AMD Next Generation Processor Technology Slides

[Ockie]

And so are Power Supply makers and your local power company

I have a PC&P 1KW, I think I'm set As for the power supply company... I have yet to figure out how to circumvent that HAHA

 

[Ockie]

Would it be possible for AMD to sell a three core chip? I was just thinking what they would do with the four core chips that were found to have one bad core, but the rest were fine. Maybe these would be their mid level chips ($300 to $600 range) with the quad core chips for the ultra high end and duo core chips on the low end.

I think the CPU and GPU and the same chip designs will have to wait for 45nm. I will be interesting to see how they cram all that onto one chip.

It will also be interesting when "Recommened Number of Cores" starts showing up on system requirement labels for games.

They would laser the 3rd core and sell it as a dual core chip. 3 Cores is possible but never feasable.

 

[mcravenufo]

When I first saw the thread title I thought, "oh no. AMD is 'sliding' downhill in technology and quality"

Anyway, looks awesome.

 

[enumae]

They would laser the 3rd core and sell it as a dual core chip. 3 Cores is possible but never feasable.

How do you laser cut 2 cores, that were designed to share L3 cache amongst all 4 core?

 

[nilepez]

Althought it seems like computers went really far in the last 10-20 years.. everything has remained the same but just got a little faster. (in some people's eyes, a lot faster --in reality, not really)

Even if we go back just under 5 years ago with the release of the Northwood core, a 2.0GHz Chip vs an FX 57 (released roughly 3.5 years later), the latter can encode an MP3 using lame in less than 1/2 the time.

A less than state of the art FX55 can encode a 17minute (roughly) audio file using lame in a bit under 1.5 minutes. A Pentium 100 would take just over 77 minutes. 56x faster.

I realize there are other differences between a P100 machine and an FX57 or 55, but placed behind a black box with now knowledge of what is doing the work, I find it hard to believe that most wouldn't have been impressed if told that 13 or 14 years from now your computer will run 50 times faster than your new P100.

Whether we stuck with that architecture for too long is way beyond me level of knowledge, but I don't get your comment that processor speed is just a little faster.

 

[w1retap]

Thats why I said in some people's eyes it may seem way faster. That is how you view it. If you look at the task at hand, its a VERY simple task. We've only sped up the process somewhat. You are only comparing it to today's standards with today's architecture. Come back and look at the same task in another 10-20 years. You will laugh at what I mean. It used to take hours and days to program a simple program to output text with a punchcard. Now it is done instantaneously in milliseconds. Thats how many million times faster? Get the idea?

To get ideas of things that will be done in the future with computers and the upcoming architecture, visit the computer engineering or nano-technology area of your college campus and sit down at talk with a computer engineering professor. You'll get a better idea.

 

[drizzt81]

Would it be possible for AMD to sell a three core chip?

From a theoretical standpoint yes, however I do not know whether there are architectural limitations that would require chips to have an even number of cores.

 

[Jakalwarrior]

Its seems everyone has missed the BIGGEST question.

So if it has a built on PLL and controls its own clocks, does that mean no overclocking?

 

[w1retap]

With a new wave of architecture, there will be a new way of overclocking. =) The processor manufacturer really doesn't take overclocking from an end user standpoint that serious I don't believe. The motherboard hardware and software end of tweaks will probably take care of it I'm guessing.

 

[(cf)Eclipse]

From a theoretical standpoint yes, however I do not know whether there are architectural limitations that would require chips to have an even number of cores.

why would there be? we've been fine with an odd number of cores until very recently

 

[DejaWiz]

Thanks for the great info, Kyle. This is certainly something I'll be keeping an eye on.

why would there be? we've been fine with an odd number of cores until very recently

Now that's GOT to be pure marketing strategy. Why add only 1 more core to your dual-core architecture when you know the competition will add two to theirs? LOL

 

[AdamNesvick]

Kinda funny if ya think about it, Intel has to go to 45nm before they can make quads, AMD's doing at 65.

