EXCLUSIVE! New EVGA Classified Dual-Proc Board - 270-GT-W555!

[Serra]

why would qpi be different?

4.0G @ 20x200 (7.2GT/s @ 4.8 QPI ratio) + 3.6G @ 18 x 200 (7.2GT/s @ 4.8 QPI Ratio)....?

That is technically possible (I suppose, I could be wrong), but you would have to seriously work to tune it so that those speeds matched exactly, down to the very last Hz (which would be a big problem unless you were willing to accept one of two or three block positions that are likely well under full OC potential). Whether the CPU's expect eachother to operate at the same speed would be a good question as well.

Serra: It probably means you can mix different models/speeds of Xeons and you don't have to run matched sets that server boards of old require(d).

Here's the thing though: no-one has ever let this occur. Not Supermicro, not Tyan, etc. If they don't, I'm not sure it's a good idea to try.


Actually, the more I think about it, the more I think it would necessarily be a bad idea or not work at all. Let's say you have some threaded code that gets dispatched to each CPU that depends on eachother... if one CPU is dramatically faster than the other, it could finish all of its threads before the other finishes the first, and that could destabilize the whole tower. This is a guess of course (I'm not a software engineer), but I would think there must be scenarios where running multiple CPUs at different frequencies would cause an issue.


And I still don't see a firm reply about different processors entirely, though I think that if the CPU speed differences pose a problem it pretty much answers that question. Potentially some wiggle room with an over/underclocked set that operate at the same speeds, but that would be a bit of a silly question. I think it seems pretty apparent they do have to be Xeons too, but I'd like to see confirmation on that as well.