What exactly makes Intel's production facilities that much better than IBM and AMD?
The question is for 2006, 2007 and 2008, not necessarily today.
Moore's Law essentially states that the number of transistors in semiconductor parts, microprocessors, doubles every 2 years. This is done by developing production facilities capable of producing smaller transistors every 2 years.
This statement needs a 10000+ page book to explain because it involves virtually everything in the universe! I exaggerate, but it involves a lot of economics and physics.
The reason Intel's production facilities will be better than IBM/AMD production facilities is because most everyone forgets to mention that the development costs for production facilities doubles every generation. This essentially means that a manufacturer has to sell twice as many processors as the previous fab did to recover the costs of development.
In broad economics terms, only the largest companies will be able to afford to develop next gen facilities and therefore produce next gen processors. Not only that, the number of companies able to do this likely decreases by half for every generation because most companies don't sell twice as many processors. What happens is consolidation.
That's why AMD had to hitch onto IBM's semiconductor R&D wagon for 130 and 90 nm. They will do the same for 65 nm. That's why Moto/Freescale had to codevelop 90/65 nm with Philips semiconductor. They couldn't afford it themselves. Freescale is almost a full fab generation behind Intel right, that's 2 years. Just think about what that means for Apple to use Freescale's processors.
Intel owns 75% of the personal computing market and therefore has the most money, the most fabs and the largest potential market to sell into for each next gen fab. That's why they got to 130 nm first, why they got to 90 nm first, why they will get to 65 nm and why they will get to 45 nm.
Their fabs where the reason why they were able to survive the stink bomb that is Prescott. I wouldn't sweat the fact that AMD's K8 architecture is better than Prescott (and that Apple will be using Intel processors) right now. Intel will have a next gen x86 microarchitecture in late 2006. Intel is moving to 65 nm in about 8 months. Intel dual-core is cheaper than AMD dual-core. Things will go against AMD again. They will always have to fight because their fabs are 6+ months behind.
Those 6 month delays are more and more important too. Processor price cycles have decreased to almost 6 months in the last couple of years. An $800 processor will go down to $200 in a mere 6 months. Intel being ahead in fab means they can recoup the cost of R&D while AMD will always have price pressure to keep the cost of their new next gen processors in order to compete with Intel's processors which are already 6 months old. Ie, 65 nm chip competing against a 90 nm chip. That's bad news for AMD.
"The whole idea is to begin to process 65-nm in the middle of 2005 and bring it into production in 2006. We're certainly on schedule," Mr. Sonderman added.
AMD doesn't have any roadmaps for 2006. Are we to assume that they are going to start making umbrellas instead of microprocessors? Or should we just do the logical thing and say, "Oh, I guess they don't release the same types of roadmaps that Intel does."?
Quote:
Yonah info is out..Merom, Conroe and Woodcrest info will be forthcoming. Intel has a well defined strategy here whilst AMD fans revel in todays architecture.
You can't run operating systems on roadmaps and well-defined strategies. Not even /. nerds can get Linux to run on roadmaps and well-defined strategies.
One thing to notice... Yonah has no 64-bit capability. And it doesn't have HyperThreading. And the second core might not even be activated in battery power, only when plugged into an outlet.
Quote:
BTW I harbor now ill will against AMD. My next homebuilt PC will use an AMD. I like the bang for buck and I want to try something different. I just realize that Intel was the ideal solution for Apple to align with
Why align with any one maker, though? It's nothing but marketing.
You don't have any ill will against AMD, sure, but you definitely want to apologize for Apple's marketing decision and try to make it seem like it is more substantive than it is. It's not. Apple sticking solely with Intel is a pure marketing move.
Apple's must be the only consumer base in the industry that actually enjoys watching their idol company choose money-making over performance and value.
Quote:
The pentium m is more than just a cool running processor. It's pipeline length is shorter thant a P4 giving a higher IPC. Just like a G5 or AMD processor.
AMD is ahead, but AMD is behind. War is peace.
THT:
Good post.
Intel has the process, AMD has the design.
Intel is showing a little improvement in design and promising big things. We'll see.
AMD claims to be able to pump out 65nm at the beginning of 2006. We'll see.