I like and respect AMD for fucking that 800lb intel gorilla in the ass on a few occasions with performance, The A64's and X2's are truly awesome chips in that it made intel change their WHOLE CHIP ARCHITECTURE to beat them.

 

[Asian Dub Foundation]

drool worthy

 

[enumae]

Kinda funny if ya think about it, Intel has to go to 45nm before they can make quads, AMD's doing at 65.

I like and respect AMD for fucking that 800lb intel gorilla in the ass on a few occasions with performance.

Not trying to bash AMD, but what performance, they don't even have engineering samples out and you have already given them your money...lol

What kind of yields do you think AMD is going to get with a chip that complicated and on 65nm?

If one core is faulty, let alone two, what are they going to do as part of the chips performance is based of it having a shared L3 cache?

When creating a chip that complicated, you do it at a lower cost per die (45nm vs 65nm) and you can deal with the yields not being exceptional?

It would seem very smart of Intel to do quad core on 45nm, they are ahead of AMD when it comes to manufacturing, and if Intel were worried about AMD, why release Kentsfield and Cloverton early, why not wait and see what AMD is offering.

Oh yeah, its about 8 to 10 months away before we even see the server based K8L, let alone a desktop variant, who knows, maybe there will be a native quadcore variant from Intel then.

I will say that AMD making X2 on 90nm is very impressive though.

 

[AdamNesvick]

Not trying to bash AMD, but what performance, they don't even have engineering samples out and you have already given them your money...lol

PAST tense pudwacker, read my post again, I added to it.

Intel didn't kill netburst own their own, AMD made em do it.

 

[enumae]

PAST tense pudwacker, read my post again, I added to it.

Intel didn't kill netburst own their own, AMD made em do it.

Well go ahead and answer my questions then.

 

[HeXeD]

Multicore processors will be utilized by the new Crytek engine for the game Crysis.. if anyone hasn't seen the demos I would DEFINITELY check it out. even though this is only one game, it is still a good looking game

 

[msde]

Doesn't the math seem funny on those slides with the "power states"? I thought 100% + (33%*3) would equate to an overall 50% power state, not 60%. Likewise, with 100% + 50%...that's only 25%+12.5% for a total of 37.5%, not the 45% they listed.

It's unclear from the diagram to me that a processor running at 33% speed consumes 33% of its usual power, or 50% consuming 50%. You can't just undervolt something 50% (1/4 power) and expect it to still run at 25% as fast.

What really excited me about that diagram: It implies that you can mix and match processors.

 

[msde]

Intel didn't kill netburst own their own, AMD made em do it.

Hah no.

Intel placed a bet on netburst allowing them to scale up megahertz. What was it, 5 GHz by 2002?

When they failed to scale MHz, they were going to move away from netburst anyway regardless of AMD. The entire purpose of netburst was high clock speed. It's a horrible architecture if other architectures can match it in clock speed, and only limped along as long as it did by bloating up the cache.

What does netburst have to do with the new yummy AMD quad core architecture anyway?

 

[Stoly]

what about "per core" performance?

will it be able to close the gap with core2?

 

[qdemn7]

So just what are the advantages, if any, of a "native" Quad-Core like AMD is proposing vs a "non-native" Quad-Core like the upcoming Kentsfield is going to be? Are we talking noticible real-world benefits, or it will be marketing hype / within the limts of error benchmarking type of stuff?

 

[Weescotty]

Why all the excitement of all this power when most programs out there cannot take full advantage of multi core? ALso i remember the day when Windows was (is) licensed per processor......does a core count as a processor?

And yes to all of you windows haters out there i know linux and unix can do what ever with them for what ever money....my point is the other 99% of us out there running windows boxes may see a licensing issue

Depends -

Win XP Home - Single CPU any number of cores.
Win XP Pro - Two CPU's any number of cores.

