Quote:
|
Originally Posted by Runnel So, now I tested the memory performance til the end, my friend. |
Indeed, you did, well, I'm impressed
But I think the conclusion is somehow false : you can't say your memory performance with 9.5x multi is 21%/14% better than with multi x11. Why? Just because you've done all the tests at different FSB to keep the CPU at the same speed, but not the memory, which was the thing you tested. CPU FSB = Memory FSB, unless you run in asymetric mode (makes no sense on nFroce2). The right conclusion in my opinion is, that :
stock Epox September bios is simply optimized for lower multipliers (probably 9-10x).
Your results with 11x230 match those of 9.5x220 - you used the same timings, so here is the difference and here you can find the right %%%%%%
While testing the influence of the multiplier on memory performance the FSB shlould remain the same - this makes the theoretical maximal memory bandwidth the same for all possible multis.
As above - this stock bios is optimizied for multi 9.5x (probably the most, but I can't tell without tests). Here, I bet you would be surprised, are different tests with different multis on different modified bioses for ABIT NF7:
D26 Manta Rays XT by TicTac :
http://forum.hslab.pl/viewtopic.php?...=asc&start=405
(3rd post on the page, Mavers, seems to be some problems with his server, plz be patient)
D26 TaiPan (one of them, there've been many
) by Merlin
http://forum.hslab.pl/viewtopic.php?...=asc&start=390
(4th post, Mavers, as above - takes time)
D26 Discovery by Merlin and D26 Black MantaRay by TicTac
http://forum.hslab.pl/viewtopic.php?...=asc&start=375
(Mavers as usual - the first post and then another one down the page)
Interesting, isn't it?
@ Wolf2000me : have a look at those tests above. You're running the Merlin's bios now, so I see it is well optimized for many multis - 9.5x and 11x are almost identical. Good to know - I'm gonna tell Merlin - he should be happy with this fact.
This is what all the modders have been trying to achieve - similar bandwidth for a wide selection of multis. Usually, the bandwidth dropps while lowering multis, but I see that Epox has prepared something special for us
Btw. Am I writting a compendium or something? I'll try do save those tests above to my HDD an upload them on my server - it takes ages untli they've loaded.