Deneidez said:
Applications don't take advantage more than two cores, because there weren't that many cores when they were made/designed. However its not trivial to use more cores than there are different kind of tasks, but its possible. For example my current simulation program takes as many cores as you have. There are no limits.(Ok, theres actually one limit. If theres more cores than work, but that would make it run real time with only few clock cycles.) As for AMD's roadmap... Those processors have way too small cache. Using stuff from ram instead of cache is way too 'slow'. :/ |
Yes, I know applications can be rewritten from the ground up to take advantage of n-core chips. (I'll ignore massively parallel tasks and server apps, because it's trivial to scale them. Deesktop/laptop users don't use massively parallel apps, but that's where the sales volume is.) However, it's all about compatibility. If you're going to require everyone to do a rewrite, you might as well make a clean break and move to a better architecture like IBM POWER. Actually you might as well dump Windows too, since compatibility is its last advantage. Seriously, 90% of the apps people will want to use (especially games) will NOT be rewritten for cores>2 (or even cores>1).
Oh, and the cache situation isn't that bad. Firstly, AMD has exclusive cache compared to Intel's inclusive cache (relating to whether data in L1 is copied to L2 and L3) so their 'effective' cache size is larger. Secondly, their cache latencies are much better than Intel's. No, the problem isn't cache: it's execution. They have a nice socket, bus, cache, chipset, graphics integration, and everything else, but Intel's core is just faster.