Now it gets tricky -

Win 2k Pro - 2 CPU's or 2 cores max.
Win 2k Server - 4 CPU's or 4 cores max.

This is because Win 2k can't tell the difference between a physical CPU and a core, so each core counts as a CPU (socket), no fix is planned from Microsoft.

 

[s10010001]

its all comming to fast...

 

[DeFex]

I love hearing about new cores and new tech coming out! But games, for example, are only just starting to take advantage of 2 cores, let alone quad and more cores. For me (I understand most people use their PC's for many, many other things), gaming is my thing on PC's and since today we are only starting to now see multithreaded games for dual core, isn't having quad core chips just overkill and more of a "B/c We Can" kind of thing for the chip manufacturers?

it means the government or RIAA can check what files you have on your PC without causing you any lag in your game

 

[TheArchitect]

what about "per core" performance?

will it be able to close the gap with core2?

the main boost core 2 had over core or K8 for that matter was the dual 128bit SSE data path. K8 uses the same as the K7 for FPU its 80bit wide. To do a 128 SSE anythign would take at least 2 cycles. K8L now has the dual 128bit SSE data path among other additions. The difference will now be marginal. Though my money is on AMD overtaking intel clock for clock (core to core) once these come out.

ps: Looks like the memory will no longer be tired to the core clock (finally). Hopefully this means that the memory bandwidth will no longer be cut in half with CnQ.

 

[AdamNesvick]

Well go ahead and answer my questions then.

I don't want to get in a debate about the technology, we all know intel is ahead of the curve here. But I mean, they got owned for how long by AMD when they invested so heavily in netburst, I took them a long time to answer back to AMD

 

[KenAF]

Why are people so excited about this chip right now?

It just taped out, which means it's anywhere from 10 months to 16 months away. Intel's modern record from tape out to availability is something like 10 months, and I don't think AMD has ever released a modern processor within one-year of tape out. Realistically, we're looking at a release some time between August and November 2007.

If we are talking about this chip, why aren't we talking about Ridgefield?

Intel is releasing the quad-core Kentsfield @ 2.66GHz on or about October 30 for $999, according to the latest Intel roadmaps. But the quad-core Athlon64 will only be competing with that chip for a few months. For the remainder of its life, it will be competing with an enhanced version of the Core2 architecture, and then the Core3.

Intel has detailed several new processors coming with the transition to 45nm in late 2007, just a few months after the quad-core Athlon64 ships. The new Intel architecture, otherwise known as Penryn, is an enhanced version of Core2 with a number of architectural improvements and 6Mb of shared L2 cache. Two new quad-core processors will be based on the Penryn architecture -- 1) Wolfdale, a low-end desktop replacement for Kentsfield featuring 3Mb L2 cache per die; and 2) Ridgefield, a high-performance desktop replacement for Kentsfield featuring 6Mb of shared L2 cache per die, for 12Mb L2 total cache. These are slated for release late 4Q 2007 or early 1Q 2008. According to Intel's latest roadmap, these processors will be replaced in 2H 2008 with Yorkfield, which will offer 8 cores per die ("native design") with 12Mb shared L2 cache.

Hence, the "K8L"--- expected to debut about 12 months from now as a high-priced Athlon64 FX--- may compete with Kentsfield for as little as three months. For the remainder of its life, the K8L will be competing against a new Intel quad-core chip with architectural enhancements and 50% more cache -- Ridgefield. And less than a year after its release, the K8L will face Intel's single-die, 8-core Core3 architecture with 12Mb shared L2 cache.

 

[brachy33]

it means the government or RIAA can check what files you have on your PC without causing you any lag in your game

Drat!! Guess I gotta get rid of all that "free" software...

 

[jackofalltrades]

I believe I see another overpriced underused processor no apps to task it after all, if you encode every movie on the planet and play doom while watching a movie and snipping an auction on ebay maybe, I would need it otherwise I believe I will wait awhile on the bloated prices to drop!

