AMD's Response to 'delays'

[CTho9305]

Originally posted by: Nemesis 1
Well its nice that you countacted Bob. For your information . Being a few years older than Bob. I am not surprized that you would put the word Chip with micro processors. ut CHIP in real world is slang for Chip off the Block. Meaning father like Son. CHIP.

I don't believe Bod doesn't remember Chip. I never said I was Chip. Bob was a little young for my group. 3 years younger than I . When your Kids 3 years is a Big differance.

Could you fill in the blank in these sentences?
"I know Bob because we were/are ______." (e.g. "students at the same school", "relatives", "coworkers", etc.)?
"I want Intel to have 60% margins because ______." (e.g. "I work at Intel", "I own a lot of their stock", "I have cash burning a hole in my pocket and want to give it to Intel when I buy my next CPU", any of the previous with "I" replaced by "a friend or family member", "IBM worked with the nazis", etc.)

Why I Hate AMD . Hate is a strong word. I just don't believe for 1 second that AMD has ever stood on its own legs. IBM has always been their forcing Intels hand. Intel made a deal with IBM not the Industry. AMD should have been allowed to make Cpu's for IBM because of the agreement. X86 was developed By intel not for the IBM pc's but for the PC industry as a whole. AMD took it upon themselves to reverse engineer intels 386 because AMD couldn't do it on its own .
Intel screwed up by not wording the contract with IBM more simply. Intel did leave the door open for AMD to Steal it tech and they did .

IBM required any CPU supplier have a second source. It wasn't a mistake that AMD could legally produce clones - it was a strict requirement from IBM. Without AMD as a second supplier, IBM would have chosen a different CPU instead of Intel's 8088 chip. Intel knew what they were getting into. As far as I know, IBM didn't restrict Intel from selling to other companies, so Intel was free to sell their chips to "the PC industry as a whole".

Everyone says Intel is monoply . What is AMD . They rely on IBM for research .

AMD has <50% of the market. It'd be hard for AMD to be a monopoly. Intel, on the other hand, has a lot of the market, and there are too many allegations of foul play on their part to not conclude that they've abused their market power.

Why is it bad if IBM and AMD share R&D for manufacturing technology? It means each side spends less, and can charge consumers less as a result. Isn't that a good thing? Anyone who wants to pay a lot for something is stupid.

I here people talking about AMD and High K / metal gates @ 45. If that happens great for them right.

Who cares? It's like SOI. The only people who should care are the people at the company that makes the product. If someone produced a fast chip for a reasonable price on 180nm technology with aluminum interconnect, I'd buy it over a high-K/metal gate, copper & low-k interconnect 32nm chip that was slower/more expensive/worse. It's just marketing bullet points... Intel is definitely milking the 45nm stuff to the limit. When I'm buying a product, I don't care why it's better - just that it is better.

There's a funny video on youtube where someone is in a retail store asking people what they think about the Hafnium technology and nobody knows or cares what it is (I think the video is by AMD...but seriously, ask ordinary people on the street how many nm their CPUs are).

Why would MS. Do this . EPIC MS is scared to death of a OS for desktop using high level language.

WTF are you talking about? Everybody uses high-level languages on every OS (outside of toasters, light switches, and ). Last I checked, you can run Windows 2003 Server on IA64 (EPIC) machines. Microsoft chose a non-x86 architecture for the Xbox360. As I understand it, managed code is generally more portable than normal C/C++. If Microsoft is scare of a non-x86 architecture, they aren't showing it.

I say this that AMD64 / IBM and MS . have already effectively brought innovation to a crawl. Witness K8.

AMD 64 isn't goog for progress it chokes it.

AMD64 made x86 much less sucky, while retaining compatibility with the many billions of dollars of existing software. If you happen to be lucky enough to not depend on legacy apps, good for you. In the real world, companies live and die by programs written 10 years ago by an intern whose source code they no longer have.

I wanted to add this bit. How it relates I will leave tuglo you.

