Timings and nf7

[antipop]

After reading that thread http://www.xtremesystems.org/forums/showthread.php?s=&threadid=14871
I decided to test different value. At first i wanted to run 5 times sandra but after the first results i saw that 3 was enough, i choose the run that was the more representative of the timing.
Here are my result :
4-2-2-2 : 3217/3037
5-2-2-2 : 3217/3040
6-2-2-2 : 3234/3047
7-2-2-2 : 3238/3050
10-2-2-2 : 3240/3055
11-2-2-2 : 3241/3057
12-2-2-2 : 3238/3052
15-2-2-2 : 3228/3036

At my surprise, the best timing was 11-2-2-2. I didn't tried 8 and 9 because i wanted to compare 4 7 and 11.
I also tried 11-1-1-2 but my comp didn't boot, but i wasn't surprised :D

I forgot to take a screenshot, but my score aren't impressive so you'll wait that i mod the nb and do L12 mod before showing my score ;)

[Major]

very interesting, I wonder antipop how 3dmark looks at the various timings. Is it possible for the settings with optimal bandwith in sandra to be less than ideal for total system perfomance ??

[felix88]

nice, i'll have to give this a try. we have very similar systems.

[saaya]

wonder if this works on 8rda+ ...brb gone for benching :D

[antipop]

Originally posted by Major Slaughter
very interesting, I wonder antipop how 3dmark looks at the various timings. Is it possible for the settings with optimal bandwith in sandra to be less than ideal for total system perfomance ?? I don't know if there would be any change in 3dmark, i have to check it maybe tonight :)
However the difference in bandwidth is minimal, so i don't think you'll see any improvement in real life use

[Lunatic]

I tried it aswell and 11 gave the best performance. On a Abit nf7-s v1.2 and Corsair 3500 ddr

[Iolao]

These are my results I tried a while ago:
Barton 2500+ @ 250x10
cas2 X-2-2

X
3 3822/3582
4 3821/3582
5 3824/3585
6 3834/3593
7 3838/3596
8 3842/3598
9 3846/3601
10 3856/3606
11 3859/3609
12 3852/3601
13 3849/3599
14 3843/3592
15 3844/3589


The best is 11 for me too. It's interesting to note that raising from 3 to 11 improves of about 30mb/s in both alu and fpu, while from 11 to 15 the score drops a little bit. It would be very important to make a similar test with 3dmark and pifast to see if there are any differences.

[saaya]

results on my 8rda+

7 3021/2836
8 3023/2838
9 3024/2841
10 3022/2842
11 3021/2838


the exact numbers on all tests ive run, i rebooted and ran the test again:

3023/2839 3022/2836 3018/2834
3024/2840 3022/2838 3023/2838
3024/2841 3027/2842 3022/2842
3023/2840 3021/2843 3022/2842
3021/2838 3022/2840 3020/2838

9 seems to be the best on my 8rda+ with twinmos 3700 and hyperx 3500 in dual channel 200fsb, cl2 2 2 x

[Lunatic]

well highest bandwidht always score highest in 3dmark, atleast for me :)

[Iolao]

Originally posted by Lunatic
well highest bandwidht always score highest in 3dmark, atleast for me :)
not sure

[I(illa Bee]

what L12 mod?

[antipop]

It's sticky on the amd forum
Here's the link http://www.xtremesystems.org/forums/showthread.php?s=&threadid=15300

ps you're lucky i've just done the first page :)