 

[duby229]

Why are people so excited about this chip right now?

It just taped out, which means it's anywhere from 10 months to 16 months away. Intel's modern record from tape out to availability is something like 10 months, and I don't think AMD has ever released a modern processor within one-year of tape out. Realistically, we're looking at a release some time between August and November 2007.

If we are talking about this chip, why aren't we talking about Ridgefield?

Intel is releasing the quad-core Kentsfield @ 2.66GHz on or about October 30 for $999, according to the latest Intel roadmaps. But the quad-core Athlon64 will only be competing with that chip for a few months. For the remainder of its life, it will be competing with an enhanced version of the Core2 architecture, and then the Core3.

Intel has detailed several new processors coming with the transition to 45nm in late 2007, just a few months after the quad-core Athlon64 ships. The new Intel architecture, otherwise known as Penryn, is an enhanced version of Core2 with a number of architectural improvements and 6Mb of shared L2 cache. Two new quad-core processors will be based on the Penryn architecture -- 1) Wolfdale, a low-end desktop replacement for Kentsfield featuring 3Mb L2 cache per die; and 2) Ridgefield, a high-performance desktop replacement for Kentsfield featuring 6Mb of shared L2 cache per die, for 12Mb L2 total cache. These are slated for release late 4Q 2007 or early 1Q 2008. According to Intel's latest roadmap, these processors will be replaced in 2H 2008 with Yorkfield, which will offer 8 cores per die ("native design") with 12Mb shared L2 cache.

Hence, the "K8L"--- expected to debut about 12 months from now as a high-priced Athlon64 FX--- may compete with Kentsfield for as little as three months. For the remainder of its life, the K8L will be competing against a new Intel quad-core chip with architectural enhancements and 50% more cache -- Ridgefield. And less than a year after its release, the K8L will face Intel's single-die, 8-core Core3 architecture with 12Mb shared L2 cache.

I just say one word, that will blow your whole argument to crap...

APM.....

 

[eblislyge]

Why are people so excited about this chip right now?

It just taped out, which means it's anywhere from 10 months to 16 months away. Intel's modern record from tape out to availability is something like 10 months, and I don't think AMD has ever released a modern processor within one-year of tape out. Realistically, we're looking at a release some time between August and November 2007.

If we are talking about this chip, why aren't we talking about Ridgefield?

Intel is releasing the quad-core Kentsfield @ 2.66GHz on or about October 30 for $999, according to the latest Intel roadmaps. But the quad-core Athlon64 will only be competing with that chip for a few months. For the remainder of its life, it will be competing with an enhanced version of the Core2 architecture, and then the Core3.

Intel has detailed several new processors coming with the transition to 45nm in late 2007, just a few months after the quad-core Athlon64 ships. The new Intel architecture, otherwise known as Penryn, is an enhanced version of Core2 with a number of architectural improvements and 6Mb of shared L2 cache. Two new quad-core processors will be based on the Penryn architecture -- 1) Wolfdale, a low-end desktop replacement for Kentsfield featuring 3Mb L2 cache per die; and 2) Ridgefield, a high-performance desktop replacement for Kentsfield featuring 6Mb of shared L2 cache per die, for 12Mb L2 total cache. These are slated for release late 4Q 2007 or early 1Q 2008. According to Intel's latest roadmap, these processors will be replaced in 2H 2008 with Yorkfield, which will offer 8 cores per die ("native design") with 12Mb shared L2 cache.

Hence, the "K8L"--- expected to debut about 12 months from now as a high-priced Athlon64 FX--- may compete with Kentsfield for as little as three months. For the remainder of its life, the K8L will be competing against a new Intel quad-core chip with architectural enhancements and 50% more cache -- Ridgefield. And less than a year after its release, the K8L will face Intel's single-die, 8-core Core3 architecture with 12Mb shared L2 cache.

