards was unnecessary as most performance boards offer 2.80-2.9v VDIMM selectable in bios. As manufacturers became more familiar with these new modules, improvements were implemented and today we are receiving some of the newest production DIMMs (dual inline memory modules) based on the Samsung TCCD revision F memory modules.


Madshrimps (c)


G.Skill ?

G.Skill is a company that was formed in 1989 by "enthusiasts" and based in Taipei, Taiwan. Their mission statement reads; "We are here to provide superior memory products and satisfactory services in order to keep pace with our customers' growing needs, and help our customers in adding value to their products. And we pledge we will continue to do so and enable both sides to obtain significant competitive advantages in the market segments". The name of G.Skill means God Skill, per Kevin Wu, U.S. Marketing Director. Upon further clarification, the meaning more closely identifies to the term Gosu or God Like Skill, not in the sense of the Christian God or Lord that some may recognize and possibly take offense to without explanation.

I have to admit that I have heard very little of G.Skill as a company, as a U.S. reviewer I feel I may be somewhat geographically isolated to certain products. However, overclocking forums tend to offer information on products worldwide, and I can recall no prior mention of G.Skill products that were discussed at length. I was browsing a forum about 1 month ago when I came upon a very interesting post. A U.S. marketing director for G.Skill named Kevin Wu was posting information on new memory modules they were releasing that offered yet unheard of performance. Modules that were tested and verified to run at DDR 4400 or 275 MHz at timings of CAS 2.5 3:3:7. Amazed with this proclamation I followed the thread and soon appeared screenshots of these modules indeed running at specifications that Kevin had alluded to. As one not easily swayed, and as some might say a skeptic, I began correspondence with Kevin regarding these modules and was eventually offered the opportunity to review them for myself.

I was eager to receive the promised memory, and once they arrived I quickly got to work on discerning how this product functioned in my experiences, compared with manufacturer data. Having no history to fall back on regarding user reports of G.Skill quality and service, this was somewhat of an abyss I was exploring with these memory modules. To qualify that statement! It is somewhat comfortable to assume that Corsair, Kingston, Mushkin, and OCZ memory for example will perform at their rated timings. On the off chance that the prior stated memory fails to live up to the quality that is promised, each manufacturer has shown the ability to service their customer’s needs and quickly provide replacement ram modules. Without such history known to me with G.Skill as a company, the question still remained in my mind, on whether this memory was as high of quality as top tier manufacturers, and whether the memory would live up to such lofty expectations as those claimed by G.Skill themselves.

Thus the challenge had been set, which would be to test this ram in accordance with the specifications of the manufacturer. Whereas tight timings of CAS 2 2:2:5 at PC3200 speeds was relatively easy to predict passing ability, my concern focused on the ability of this ram to run at 275mhz at CAS 2.5 3:3:7 timings. So what did we find out? Did the ram lead us to the Promised Land, or leave us in the desert wanting! Continue reading to find out.
  • next