I'm kinda 'bummed', I could only get my C2D up to 3GHz last night. It would do 3.4, but not stable. I've got the fsb:ram at 1:1 as someone suggested, but then I read this post on another board:
First assume the ram chip timings (ras, cas, etc.) don't change (much) when you change the ram speed, FSB speed or ratio.
Memory latency is proportional to the FSB/ram ratio (and the ram timings).
Memory bandwidth is proportional to the (FSB) ram speed.
Both of these affect the efficiency and performance of the ram.
On AMDs the latency is the most important factor because of its short pipelines - fast on-demand access to the memory is important. So for AMDs it is best to run the FSB synchronized with the ram - a 1:1 ratio is best.
On Intel P4s the long pipelines masks the latency so bandwidth is more important. So for P4s a higher ram speed improves efficiency and it doesn't matter what the FSB/ram ratio is - adjust the ratio to get the highest ram speed.
C2Ds have a memory controller which optimizes memory access very nicely, the C2D does not have particularly short pipelines and the combination of the optimized memory controller and large cache can easily keep the pipes full, there is no need to run the memory synchronously. C2Ds are usually not bandwidth starved as a result.
So, C2Ds are most efficient when the ram speed is highest - adjust for the highest ram speed at whatever FSB/ram ratio is workable. In this case the latency of the ram is also somewhat important, a lower latency ram is obviously best since at higher clock speeds the (higher) latency affects performance. The limiting factor for FSB speed with a C2D is the Northbridge chip which at high FSB speeds can get very hot and unstable.
For C2Ds this means that a higher (the highest) multiplier will probably give the best performance since that will cause the Northbridge to run at the lowest possible FSB speed.
C2Qs (quads) have been shown to be somewhat bandwidth starved, four cores on one bus causes some memory contention. But, because of the large cache, C2Qs also benefit most from having the highest ram speed possible regardless of the FSB/ram ratio.
So now I'm considering running it asynchronously because it's at 333MHz now and I can run my memory (ddr2 1000) at 500MHz, and from what that post says, that would be a better situation with a C2D.
Thoughts?
Author
