3.
The lack of top performance in games is really quite irrelevant because these processors will run any game far above the 60fps threshold of most monitors: the gaming bottle neck can only really be the graphics card if you have one of these processors.
The point of gaming benchmarks with processors is really academic. Such benchmarks are good test of CPU performance because they emphasize a diverse and complex instructions that simultaneously tax several elements of processor architecture: cache size, cache speed, clock speed. In fact, one might conjecture that what this article has shown is that with complex instruction sets (e.g. video games), the demands on the cache actually slow processing: since each thread effectively requires its separate data in the cache, the demands on the cache are very high when multi-threading is on. Don't know that for sure, it could be something more buggy, but since hyperthreading gives such a radical performance increase in other applications, it suggests that there's nothing wrong with hyperthreading in Nehalem per se, but with hyperthreading in this particular task. I wish I understood this stuff better.
0
0
[
Posted by: philosofool

|
Date: 06/05/09 01:47:12 PM]