duals quads mem speeds

[benniebeeker]

why do quad cores have slower read times and latencies than dual cores...

example... e6600 and q6600 at 400x8 for 3.2 everest results:
................e6600..........q6600
read...........10k...............8k
write...........7k................7k
copy...........6.5k.............6.5k
latency.......49ms............63ms

same chipset... same hardware... only difference is the cpu...

this was brought up in another forum with no definitive answer...

[zanzabar]

it looks like the ram setting arnt identical, and what chipset is that with

[benniebeeker]

it looks like the ram setting arnt identical, and what chipset is that with
you would think the ram settings aren't identical but they are... write and copy always match within a few points... but read and latency are always off... someone suggested on the other forums that a quad has to split the fsb and cache once over and it effects the latency... which makes sense to me but im looking to verify this...

800mhz 1:1 4-4-4-12-t

evga 680i a1

nice 3dmark score by the way... i also have a 3870... not a bad card for the price...

http://forums.extremeoverclocking.com/showthread.php?p=3052893#post3052893

[zanzabar]

sure u have 5 of teh settings but what i ment was dose it change teh sub timings i think that do be definitive u would need to get memset or optimize every setting to make sure that they are identical, for instance there could be a different default performance lvl and that would make a huge difference on latency


also the NV chipsets arnt optimized and dont natively support duel die cores and use emulation to use them so that might be the problem

[benniebeeker]

sure u have 5 of teh settings but what i ment was dose it change teh sub timings i think that do be definitive u would need to get memset or optimize every setting to make sure that they are identical, for instance there could be a different default performance lvl and that would make a huge difference on latency


also the NV chipsets arnt optimized and dont natively support duel die cores and use emulation to use them so that might be the problem

its not the sub timings and im not the only one that notices this on my chipsets.... it happens on other chipsets too... i have a quad in my p35 board (p5k premium) and i get extremely similar results to the quad on a 680i... i asked a few fellow gaming clan member to bench their systems with similar clocks and sure enough... same results... its the same results across operating systems same results across chipsets... it has something to do with the way the cpu handles the memory... im just looking to find out exactly why...

[STEvil]

Four cores fighting for limited bandwidth is going to be slower than only two fighting over the same bandwidth.

[benniebeeker]

thats what i figured...