Given that research is invariable limited by resources and time, I will try my best to declare what I feel are weaknesses of this article for you to keep in mind when reading.
We are only testing on one Socket 939 MSI K8N Neo2 Platinum for AMD64 tests and one Asus P4C800E-Deluxe motherboard for Intel tests. While data provided can certainly be used to cross compare memory, I would be cautious in that all platforms and motherboards vary in their compatibility and overclockability with various memory modules. Test SetupWindows XP Professional SP1, DX9C, no OS tweaks
Testedat 200mhz, 217mhz and 265mhz levels to show performance deltas in various tests.
OCZ PC3200 EL Rev 2 Cas2 2:2:5 (2x512mb Dual channel Kit-Retail) Timings:
- Cas 2 2:2:5 @ 200mhz
- Cas 2.5 3:3:8 @ 217mhz
- Cas 2.5 4:4:8 @ 265mhz
OCZ PC3700 Gold Rev3 Cas 2.5 3:3:8 (2x512mb Dual Channel Kit-Retail) Timings:
- Cas 2.5 3:3:8 used in all testing
ScienceMark 2.0 Beta seems to prefer the 3200EL at 240mhz and below. At 265mhz the lead solidly belongs to the 3700Gold.
This was interesting. The 3200EL Rev2 held compiling times steady throughout the mhz spread. Timings being loosened seemed to be offset by the higher HTT speeds. The 3700 Gold Rev3 shows the potential of high HTT though. We see solid gains as mhz scale on ram and timings are held constant. SuperPi seems to really like cpu mhz and at 265mhz we are even a few mhz lower on the cpu vs. 200 and 217mhz.
Sandra buffered scores show us good bandwidth across the board. The 3200EL Rev2 taking the lead at 200mhz due to very tight timings. At 217mhz timings are identical, but the 3200EL still wins this round. 240mhz sees both ram trading blows as timings are the same and scores are within variance. At 265 though the 3700Gold Rev3 enjoys the better timings now and shows the advantage.
Again we see the OCZ PC3200EL Rev2 lead at 200mhz in read tests. The PC3700 really hits its stride past PC3500 speeds and just scales from there. By 265mhz the slack timings needed to run the PC3200EL are really hurting performance. Then again, how many 3200 sticks will run at 265mhz!
I like to use 3DMark as a system performance gauger as this benchmark reflects changes in ram timings and mhz in the results you obtain. The 3200EL Rev2 illustrates the importance of low latency. Tight timings at 200mhz produce scores that are stout enough to hold off ram up till about 240mhz when using slightly slower timings. I would have liked to have run the 3200EL Rev2 at 265HTT with a PC2700 ram divider which should have placed it around 212mhz. I think this ram could have still held tightest timings at that speed and should have by all accounts been higher than the 200mhz run. However, the board would not post with these settings in bios and sadly, I cannot provide these results. That said, the 3700Gold Rev3 begins to pull ahead at 217mhz and above and by 265mhz is beating the 3200EL Rev2 by a solid 800pts.
Doom3 follows the trend set in 3DMark. 3200EL Rev2 leads at 200mhz levels and beats both rams even after they are scaled to 217mhz. At 240mhz they are basically tied, as cpu mhz and timings are standardized. Once either ram is scaled to 265mhz they both beat tighter timings at 200mhz. However, the importance of tight timings is now lost as the 3700Gold Rev3 having lower timings than the 3200EL Rev2 at 265mhz enjoys a very significant lead.
Onto our conclusion ->