2x512 vs. 4x512

When I initally built my system, just went with 1GB of mem (2x512). Recently I had the opportunity to get another 1GB kit (2x512) loaner to do some testing.

Some initial info, this is for an intel platform, and is DDR2 ram. System specs is Rig1 in sig.

For starters the ram is as follows:
KHX5400D2K2/1G - Kingston HyperX Dual Channel Kit (2x512) 4-4-4-10; PC2-5400
OCZ26641024VDC-K - OCZ 1GB Dual Channel Kit (2x512) 5-5-5-10; PC2-5400

The kingston is whats been in my system from day one and has been rock solid at a 3:5 ratio, 4-4-4-10 timings.

These are the numbers I get from a few sources; 3003MB/s (MemTest B/W), 5031 (SiSoft INT ALU), 5031 (SiSoft FloatFPU), 17,044 (3DMark03), 7,555 (3DMark05), 4,108 (3DMark 06), 79,181 (Aquamark 3).

Thats using the ram at the specified settings. When I loosen the timings to match what the OCZ is, these are the numbers i get: 2822MB/s (MemTest B/W), 4985 (SiSoft INT ALU), 4980 (SiSoft FloatFPU), (didn't do 3dmark at this or aquamark). Now, the change in memtest i understand since I'm changing the timings. I did this to even up the testing when I toss the other 2 512's in.

So, final test, I tossed the OCZ sticks in, and got the following results; 2768MB/s (MemTest B/W), 4983 (SiSoft INT ALU), 4987 (SiSoft FloatFPU), 17,056 (3DMark03), 7,539 (3DMark05), 4,111 (3DMark 06), 78,709 (Aquamark 3).

Since I'm mixing different brands of ram for this test (which probably isn't the ideal situation), one could almost think that if these were the same kits i'd have virtually no benchmarking differences?

Different brands aside, the memtest bandwidth lowered by 1.95%, and the remaining numbers also had rather small changes +/-. This is running all 4 sticks at DDR667, w/ timings of 5-5-5-10.

To basically summerize, could it be said, that if I put another identical kit of kingston hyperx in, run at 4-4-4-10 timings, pc2-5400. That I probably won't see any degredation in any benchmarks, and just an increase in overall performance? Or would I still be better off going with a 2x1GB kit instead of 2, 2x512mb kits to achieve 2GB system ram.

Any thoughts or feedback would be greatly appreciated.

Related issue

check this post out (link below), i had the same question about ram. Im not very good at intel chipsets and ddr2 ram but i suspect that your mobo can only run 4 sticks at lower frequency, compared to 2 sticks (that is basically what all Nvidia chipsets do). I recomend u check your mobo manuals ram pages for more info. This is what my manual say about using 2 ram slots / 4 ram slots

DIMM1 | DIMM2 | DIMM3 | DIMM4 | DEFAULT SETTING

Single___Single___Single___Single
_Rank___Rank____Rank_____Rank____DDR333

Single___Single____NA______NA
_Rank___Rank_____-________-______DDR400

I also had a significant loss in memory bandwidth using 4x512 compared to 2x512 ram. (from ~5300 using 2x512 to ~3800 using 4x512)

check out this thread:
http://www.ocforums.com/showthread.php?p=4271754#post4271754

I returned my 4x512 and now im buying 2x1024 sticks instead.
Hope it helps

The thing is, I had no loss in bandwidth, other then < 2% unbuffered in memtest, every windows test was the same, or +/- 0.5%, all game bench's the same (again +/- 0.5% if that). And the ram, both kits were DDR2-667, and I was able to run at that speed, had to force it tho; and there were no problems using that speed.

According to the manual, DDR2-667 with CL=4 will be downgraded to run at DDR2-533 by default, altho, this is not the case, my system is rock stable at forcing DDR2-667 and its even autodetecting ram speeds fine. (on 2 sticks, dunno about 4 as the other ram i tested with was slower timings)