some memory benches for your viewing pleasure

[hurtstotalktoyou]

I've been overclocking my Sempron 2800+ from 1.6 GHz to 2.4 GHz. To do this, I set my HTT to 900 MHz (3x), and my memory remains at an even 200 MHz (133 setting). The thing is, though, I have some piss-poor generic memory, a 512 MB PC3200 Xerox module with 3-4-4-8/2T timings.

The module is certified to run at 166 MHz with 2.5-3-3-7/1T timings. Since one of tasks I do most often is video editing, I thought I'd do some testing to figure out which configuration worked best. However, since I was overclocking, instead of dropping the memory clock to 166 MHz, I had to go all the way down to 150 MHz. After this, I went back up to 200 MHz, but kept the lower timings. Here are the results...

test setup:
TMPGEnc Plus 2.58.44.152
Sempron 64 2800+ E6 @ 2.4 GHz
nForce3 250GB northbridge (Epox EP-8KDA3J)
512 MB DDR SDRAM (Xerox 3-4-4-8/2T stock)
80 GB SATA/150 Seagate Barracuda 7200.7
original file: Quicktime video + 44.1 KHz PCM combination
converted file: 1850 kbps MPEG-2 (DVD)
5000 frames (2:46 min:sec)

150 MHz, 2.5-3-3-7/2T/2.65v: 1:41 (50 fps) -6.9%
150 MHz, 2.5-3-3-7/1T/2.75v: 1:36 (52 fps) -2.1%
200 MHz, 3-4-4-8/2T/2.65v: 1:34 (53 fps) stock
200 MHz, 2.5-3-3-7/1T/2.75v: 1:29 (56 fps) +5.6%

As you can see, the slower clock, even undertimed to 1T, couldn't overcome the faster 200MHz at stock timings. Undertiming at 200 MHz led to a nice little boost in performance, more than 5%! I've only begun to run Prime95, though, so we'll have to wait and see if it's stable.

 

[Lithan]

You should note a few things.

First: this is Single channel. It doesn't reflect dual channel (939) scenarios.

Second: you only ran one benchmark. I'd recommend at least three or four different modern games to give a gaming performance estimation.

 

[hurtstotalktoyou]

Oh, I wasn't trying to show all-around performance. I just ran some benchmark tests, and I thought I'd post them publicly. They're probably not of much interest right now, but you never know who might be searching for similar data some time in the future. I don't play games, period--except my fiance plays Sims 1 now and then--but I may run some audio benchmarks tomorrow. I was kind of surprised at the results. I expected a ~2% boost from stock, but I got 5.6%. It goes to show that (1) you can undertime cheap RAM while overclocking your CPU and (2) spending the extra $10 or so on memory with low timings, which you might be able to further undertime, may be worth while. By the way, Prime95 passed for 3 hours at 2.65v, so I'm going to leave it there until I can run a 24-hour test.