Dual-core CPU Duels
Via Scott M. Fulton, III, through Jeff's pointer to Dueling Multicores- Intel and AMD Fight For the Future (Hardware):
Thanks for bringing attention to the dual-core article on Tom's Hardware Guide. You wanted to know why I didn't take apart Pat Gelsinger's comment about multiplying performance by a factor of 10 over 4 years. The reason is because any attempt to do so on my part might be construed as opinion, and I didn't want the reader's own opinion on this topic (yours) to become obstructed.
You think it's a vague formula? It may very well be because it _is_ vague. A lot has to happen in order to enable a ten-tupling of performance. Both Longhorn and future Linux must evolve substantially; otherwise, application performance could conceivably _decrease_ with the addition of new processing cores beyond the second one. The multithreading scheme of choice must evolve; in the update to the article, Intel finally admits that hyperthreading will evolve, and that's a significant statement on their part. But Gelsinger did manage to set a very high bar without detailing how that bar can be reached, especially with the performance curve not increasing along with Moore's Law. A lot of this absence of detail is self-evident…and you caught right onto it.
Again, my thanks for providing Lockergnome with a link, and I hope you'll stay tuned to Tom's for more news that Lockergnome can share.
Maybe a few weeks before Longhorn gets released, I'll update my primary system with all the latest bells and whistles.









Recent Comments