A previous poster summed it up nicely. APM

Amd's manufacturing capabilities are second to none, Intel may be able to produce alot of chips, but look at all the fabs they use to do it. Now step back and look at how much silicon AMD can pump out with just a couple fabs.

Furthermore Amd has extremely high yields now, With Apm they can controle processor assembly down to a single transitor. They got all the pieces in place so reguardless of how there quad cores perform there won't be a shortage.

 

[brentpresley]

A previous poster summed it up nicely. APM

Amd's manufacturing capabilities are second to none, Intel may be able to produce alot of chips, but look at all the fabs they use to do it. Now step back and look at how much silicon AMD can pump out with just a couple fabs.

Furthermore Amd has extremely high yields now, With Apm they can controle processor assembly down to a single transitor. They got all the pieces in place so reguardless of how there quad cores perform there won't be a shortage.

I've gottta call BS on this one. If AMD's process technology was THAT advanced then:

1) They wouldn't be almost a full YEAR behind Intel in the 65nm transition (Yonah has been SHIPPING since the BEGINNING of January, with production run-up 2 months before that, and we aren't going to see 65nm X2 from AMD until December - according to AMD's OWN roadmap).

2) You and I have NO CLUE what AMD's yields are unless you work for AMD. RARELY does a company disclose yields, ESPECIALLY if they aren't stellar and well above the competition. We have had no such announcement from AMD.

3) IF THEIR yields and production were that high, the would NOT have outsourced to Chartered for ~20% of their currently CPU production. You don't pay someone else to make your chips when you have the in-house capacity to do it yourself. That's just bad business. (And FYI, there is a REALLY OLD x86 licensing agreement with Intel that prevents AMD from outsourcing ANY MORE than 20% of their total production to a 3rd party, so they are already at the limit of what they can ask Chartered to produce)

And if you want to compare fabs to fabs, you should really do your homework and check your numbers. Intel has 29 fabs or lines (fabs within a fab essentially), but only a few of them are current in technology and used to produce CPUs. The rest are used for legacy products.

Here is the breakdown:
http://www.intel.com/pressroom/kits/manufacturing/manufacturing_at_a_glance.pdf

2006 Intel 65nm CPU production (all on 300mm wafers):
Fab 12, Fab 24, Fab 11-X, and Fab D1D-

Fab 12 = 65nm 300mm
Fab 22 = 0.13 200mm
Fab 32 = 45nm
Fab D2 =65nm (Flash, cellular, handheld)
Fab 23 = 0.13 (communications)
Fab 10/14 - 0.13 (Flash)
Fab 24 = 65nm
Fab 24-2 = 65nm
Fab 8 = 0.35 (logic)
Fab 18 = 90nm (flash)
Fab 28 = 45nm
Fab 17 = 0.13
Fab 11 = 0.13 (flash
Fab 11x = 0.13 (flash)
Fab D1C = 90nm (logic)
Fab 20 = 0.13 (logic)
Fab D1D = 45nm/65nm
Fab CV1 = flash
Fab Cv2 = flash
Fab PD1 = flash
Fab Cd1 = logic
Fab Cd6 = logic
Fab PG6 = comminications
Fab PG7 = logic and communications
Fab KM1 = logic
Fab KM2 = logic
Fab KM5 = communications
Fab CR1 = logic
Fab CR3 = logic

90nm = 2 fab facilities
65nm = 4 fab facilities
45nm = 3 fab facilities currently in ramping

Intel still produces ~80% of the world's x86 microprocessors, and does so using 4-5 fabs. That would put it right in the same ballpark as AMD in terms of CPUs per fab w/ 1 (soon to be 2) fab that produces ~20%.

 

[moon02]

Why are people so excited about this chip right now?

It just taped out, which means it's anywhere from 10 months to 16 months away. Intel's modern record from tape out to availability is something like 10 months, and I don't think AMD has ever released a modern processor within one-year of tape out. Realistically, we're looking at a release some time between August and November 2007.