I don't see what the disadvantage is for AMD & IBM to team up on process, I can't see it as anything but beneficial for both companies to pool resources instead of just re-inventing the wheel in their own separate little caves. You could have fighting between the two companies, sure, but with a company as huge as Intel you could have infighting just as easily.
Intel has a 32nm roadmap. . . 2009. I read it in EETimes yesterday.
Holy smoke Splinemodel. You couple this with Intel transitioning to 450mm wafers and you have the potential for low cost but highly dense processors. I guess by then 8 cores will be trivial on a die. Looking at where virtualisation is going in Servers I could see a 42U rack containing 100s of Servers.
Intel and the PCI SIG group just discussed extensions to PCI Express that will allow concurrent use of PCI-Express by virtualized servers and the ability to run PCI-Express over copper
Ahem back on topic with AMD. I'm really looking forward to seeing what K9 offers. I don't think AMD will be far behind but Intel will hit them where it hurts and that's money. I don't doubt that AMD will be at 65nm, they will, I just don't think their microprocessor design is as efficient as Intels at 65nm.
I guess we'll see soon enough even before the first Macintel ships
AMD claims to be able to pump out 65nm at the beginning of 2006. We'll see.
Your URL stated "bring it into production in 2006," not the beginning of 2006. This is a 12 month window for their "on schedule" statement. Qualification on memory ICs is occuring 2H 05. I think AMD will be doing really well if they can deliver a 65 nm uprocessor in July 2006.
Quote:
I don't see what the disadvantage is for AMD & IBM to team up on process, I can't see it as anything but beneficial for both companies to pool resources instead of just re-inventing the wheel in their own separate little caves. You could have fighting between the two companies, sure, but with a company as huge as Intel you could have infighting just as easily.
With the same resources, one company will typically get something done faster than two companies in a partnership.
However, that could be fine as you say. However, I'm negative on IBM's, and therefore AMD's, process technology. IBM was thinking they could deliver a 39 Watt, 2 GHz 970fx in Q1 2004. It's now been more than a year, and the closest they have appears to be 45 Watts. No low-k process in sight either.
In the meanwhile, it looks like Intel has put the engineering solutions and process tweaks in for solving the technical problems at 90 nm with a 2.13 GHz P-M burning 27 Watts. In fact, it even looks like IBM's fancy smancy SOI 90 nm fab produces higher Watts/transistor than Intels.
If I was making a choice going into the future, it seems very obvious who to partner with. If IBM/AMD is already behind, they will stay behind because the game really is about money. More so than ever.
In the meanwhile, it looks like Intel has put the engineering solutions and process tweaks in for solving the technical problems at 90 nm with a 2.13 GHz P-M burning 27 Watts
Pentium-M is the product of the Israeli design teams -- hence the 'biblical' sounding code names.
About 5 years ago Transmeta was the new Si-valley darling. It claimed that it was going to take the low power ( i.e. mobile and blade ) business away Intel. They even convinced Linus Torvalds to come work with them. They proceeded to burn through a few hundred million bucks and they're now down to their last few million. Soon they will be gone.
But, for a while Transmeta was getting good press and they were getting designed into some products. That gave the kids in Israel a big shot in the arm. Their designs for more power efficient x86 chips won out over the other design teams. That lead to several generations of designs. The current 90nm design is generation 3 from these guys.
The new 90 nm Pentium-M designs have just incremental improvements over the last two generations. That's not a slam on the current generation. That's just saying that all of the generations have excellent power profiles.
Not a chance. There's no money in this deal for Intel, its a PR and Branding play. Apple will be locked into a five year exclusive deal similar to the Dell deal.
That Apple has a deal like this, and that Intel can push for a deal like this and get it, I wouldn't doubt.
But I'd also think, so long as Apple was in strong enough a negotiating position while making this deal with Intel, that Apple would have insisted that Intel meet some specific targets by specific dates, otherwise Apple can activate an escape clause to buy from AMD if AMD has something better to offer.
Apple has been burned too many times not to have at least pushed for a way out if Intel fails to perform.
that wont happen, intel has never underperformed. AMD is not yet to 3GHz. Intel is at 4. The future is bright for Intel because of their money. It takes money to make money. It takes money to make kickass processors.
Comments
Originally posted by hmurchison
:P while I enjoy your vain attempts at humor the question still remains. Where is the comperable low power 65nm roadmap from AMD?