RISC vs. CISC is over. I'm not going to debate it with someone who won't come out and say what he's thinking but instead likes to drop little tidbits and expect us to figure out what they mean.

 

[Fox5]

Want proof Intel will raise prices?
Check out Intel's pricing guidelines for next quarter. Everything that matches with AMD is dirt cheap, after that prices skyrocket.

Also, Nemesis, something like the Power6 can hardly be considered RISC at this point. Considering RISC was about high clock speeds and low complexity (ie, shifting the burden to the programmer), whereas the Power6 has some of the highest clockspeeds around in addition to being one of the most complex chips the planet's ever seen. It may be a better ISA than x86, but it's not RISC.
In addition, the instruction set of a cpu has greatly declined in importance. As long as two architectures are geared toward the same paradigm (such as Power and x86), the burden of the architecture has been so marginalized at this point so that it doesn't matter significantly.
Oh, and reverse engineering is legal and IBM was SMART to mandate a secondary supplier of x86 processors. It kept Intel from charging through the roof, made up for shortages, and even more recently probably guaranteed IBM some good bargains for its x86 servers.

 

[Nemesis 1]

Power 6 isn't RISC . OK If you say so .


IF ya can find it guys go here.



The Rise Of Functional Programming: F#/Scala/Haskell and the failing of Lisp January 13th, 2008 Over at Lambda The Ultimate, the best academic programming blog on earth, there is a large debate going on regarding what the future of languages will be for 2008.

 

[Accord99]

Originally posted by: Vee
Please keep your wits together. Your answer is mindboggling, and so are the many roads drowning your extremely unlikely little future scenario. I don't even know where to start.
There won't be any saturated market. Why would Intel invest huge sums constantly building new manufacturing facilities to keep the world market "saturated"?

The developed world is a saturated market for computers. Just about everybody who needs a computer, has a computer (or two, or three). The only incentive to buy new CPUs is if they offer improved performance. The emerging market would be demanding cheap CPUs.

Intel's purpose is to make money. Not to supply the world population with cheap computing power.

Which is why Intel has always had products covering the entire range of the market, with regular product advancements, because that is how you make the most money.

It would definitely be in Intel's short term interest to have a much leaner manufacturing structure that offers better returns. And in the long term their revenues will rather rise than drop until the day they are encountering competition.

But it would hurt their long-term revenues.

Another thing you will see evolving is a dramatic difference in price, capability and performance between highend and lowend platforms. Today, competition from AMD has pushed up the lowend to almost close the gap to highend. This doesn't just benefit Joe&Jane.

The power of the lowend platforms comes from the natural performance increases of x86 vs the slower performance demand increases of software.

It also exposes the failure of Itanium and makes it hard to justify transistor counts, price, power consumption.

Of course, the Itanium is growing rapidly in the big iron market. At least it's making money, unlike AMD.

There wouldn't have been any cheap, powerful X86 solutions without AMD. With AMD gone Intel can start the progress to extract premium charges for hardware that meets the needs of industry & business. My guess is that Intel's primary means to accomplish this stratification is the capabilities they will offer in the various platforms. Meaning, any need above the typical business/consumer laptop - pay up! Pay up big.

And? Once a business has paid up once, without a meaningful performance difference, they're not going to pay up again.

Also ask yourself this question, Why does intel exist? To make money. They have no reason once amd is gone to keep prices low. As long as there is a need for processors they can charge what they want(certain limitations apply like if they were to raise the price of their cpus above what their customers can afford or if the government regulates them somehow).

And if replacing existing CPUs with new products that aren't very much faster, and are very expensive, you're going to see customers refuse to buy.

Again your answer is mindboggling. Again I don't know where to start...
They will buy them because they need to. And you will too. Trust me, the world is not about to experience any lack of demand for processors.

But they don't need to for the vast majority of applications. They just do, because a) it's cost effective because of increased performance and other capabilities b) it's cheap or c) a combination of both.

 

[JimiP]

Originally posted by: Accord99

Originally posted by: zach0624
I think jimip is pretty spot on here. Also what reason does intel have to improve tech if there is no competition?

