Mattb2e

I know everyone on here has heard that running ram in a 1:1 ratio is supposed to be better performance, some others say it isnt, well today I figured that I would do a little testing on my own setup to see the results. I used a benchmark program called Performance Test 6.1 by Passmark. I tested the difference between running my ram at 4-4-4-12 timings at 667mhz vs 5-5-5-15 timings at the stock speed of 800mhz. I like using performance test as it isnt isolated to just one individual part of your system, it does several tests on all parts of your system, except audio. So I ran the test twice for both settings my results showed that I got an overall score for my system of 850.7 with underclocked ram and tight timings and 448.3 for my rams stock timings. When I looked at the individual groups of testing I found that I had gained about 1-2fps in the complex video test, and about a 5 percent gain in overall video performance, also the memory tests came up to be 5.3 percent faster with stock timings and 7 percent faster for disk tests. The only deficits were in the cpu tests and the cd-rom tests both about 4 percent slower than the underclocked timing, overall though having the stock ram timing affected my overall system performance by -.3 percent. Considering that there is an increase of 1-2fps in video testing and there is a 5 percent gain in memory tests I think a .3 percent loss in performance is well worth having a divider in place to run my ram at 800mhz.

I hope that this was coherent enough to understand, if anyone wants to try this out on their own and post their results that would be cool, not just with performance test but maybe with some other benchmarking programs.

-n7-

This isn't news, though it's good you tested for yourself.

As soon as i went to an Intel platform, i started testing RAM ratios.

This is back with a P5B-D + E6400 nearly two years ago.

I quickly discovered 1:1 is NOT the best performance, as on average, all the different test i ran showed 4:5 or higher speeds > lower speeds with tighter timings.

In games, in video encoding, in benchmarks, etc.

Same thing holds true today with all Intel chipsets.

Mattb2e

Yeah I figured that some people knew already, but the fact is that some people dont know, so I figured I would put it to the test to see for myself instead of reading about it and trying to find information. Just posted my results as proof and to see if anyone else had any benchmarks to prove the same and see which ratios are the best.

-n7-

t4ct1c47

I've made some runs on SuperPi and it turned out slightly better to have faster running RAM with higher timeings, than RAM with lower timings and 1:1 ratio. I ran each test 30 times and although there was very little difference between the times, the higher speed proved more beneficial in this benchmark.

Right now I'm still working on the intro of my report. This may seem odd to some people but I need to explain exactly what I'm looking into so that even laymen who don't know what FSB and HTT are will understand it. This also requires that I have credible references which means no wikipedia unfortunately. I'm basically doing what I can, when I'm in the mood for it as I currently have a web technologies assignment from my Uni that's taking priority right now. I intend to run 3DMark06 and 3DMark01SE at least 5 times in each configuration, so that would be pretty much a whole afternoon taken up just running the benchmarks.

I must stress that any tests made by me will only have been run on my own machine, so others with different MCH's (Intel chipsets) may see different results.

Low timings and low speed;


High timings and high speed;

__________________
3DMark Vantage - 18201pts E6600 @ 3.75Ghz
Q6600 @ 3.50Ghz

私は作るアマチュアポルノグラフィー

-n7-

t4ct1c47, Everest is a really good way to slow tiny differences in ratios, speeds, timings, etc.

For more real world comparisons, use games (at lower resolutions; higher will be GPU-bound, not CPU/memory-bound), WinRAR, encoding (though the difference will be slight).

Truth is, buying more expensive RAM is basically always a waste of money in price/performance terms.
A slightly better CPU/GPU or higher overclock will make a far bigger difference.

That doesn't stop me from spending more on RAM though, or stop me from spending hours overclocking/tweaking mine.

kpo6969

Here's my $85 2x2GB kit with modest settings running 4:5



Super PI
http://i239.photobucket.com/albums/f...o6969/PI38.jpg

-n7-

For Everest on my mobo, higher FSB is better than lower, so i'm a 5:6 user right now.

Poor numbers since i have to run loose subtimings to maintain a stable 8 GB.

18 is # 1

Yep, mine runs faster with higher bandwidth too...of course lowering your timings doesn't hurt once you've found your sweet spot.

:

__________________
http://valid.x86-secret.com/show_oc.php?id=410597

Super Pi 1M 11.672s @ 3.9GHz

For Sale: 2x1GB OCZ CL3 6400