Intel Corporation (INTC) Message Board

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Intel Corporation (INTC)

INTC 36.78

+2.7%

Nov 26 3:59 PM EST

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next

To: Tenchusatsu who wrote (128438)	2/27/2001 11:16:35 PM
From: Rob Young	Read Replies (1) of 186894

Tench, "It's sometimes called RUMA (i.e. Reasonably UMA)." Another take: SKUMA Sorta-Kinda UMA "In my opinion, that's true only for up to 16 EV7 processors (in a 4x4 arrangement). Beyond that, you'll have to go totally NUMA. You can't make 256-CPU UMA or RUMA systems, not even with EV7's architecture." In a two-D torus, and 4x4 at most you are 4 hops away, the Router is 13 cycles pin to pin, so that would be 52 cycles at 1000 MHz or 52 ns, setup takes time too so it isn't near that good. Are all things accounted for including arbitration? It appears that way , at least it looks that arbitration tears up 7 cycles, but see for yourself: alphapowered.com Worst case memory access in a 16 processor EV7 appears to be less than 100 ns. Re: 256 CPUs? "Or are you saying it can be done?" Why not? With 256 CPUs what is the worst case hop? When you plot all combinations of hops worst case is a very low percent. Average latency is very good. Nice too is aggregate memory bandwidth, 12.8 GByte/s direct attached, 25.6 GByte/s routed. Heard at IDF McKinley has 3X the memory bandwidth of Itanium. 6.4 GByte/s... shop and compare! Rob

Report TOU Violation

Share This Post

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next