Constantine, Re: "you realize I was just kidding around, right?"
It's ok. I've seen the results of Hyperthreading: Round 1, and I'm not all that impressed, either. It seems to have a definite advantage in UP systems running scalable server software, but all this negative scaling on consumer apps is disappointing. The way I see it, Intel wouldn't continue to "hype" Hyperthreading if they didn't see improvements in the future. What you say may contain a grain of truth, too. Some applications (and benchmarks, of course <g>) may need to be rewritten to take better advantage of the hardware. I'd expect that this will be one part of it, and hardware enhancements to lessen the overhead will be the other part. Intel needs to get it to the point where unoptimized applications have a zero net gain/loss, while optimized applications (or ones that are already multithreaded) get a large net benefit.
Take a look at Aces Hardware. The new Northwood chips seem to have received some micro-architectural enhancements. Since Intel went through the trouble, I think there is a possibility of Hyperthreading being changed as well. Otherwise, I don't see Intel as having any more time between now and the end of the year when they introduce a 3.06GHz Pentium 4 with Hyperthreading.
wbmw |
