My "hobby" project at the moment is a quad socket Tyan 4985 I had laying around running 4x 8381 (2.5ghz high efficiency) procs on 32gb RAM. I have two "real" PCIe x16 duoble wide slots available. For those who don't know, each of the x16 slots on this board run through it's own chipset...with one married to CPU0 and the other married to CPU1. They are both however 1.0 specced (iirc).
At some point, SLI becomes more effective due to proc and pipe bottlenecks, and the boards unique efficiency of hardware splitting the CPU load. SLI is tempremental on this board, and I'm willing to experiment. I would like this to be it's "final" nostagia build....simply the best it can do.
Where is the threshold of bottlenecking for modern GPUs...as knowing this, I can populate each "half" with close to maxed cards. Keep in mind, this is different from the ordinary bottlenecking questions, as I can completely max both physical CPUs, and still have another 8 cores (2x physical CPUs) to run on. I also have some proccesor affinity tricks to force games/progs to CPU2&3.
Thanks....lonnnnggg time member here, but haven't posted in prob at least half a decade, lol
At some point, SLI becomes more effective due to proc and pipe bottlenecks, and the boards unique efficiency of hardware splitting the CPU load. SLI is tempremental on this board, and I'm willing to experiment. I would like this to be it's "final" nostagia build....simply the best it can do.
Where is the threshold of bottlenecking for modern GPUs...as knowing this, I can populate each "half" with close to maxed cards. Keep in mind, this is different from the ordinary bottlenecking questions, as I can completely max both physical CPUs, and still have another 8 cores (2x physical CPUs) to run on. I also have some proccesor affinity tricks to force games/progs to CPU2&3.
Thanks....lonnnnggg time member here, but haven't posted in prob at least half a decade, lol