Intel has a 32nm roadmap. . . 2009. I read it in EETimes yesterday.
Originally posted by Ben Huebscher
What exactly makes Intel's production facilities that much better than IBM and AMD?
The question is for 2006, 2007 and 2008, not necessarily today.
Moore's Law essentially states that the number of transistors in semiconductor parts, microprocessors, doubles every 2 years. This is done by developing production facilities capable of producing smaller transistors every 2 years.
This statement needs a 10000+ page book to explain because it involves virtually everything in the universe! I exaggerate, but it involves a lot of economics and physics.
The reason Intel's production facilities will be better than IBM/AMD production facilities is because most everyone forgets to mention that the development costs for production facilities doubles every generation. This essentially means that a manufacturer has to sell twice as many processors as the previous fab did to recover the costs of development.
In broad economics terms, only the largest companies will be able to afford to develop next gen facilities and therefore produce next gen processors. Not only that, the number of companies able to do this likely decreases by half for every generation because most companies don't sell twice as many processors. What happens is consolidation.
That's why AMD had to hitch onto IBM's semiconductor R&D wagon for 130 and 90 nm. They will do the same for 65 nm. That's why Moto/Freescale had to codevelop 90/65 nm with Philips semiconductor. They couldn't afford it themselves. Freescale is almost a full fab generation behind Intel right, that's 2 years. Just think about what that means for Apple to use Freescale's processors.
Intel owns 75% of the personal computing market and therefore has the most money, the most fabs and the largest potential market to sell into for each next gen fab. That's why they got to 130 nm first, why they got to 90 nm first, why they will get to 65 nm and why they will get to 45 nm.
Their fabs where the reason why they were able to survive the stink bomb that is Prescott. I wouldn't sweat the fact that AMD's K8 architecture is better than Prescott (and that Apple will be using Intel processors) right now. Intel will have a next gen x86 microarchitecture in late 2006. Intel is moving to 65 nm in about 8 months. Intel dual-core is cheaper than AMD dual-core. Things will go against AMD again. They will always have to fight because their fabs are 6+ months behind.
Those 6 month delays are more and more important too. Processor price cycles have decreased to almost 6 months in the last couple of years. An $800 processor will go down to $200 in a mere 6 months. Intel being ahead in fab means they can recoup the cost of R&D while AMD will always have price pressure to keep the cost of their new next gen processors in order to compete with Intel's processors which are already 6 months old. Ie, 65 nm chip competing against a 90 nm chip. That's bad news for AMD.
But neither am I against well-engineered cooling solutions that allow higher-power components. Like the Shuttle XPC.
hmurch:
Where is the comperable low power 65nm roadmap from AMD?
Quoted for the second time:
"The whole idea is to begin to process 65-nm in the middle of 2005 and bring it into production in 2006. We're certainly on schedule," Mr. Sonderman added.
AMD doesn't have any roadmaps for 2006. Are we to assume that they are going to start making umbrellas instead of microprocessors? Or should we just do the logical thing and say, "Oh, I guess they don't release the same types of roadmaps that Intel does."?
Yonah info is out..Merom, Conroe and Woodcrest info will be forthcoming. Intel has a well defined strategy here whilst AMD fans revel in todays architecture.
You can't run operating systems on roadmaps and well-defined strategies. Not even /. nerds can get Linux to run on roadmaps and well-defined strategies.
One thing to notice... Yonah has no 64-bit capability. And it doesn't have HyperThreading. And the second core might not even be activated in battery power, only when plugged into an outlet.
BTW I harbor now ill will against AMD. My next homebuilt PC will use an AMD. I like the bang for buck and I want to try something different. I just realize that Intel was the ideal solution for Apple to align with
Why align with any one maker, though? It's nothing but marketing.
You don't have any ill will against AMD, sure, but you definitely want to apologize for Apple's marketing decision and try to make it seem like it is more substantive than it is. It's not. Apple sticking solely with Intel is a pure marketing move.
Apple's must be the only consumer base in the industry that actually enjoys watching their idol company choose money-making over performance and value.
The pentium m is more than just a cool running processor. It's pipeline length is shorter thant a P4 giving a higher IPC. Just like a G5 or AMD processor.
AMD is ahead, but AMD is behind. War is peace.
THT:
Good post.
