Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: dumbmoney who wrote (81081)	5/30/2002 1:35:04 PM
From: wanna_bmw	Read Replies (4) \| Respond to of 275872

dumbmoney, Re: "you sure do view IA-64 with rose colored glasses." Give me a break. Just about everyone else on this thread thinks that it's an absolute certainty that Hammer will rule the server market, and that Itanium will be thrown out with yesterday's trash. Even a mention that this might not be true is met with harsh criticism - that's how convinced people are! Are you 100% certain, too? If so, who do you think is the one with rose colored glasses? Re: "Don't you think it's a little strange that a brand new, cost-no-object design using a shiny new (and supposedly much superior) architecture can't outperform a low-cost commodity x86 CPU, loaded down with 23 years of architectural crud?" First, x86 may have 23 years of architectural crud, but they also have 23 years of compiler optimizations around that architecture. As far as what I think, Itanium performance has nowhere to go but up, as compilers and applications get fine tuned around the new architecture. Second, integer applications have very limited parallelism, so it's very difficult for any parallel architecture, whether it be superscalar or explicitly parallel, to gain massive amounts of improvement. Having said that, I think it's more than a feat that McKinley was able to get a >2x performance improvement over the previous generation Itanium. Like I said above, I think a lot of improvement will come from figuring out how to extract further parallelism at the compiler level. Adding architectural improvements like Hyperthreading, larger caches, higher frequencies via extended pipeline, etc, will all help, and Intel has several new cores already on the roadmap. Look at floating point performance, and you can see how well Itanium's parallel computing comes in handy. The fastest commodity CPU on the planet, the 2.53GHz Pentium 4, manages 861 on SPECfp, but Itanium 2 with its explicitly parallel engine gets almost 57% better performance at only 2/5 the clock frequency! And this is without the benefit of the better manufacturing process on the Pentium 4. Itanium 2 is still at .18u. But early next year, there will be yet another core. I won't say "wait for Madison", because McKinley is already here, and it performs great. But when the time comes, you can compare a .13u Itanium 2 against the fastest .13u CPU - whether it be Pentium 4 or Hammer, and I bet you'll still see a floating point score that embarrasses the commodity processors. The ideal would be if Intel could find a way to extract enough parallelism from integer code to make it as high as its floating point performance. They may or may not be able to make that miracle happen, but integer performance is certainly no slouch, and I believe that the bandwidth of the Itanium platform (6.4GB/s) will allow it to top the charts in transaction processing and other enterprise applications. Later improvements in the compiler may make things even better, but I'll take my "rose colored glasses" off and admit that it may take years, or it may not happen at all. All I can say is that performance is already very promising for the architecture, it will definitely get better with new cores on the horizon. IA-64 hasn't even started, and it's already competitive with any other CPU out there. Its demise is certainly not definite, as many people here like to think. The industry may yet take a serious interest in it, and it still has support from a couple of the largest server OEMs out there, so it's certain that efforts will continue on Itanium, at least in the near future. What happens after that is anyone's guess, and I'm sure your speculation is no worse than mine, but no less rosier, either. wbmw