Justanote, Lets try to focus on the topic and not rip on each other.

awestruck,
Imangine you weren't a F!#N@ jerk (edited for content)...

Lets try to focus on the topic and not rip on each other...

Thanks!

Really, can you imagine four processors running together in the future? Never! it's not like there's countless examples already of computers running thousands of Intel CPUs as one machine. Gee you think they might....nah it's just too outlandish...but imagine if they started using the next gen Intel chips in the same way! Or imagine if someone built a small cluster from say 10 dual-CPU server boards and used it as their computer! Imagine if you could get two graphics cards on one motherboard and each of them had two processors!

this is cool. so finally we're going to see working hydra engines? I'm very curious.

I can imagine four Larrabees running together in the future.

Any same manufacturers cards are said to work together aggregated and parallel processing just-like-that? Not so sure that will happen in practice - if they use the same instruction sets it should work.

What about shared memory (eg a graphics process is split between more than one card, so that means that to keep the processing consistent the cards will have to share address space (but they have their own memory per card) or fetch it from another location (ram, the hydra itself?) - or ! duplicate it (hopefully not)....if the hydra will be storing addresses or data for 1+ card, then I'm thinking it will very much be needing a good cooler on it at the least.

It sounds great though - having extra onboard chips to do any processor-type sharing is probably the best way to deal with 'how do we split up processing between multiple *PUs?', for now. Essentially having more chip(set)s or a return to co-processor onboard days.