Intel has the process, AMD has the design.
Intel is showing a little improvement in design and promising big things. We'll see.
AMD claims to be able to pump out 65nm at the beginning of 2006. We'll see.
I don't see what the disadvantage is for AMD & IBM to team up on process, I can't see it as anything but beneficial for both companies to pool resources instead of just re-inventing the wheel in their own separate little caves. You could have fighting between the two companies, sure, but with a company as huge as Intel you could have infighting just as easily.
Originally posted by Splinemodel
Intel has a 32nm roadmap. . . 2009. I read it in EETimes yesterday.
Holy smoke Splinemodel. You couple this with Intel transitioning to 450mm wafers and you have the potential for low cost but highly dense processors. I guess by then 8 cores will be trivial on a die. Looking at where virtualisation is going in Servers I could see a 42U rack containing 100s of Servers.
Intel and the PCI SIG group just discussed extensions to PCI Express that will allow concurrent use of PCI-Express by virtualized servers and the ability to run PCI-Express over copper
Ahem back on topic with AMD. I'm really looking forward to seeing what K9 offers. I don't think AMD will be far behind but Intel will hit them where it hurts and that's money. I don't doubt that AMD will be at 65nm, they will, I just don't think their microprocessor design is as efficient as Intels at 65nm.
I guess we'll see soon enough even before the first Macintel ships
Originally posted by groverat
AMD claims to be able to pump out 65nm at the beginning of 2006. We'll see.
Your URL stated "bring it into production in 2006," not the beginning of 2006. This is a 12 month window for their "on schedule" statement. Qualification on memory ICs is occuring 2H 05. I think AMD will be doing really well if they can deliver a 65 nm uprocessor in July 2006.
I don't see what the disadvantage is for AMD & IBM to team up on process, I can't see it as anything but beneficial for both companies to pool resources instead of just re-inventing the wheel in their own separate little caves. You could have fighting between the two companies, sure, but with a company as huge as Intel you could have infighting just as easily.
With the same resources, one company will typically get something done faster than two companies in a partnership.
However, that could be fine as you say. However, I'm negative on IBM's, and therefore AMD's, process technology. IBM was thinking they could deliver a 39 Watt, 2 GHz 970fx in Q1 2004. It's now been more than a year, and the closest they have appears to be 45 Watts. No low-k process in sight either.
In the meanwhile, it looks like Intel has put the engineering solutions and process tweaks in for solving the technical problems at 90 nm with a 2.13 GHz P-M burning 27 Watts. In fact, it even looks like IBM's fancy smancy SOI 90 nm fab produces higher Watts/transistor than Intels.
If I was making a choice going into the future, it seems very obvious who to partner with. If IBM/AMD is already behind, they will stay behind because the game really is about money. More so than ever.
In the meanwhile, it looks like Intel has put the engineering solutions and process tweaks in for solving the technical problems at 90 nm with a 2.13 GHz P-M burning 27 Watts
Pentium-M is the product of the Israeli design teams -- hence the 'biblical' sounding code names.
About 5 years ago Transmeta was the new Si-valley darling. It claimed that it was going to take the low power ( i.e. mobile and blade ) business away Intel. They even convinced Linus Torvalds to come work with them. They proceeded to burn through a few hundred million bucks and they're now down to their last few million. Soon they will be gone.
But, for a while Transmeta was getting good press and they were getting designed into some products. That gave the kids in Israel a big shot in the arm. Their designs for more power efficient x86 chips won out over the other design teams. That lead to several generations of designs. The current 90nm design is generation 3 from these guys.
The new 90 nm Pentium-M designs have just incremental improvements over the last two generations. That's not a slam on the current generation. That's just saying that all of the generations have excellent power profiles.
Originally posted by vinney57
Not a chance. There's no money in this deal for Intel, its a PR and Branding play. Apple will be locked into a five year exclusive deal similar to the Dell deal.
That Apple has a deal like this, and that Intel can push for a deal like this and get it, I wouldn't doubt.
But I'd also think, so long as Apple was in strong enough a negotiating position while making this deal with Intel, that Apple would have insisted that Intel meet some specific targets by specific dates, otherwise Apple can activate an escape clause to buy from AMD if AMD has something better to offer.
Apple has been burned too many times not to have at least pushed for a way out if Intel fails to perform.