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: semiconeng who wrote (81294)	6/1/2002 8:16:30 PM
From: Elmer	Read Replies (1) \| Respond to of 275872

-- By the time AMD gets their x86-64 act together, intel will be selling 3rd Generation 0.13u Madison Lower Power IA-64, with reported 3MB of Cache Most likely offered by almost every, if not EVERY (If SUN wakes up) Major Server Maker on the planet. The Madison rumors are for 6Meg of ondie cache. BTW, don't forget Pete's the guy who insisted that Intel has a capacity of 200K Wafers per week! Message 17406539 EP

To: semiconeng who wrote (81294)	6/2/2002 10:02:53 AM
From: Charles Gryba	Read Replies (2) \| Respond to of 275872

Semi, I agree with you on the risk starts but I don't care much about the itanic ( no matter how much cache Intel adds to it ). Enterprise purchasers won't go near it with an 8-foot pole unless it can run their legacy software faster and more reliably than their existing hardware. In the enterprise, software drives hardware purchases not the other way around..... C

To: semiconeng who wrote (81294)	6/2/2002 11:56:25 AM
From: pgerassi	Read Replies (3) \| Respond to of 275872

Dear Semi: Even Intel does not go whole hog onto a new revision on all lines all at the same time. Only if it was proven. That should mean running on one line or maybe, two. Only after bin splits are realized and functional real silicon after device testing in a real world environment. Intel has not gone to a new revision on all lines running once in all the time I have been watching. Even slower during a new CPU core. When they could simply add extensions, but not tell anyone that they are active, until ist was proven to work and then they unveil that these CPUs do the extensions. It might appear that they switched to a new revision but, they just told people that it could do it when it could for a while. They do not have that luxury now. Especially with IA-64 being dead in the water. Out how long? And not even to 10K units actually sold for money. Itanium 2 will fail. Its still too slow, too hard to work with and ignored by the market. They have gone up in performance less than their own IA-32 lines over the same time frame and still not as fast as the competition. And for all your wishes, Opteron is nailing the cover on IA-64's coffin either by itself or forcing Intel to use a x86-64 or variant thereof. And the latter two are not even out for third party testing yet. And when third parties test outside of the limits wished by Intel, even that highly optimized P4 software on the flagship NW fails to match the performance of AXP. IA-64 can't even get up to the level of a plain Pentium on normal server random workloads. Look at what Lightwave 7.0b does when used on other than the Intel recommended image set. Even in Photoshop, a very frequent 1 degree rotation done all of the time when an image is scanned (alignment and registration is never perfect when scanning), is much faster on AXP than any P4. And without overall performance of even P4 much less AXP, AMP, Power4, et al, IA-64 will not be sold to this market. Too hard to work with, not predictable in performance or development time, not fast enough even after much time spent optimizing and not being able to upgrade without recompiling (Stuff optimal on Itanium 1 is not optimal for Itanium 2 (If you do not believe, that is what is implied when the compiler optimizes for resources available to a given CPU revision) unlike that for x86, until P4 that is). LCD is what happens for widely available software and that means that core resources must be untouched for years. And we see what that means, poor performance forever. With the CPU optimizing its own execution, the level of abstraction rises. Compilers may still optimize for a given CPU but that gets just a few percent of additional performance intead of 50%, 100% or more. And underlying architecture changes require little if any software change. The RISC backend change now in all modern x86 CPUs could not be done with the compiler doing the optimizations (at least it would be very hard and require extensive redevelopment time). OOE is not allowed by IA-64 by definition! In a environment that continually ups the number of levels of abstraction, Intel is swimming against the current by trying to throw away a number of them. Why? To make the hardware easier to design and make. That is that stated reason for EPIC. Yet, it has not done so. Itanium is harder to make, not easier than either RISC or x86! And that is the supposed easy part! IA-64 is a dead end. Intel just refuses to see that. Evidently, so do you. IBM pioneered it and dropped it. TI did it for embedded apps and it worked ok for that but, they still went away from it. A number of others tried it and walked away. EPIC is far more complicated hardware wise than VLIW, yet for all of that still didn't make the compiler any easier to do. It is no where near ready. And until it is ready, IA-64 is dead. So dream on. You know much about process engineering. You may know a little about how process affects design. You seem to know little about systems analysis of the whole picture from initial design decisions to working systems in the field. And even less about the business of the above. I have a much more varied experience seeing this from many POVs. Asd a software developer of long standing, EPIC has failed, thus far. The trends are not good for that to change. It is another blind alley. I am sorry that Intel went this route and hope they can recover from it. Competition is good from a customer's POV and a developer's POV. From a stockholder's POV, it stinks. I own Intel stock through mutual funds in my 401K and AMD both through funds and directly. Granted, I make more if AMD rises to near industry standard values and Intel falls to same. Even more if Intel and AMD become more equal. And the most, if Intel stops being an ex-monopolist and determines to live and let live by stopping the price war. Then everyone wins. AMD gets more ASP, Intel gets that entrepenurial fire of its youth and the customers get the results from the race. AMD makes money for us, Intel makes money for us and the results allow the customers to increase productivity to the point where even our poor live far better than the richest among us. Pete