AI
Friday, June 24, 2005 2:39:58 AM
Re: It is not a "design", not any more than San Diego, Venice and Toledo are different designs. It is just an arrangement of components on the die. Rev E was a design, various configuration of components are less than a full fledged "design". Once a new design, say Rev F, acquires 4 ports for cores on SRQ, Quad core design will be just one of the implementations, just like Sempron.
You don't know what you're talking about here. You might argue that San Diego and Venice share a lot of reuse, except for a chopped cache on the 512k part, but Toledo is a different design altogether. Quad cores will require a new design as well, unless AMD does something along the lines of what Intel has done with Presler.
Re: If you argument is correct Itanium is currently losing a bundle, since its volume will be on the same order of magnitude as a potential QC would have, and there is no further reuse of Itanium design.
It's hard to tell if Itanium is making or losing money any more. Intel does get a premium ASP for it, that's for sure, and it looks pretty heavily weighted at the high end skus as well. I'd bet that the designs themselves at least break even, but the infrastructure and software investments continue to borrow from the future.
An AMD quad core, on the other hand, depends on a die size that AMD has not yet ever successfully brought into high volume manufacturing on a process that is either too big/slow for quad cores (90nm) or too new and risky for a large die introduction (65nm). Notice that even Intel builds Itanium on very mature processes.
Re: The reason why AMD is not selling out its production, and people instead buy higher priced, CPUs from Intel, that are clearly inferior in desktop and server segment, that is a subject for another discussion.
In this case, Intel's dual core desktop CPUs are lower priced than AMD's, and by a fairly significant margin, too. Not that I would buy one, but the price/performance is there this time.
Re: I thought that was 32 bit only. Gee, if a real quad Opteron would be lame in 2006 in your opinion, I hesitate to ask what it would make a 32 bit only QC CPU in 2007.
I don't know where you got this particular piece of FUD (let me guess: Dougie?), but it sounds ridiculous to me. Whitefield as a code name as been leaked for more than a year now, and since the beginning it's been planned as a native quad core based Merom design with integrated memory controller and CSI.
http://www.theinquirer.net/?article=23744
http://www.theinquirer.net/?article=23055
And Merom is, of course, an EM64T enabled core.
Re: Impression in eyes of people with a clue or those without a clue? It makes a big difference, because in eyes of those without a clue, Intel is now ahead with their donkey.
People with a clue ought to be able to change their opinion depending on when Intel is truly ahead or behind competitively. I am most certainly not impressed with Smithfield's performance, but I also recognize that Intel did what they had to, and it ended up being the best tradeoff given what they had. AMD's solution is superior right now, but I can already tell that the 'Droids on this forum are dreaming up new kinds of FUD for Intel's next generation. Things like "32-bit Whitefield" for example.
In the end, it's revenues and profits that make the difference, and Intel has been doing pretty well in spite of being competitively disadvantaged. AMD is right to go the dual core route in servers where their parts are clearly superior over single core designs and a mighty premium can be achieved. On the desktop, it only comes across as trying to give the "illusion" of superiority, even though they don't have the volumes to back it up.
You don't know what you're talking about here. You might argue that San Diego and Venice share a lot of reuse, except for a chopped cache on the 512k part, but Toledo is a different design altogether. Quad cores will require a new design as well, unless AMD does something along the lines of what Intel has done with Presler.
Re: If you argument is correct Itanium is currently losing a bundle, since its volume will be on the same order of magnitude as a potential QC would have, and there is no further reuse of Itanium design.
It's hard to tell if Itanium is making or losing money any more. Intel does get a premium ASP for it, that's for sure, and it looks pretty heavily weighted at the high end skus as well. I'd bet that the designs themselves at least break even, but the infrastructure and software investments continue to borrow from the future.
An AMD quad core, on the other hand, depends on a die size that AMD has not yet ever successfully brought into high volume manufacturing on a process that is either too big/slow for quad cores (90nm) or too new and risky for a large die introduction (65nm). Notice that even Intel builds Itanium on very mature processes.
Re: The reason why AMD is not selling out its production, and people instead buy higher priced, CPUs from Intel, that are clearly inferior in desktop and server segment, that is a subject for another discussion.
In this case, Intel's dual core desktop CPUs are lower priced than AMD's, and by a fairly significant margin, too. Not that I would buy one, but the price/performance is there this time.
Re: I thought that was 32 bit only. Gee, if a real quad Opteron would be lame in 2006 in your opinion, I hesitate to ask what it would make a 32 bit only QC CPU in 2007.
I don't know where you got this particular piece of FUD (let me guess: Dougie?), but it sounds ridiculous to me. Whitefield as a code name as been leaked for more than a year now, and since the beginning it's been planned as a native quad core based Merom design with integrated memory controller and CSI.
http://www.theinquirer.net/?article=23744
http://www.theinquirer.net/?article=23055
And Merom is, of course, an EM64T enabled core.
Re: Impression in eyes of people with a clue or those without a clue? It makes a big difference, because in eyes of those without a clue, Intel is now ahead with their donkey.
People with a clue ought to be able to change their opinion depending on when Intel is truly ahead or behind competitively. I am most certainly not impressed with Smithfield's performance, but I also recognize that Intel did what they had to, and it ended up being the best tradeoff given what they had. AMD's solution is superior right now, but I can already tell that the 'Droids on this forum are dreaming up new kinds of FUD for Intel's next generation. Things like "32-bit Whitefield" for example.
In the end, it's revenues and profits that make the difference, and Intel has been doing pretty well in spite of being competitively disadvantaged. AMD is right to go the dual core route in servers where their parts are clearly superior over single core designs and a mighty premium can be achieved. On the desktop, it only comes across as trying to give the "illusion" of superiority, even though they don't have the volumes to back it up.
