Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
CPU Intel P4 840 D P4 820 D P4 630 P4 640 P4 650 P4 660 P4 670 AMD Athlon64 3500+ 3700+ 3800+ 4000+ X2-3800+ X2-4200+ X2-4400+ X2-4600+ X2-4800+ 1-Way Opteron Opteron 144 Opteron 146 Opteron 148 Opteron 150 Opteron 152 2-Way Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Opteron 248 Opteron 250 Opteron 252 2-Way Dual Core Opteron Opteron 270 Opteron 275 nVidia GF 7800GT GF 6800GT GF 6600GT ATI R X850 XT PE R X850 XT R X800 XT PE R X800 XT R X800 XL Memory Corsair Crucial Kingston Mushkin OCZ What are you shopping for?

Test Setup

Results

On both systems, the modules were capable to run at 2.5:3:3 at 250 MHz stable through all benchmarks, which included 3DMark2001SE, Comanche4, Final Fantasy XI, UnrealTournament and a number of other benchmarks to stress the memory / memory bus. However, above 252 MHz, the Twister would crash, even if the voltage was increased to 2.9 or 3.0V. When we increased the latencies to 3:4:4 to push the memory further, the DIMMs lost stability and developed a tendency for crashing, even at 245 MHz and, once again, the voltage did not seem to make any difference. To be fair, the higher latency settings worked better on the P4P800 but it is really a moot point since there is no reason to run memory at latencies that are higher than what the modules support. The point here is that for every memory component, setup and hold times need to be balanced with the frequency range and only at the sweet spot will the modules run stable AND fast.

Keep in mind that memories are passive devices and as such, the performance, that is, how much bandwidth they provide is limited by the memory controller alone. The only contribution of the memory itself is to either pass or fail, differences in performance between different modules at the same settings are either error / retry events or else glitches in the benchmarks. The only exceptions in that case are either a hardware malfunction as we saw it in e.g. the case of the Kingston Registered DDR400 modules, or else, there can be a wrong SPD setting that. e.g. prohibits bank interleaving on the memory controller level.

What is more interesting is, how do different latency settings at the same speed influence the performance. As always, there is a very general answer, that is, it all depends on the application. For example, running HDTach at different memory latency settings will show zero differences.

In any case, we have a few benchmarks to position the Twister PC4000 between a ultra low latency (2:3:2) memory module (Mushkin Level II) and another PC4000 or higher running at 3:4:4 settings (Mushkin PC4000) for CL:tRCD:tRP. All benchmarks were done on the DFI LAN Party running at 250 MHz x 10 in synchronous memory mode.

   

       

Top row: SiSoft Sandra buffering enabled -- Disabled
Bottom row: Comanche4 -- Final Fantasy XI -- 3DMArk2001SE

SiSoft Sandra shows up to 6% spread between the three contenders, gaming benchmarks up to 3%. This does not appear too much, but first of all, there are other system limitations and second, on average, the difference between the slowest and the fastest is approximately one full CPU speed grade. We have read only recently somewhere else that quality memory does not matter anymore because there is very little performance gain. I would argue the same for any processor or other system component with maybe the exception of a graphics adapter. On the other hand, spending a little bit more on any component will either cover up the lesser grade (of e.g. CPU) or else sum up to getting the top performance. Those who wish to go commodity should be recommended to stick with Dell or Gateway anyway.

Conclusion

It took a long time (over 5 months) to get the TwinMOS samples, when they finally arrived, it was at a very busy time and so they were sitting here for about 2 months until all data were ready and confirmed and we apologize for the delay. The modules will run at the specified latencies but will not work at CL-2 or any other more aggressive latency settings, nor will they run reliably at more relaxed latencies but there may be differences between different mainboards. In the final analysis, the TwiSTER PC4000 meets the manufacturer's specs but does not go beyond those. In other words, there are practically no margins, the modules are specified to run at 250 MHz max and they do no less and no more. Personally, I would rather characterize them as PC3700 but in times where it has become usus to spec modules to what they can potentially run under optimal conditions, we won't hold aggressive marketing against TwinMOS. At the same time, a price point of ~ $78 makes the TwiSTER pair relatively attractive for the moderate overclocker community.

next page:    => more Memory =>