If we are talking about this chip, why aren't we talking about Ridgefield?

Intel is releasing the quad-core Kentsfield @ 2.66GHz on or about October 30 for $999, according to the latest Intel roadmaps. But the quad-core Athlon64 will only be competing with that chip for a few months. For the remainder of its life, it will be competing with an enhanced version of the Core2 architecture, and then the Core3.

Intel has detailed several new processors coming with the transition to 45nm in late 2007, just a few months after the quad-core Athlon64 ships. The new Intel architecture, otherwise known as Penryn, is an enhanced version of Core2 with a number of architectural improvements and 6Mb of shared L2 cache. Two new quad-core processors will be based on the Penryn architecture -- 1) Wolfdale, a low-end desktop replacement for Kentsfield featuring 3Mb L2 cache per die; and 2) Ridgefield, a high-performance desktop replacement for Kentsfield featuring 6Mb of shared L2 cache per die, for 12Mb L2 total cache. These are slated for release late 4Q 2007 or early 1Q 2008. According to Intel's latest roadmap, these processors will be replaced in 2H 2008 with Yorkfield, which will offer 8 cores per die ("native design") with 12Mb shared L2 cache.

Hence, the "K8L"--- expected to debut about 12 months from now as a high-priced Athlon64 FX--- may compete with Kentsfield for as little as three months. For the remainder of its life, the K8L will be competing against a new Intel quad-core chip with architectural enhancements and 50% more cache -- Ridgefield. And less than a year after its release, the K8L will face Intel's single-die, 8-core Core3 architecture with 12Mb shared L2 cache.

Oh I dont know maybe because this is the amd forum and this is a topic for people interested in amd's upcoming chip? It could be possible that people are not solely interested in intel products. just maybe we dont need people like you thread crapping and deviating our threads to intel vs amd thread when all we want to talk about is a upcoming amd product.

 

[SilverBack]

Win XP Pro - Two CPU's any number of cores.

I thought XP Pro would just see two cores as well as 2 cpu's?

Also will Vista have quad support out of the box?

 

[DejaWiz]

I've gottta call BS on this one. If AMD's process technology was THAT advanced then:

1) They wouldn't be almost a full YEAR behind Intel in the 65nm transition (Yonah has been SHIPPING since the BEGINNING of January, with production run-up 2 months before that, and we aren't going to see 65nm X2 from AMD until December - according to AMD's OWN roadmap).

2) You and I have NO CLUE what AMD's yields are unless you work for AMD. RARELY does a company disclose yields, ESPECIALLY if they aren't stellar and well above the competition. We have had no such announcement from AMD.

3) IF THEIR yields and production were that high, the would NOT have outsourced to Chartered for ~20% of their currently CPU production. You don't pay someone else to make your chips when you have the in-house capacity to do it yourself. That's just bad business. (And FYI, there is a REALLY OLD x86 licensing agreement with Intel that prevents AMD from outsourcing ANY MORE than 20% of their total production to a 3rd party, so they are already at the limit of what they can ask Chartered to produce)

1) They aren't a year behind anyone. They are on track for their architecture transition to 65nm. AMD didn't ever have the need to transition to 65nm within the last year. Power consumption and the resulting heat output of the current AMD cpus proves this to us.

2) I agree. But I don't think AMD is in a position to put on golashes unless they are ready to blow the dam.

3) Outsourcing does not equal a bad strategy. For AMD, it has granted them an alliance and partnership which has taken their manufacturing processes to new heights. Besides, why spend billions in the short term to revamp current fab plants when you can contract the work to someone else ready for it for much cheaper and then revamp over the long term? Going from 90nm to 65nm to 45nm in a short period of time could mean that haste very well might cause waste, right?

my .02

 

[KenAF]

Also will Vista have quad support out of the box?

Yes. Moreover, Vista makes significantly better use of multi-core than XP. You can expect many new "for Vista" software releases to better leverage multi-core as well.

Today, very few applications take advantage of more than two cores. However, the situation will be very different one year from now. One of the positive things about Intel releasing the first quad-core cpu on October 30 is that it gives developers a processor with which to test and improve their multicore optimizations. By the time the K8L is released in the second-half of next year, there should be plenty of applications and games that make full use of quad core.

 

[enumae]

1) They aren't a year behind anyone. They are on track for their architecture transition to 65nm. AMD didn't ever have the need to transition to 65nm within the last year. Power consumption and the resulting heat output of the current AMD cpus proves this to us.

2) I agree. But I don't think AMD is in a position to put on golashes unless they are ready to blow the dam.

3) Outsourcing does not equal a bad strategy. For AMD, it has granted them an alliance and partnership which has taken their manufacturing processes to new heights. Besides, why spend billions in the short term to revamp current fab plants when you can contract the work to someone else ready for it for much cheaper and then revamp over the long term? Going from 90nm to 65nm to 45nm in a short period of time could mean that haste very well might cause waste, right?

my .02

1. Power consumption is not everythng, I think his point is about profits, if they can make more chips with less wafers, thats more profits.

3. Because your paying someone else to make your product...If you are the size of AMD then yes it would be wasteful to go from 90 - 65 - 45nm, but in a few years when they are creating chipsets, and graphics it will benefit them very much. Each transition allows other departments the ability to move down as well.

Also by that time they will have more fabs, and the ability to transition without losing production.

Hence Intel manufactoring wise is a about a year to a year and a half ahead of AMD.

 

[Mohonri]

There's another good reason not to lead the charge on shrinking feature size: risk. Unlike Intel, AMD is not in a position to invest billions of dollars in retooling, then going through all the pain and expense of figuring out how to make the new process work. They let Intel do it all first, then hop on it once the process is better-understood.

The company I work for (nothing to do with computers) does the same thing. It's called "being an aggressive follower".

 

[DejaWiz]

1. Power consumption is not everythng, I think his point is about profits, if they can make more chips with less wafers, thats more profits.

2. Because your paying someone else to make your product...If you are the size of AMD then yes it would be wasteful to go from 90 - 65 - 45nm, but in a few years when they are creating chipsets, and graphics it will benefit them very much. Each transition allows other departments the ability to move down as well.

Also by that time they will have more fabs, and the ability to transition without losing production.

Hence Intel manufactoring wise is a about a year to a year and a half ahead of AMD.

1) True. AMD is once again gaining ground in the market, especially since Dell recently stated that they are going to start using AMD cpus as well as the recent significant price drops. They may not have had to worry about yields too much over the past decade, but they will have to focus on this more and more as demand for their processors increases.

2) I'm sure they aquired all of ATI's manufacturing plants/partners with the aquisition. And whether or not it will benefit them is subjective. If it remains cheaper to contract the work out than spending to revamp current fab plants, then it won't be of much benefit.

In my opinion it's really not fair to directly compare AMD to intel. They both do business differently. AMD proved it would no longer be a wannabe clone when they introduced the later variants of their Socket 7 processors as well as their Athlon line.

I still think it's silly for anyone to think AMD is somehow behind intel manufacturing-wise. intel certainly isn't a model citizen in the manufacturing world. Meaning, they don't have the "be all end all" perfect manufacturing processes in place. If they did, then every cpu company, including VIA and Transmeta, would be following suit. And that, especially in AMD's case, is not happening.

 

[kemist1117]

The two biggest differences between AMD and Intel are money and capacity. Intel has money to throw at aggressive process ramps, and also has the other benefit of greater fab capacity to spare, while AMD does not. This does not mean they are an inferior company technology wise and if you read any type of article on APM its considered state of the art stuff. The main reason they are lagging is they dont have the huge amount of resources needed to compete that aggressively.