The answer is to keep selling CPUs in a saturated market.

Also ask yourself this question, Why does intel exist? To make money. They have no reason once amd is gone to keep prices low. As long as there is a need for processors they can charge what they want(certain limitations apply like if they were to raise the price of their cpus above what their customers can afford or if the government regulates them somehow).

But if CPUs are expensive and offer little improvement, then people won't buy them.

I don't feel that to be true. What would we do without CPU's? No matter how costly they get, there will always be a market for them.

 

[Accord99]

Originally posted by: JimiP
I don't feel that to be true. What would we do without CPU's? No matter how costly they get, there will always be a market for them.

Yes, there will always be some who need the absolute cutting edge; but it's not enough to generate $40 billion in revenues and $9 billion in profits every year for Intel if it massively increases prices and greatly reduces product advancements. Intel has grown to where it is today because it sells to the entire market and has enough product advancement to get people and business to upgrade regularly. People used to spending $500 on a whole system are not going to spend $1000, especially if the new system is not much better.

 

[Extelleron]

Originally posted by: Accord99

Originally posted by: JimiP
I don't feel that to be true. What would we do without CPU's? No matter how costly they get, there will always be a market for them.

Yes, there will always be some who need the absolute cutting edge; but it's not enough to generate $40 billion in revenues and $9 billion in profits every year for Intel if it massively increases prices and greatly reduces product advancements. Intel has grown to where it is today because it sells to the entire market and has enough product advancement to get people and business to upgrade regularly. People used to spending $500 on a whole system are not going to spend $1000, especially if the new system is not much better.

I don't think Intel would raise prices, I just think that their pace of advancement would decrease significantly. Instead of releasing Nehalem, a big ~270mm^2 chip, why not continue selling small, profitable Penryns and just offer minor speed bumps as the proces improves? Why not keep the same architecture for 4 years instead of 2 and shrink Penryn to 32nm, then introduce Nehalem when it is cheaper to do so?

That's the kind of thing I see happening. Intel will not raise prices, but they will stop releasing new products as frequently and certainly reduce R&D expenditures.

 

[Vee]

Originally posted by: Accord99

Originally posted by: Vee
Please keep your wits together...?

The developed world is a saturated market for computers...
...

cut for brevity.

Intel has not always had products covering the entire market. And one of my points was that you're going to see large differences (virtually nonexistant today) and very major price markups (also pretty trivial today) as you move up the range. That is how they will make the most money.

Also I don't think the CPU market is anywhere near saturated. In a decade or so you will find yourself being the owner of a rather large number of Intel 'platforms'. In various things and gadgets. Regardless if AMD survives or not.

Accord99, I can't see you're making any viable argument at all. Everything I said still stands. And it seems a waste to write it all over again since it seems worded clear enough.
Just read it again.
You didn't, for instance, get that the part of having to pay up is going to hit YOU (or your children). What future uses of a computer a consumer will afford to be part of. I'm sure industry and business will find the money. I'm sure Intel is counting on that too.

But to make some further effort, Let's state something obvious very clearly (you might be stupid, what do I know?): I'm talking about what the future will evolve into with Intel monopoly. Not what Intel will do from one day to the next when they achieve monopoly.
New processors are going to be put on market at progressively increased prices. Old lower level CPUs are going to be replaced by new ones at the same price and the same performance but cheaper to manufacture and maybe improved in other ways. By constantly building on the top with sharply increased prices, and roughly keeping the bottom, they are over time going to achieve the wide separation and market stratification they want for maximum profits.

I think Intel would be just fine with people buying computers less often. They will earn more money from leaner manufacture, longer lasting generations and several times higher price on each unit anyway.
Even if people only buy a new computer every 10'th year instead of every 3'rd year Intel will still make much more money if they earn 4 times as much on each due to higher prices. Also they won't stay a CPU provider. They will increase their earnings by advancing their monopoly over other computer hardware.
They can and they will.

However, people will still need new computers more often than every 10'th year . Major reason people buy new computers is rather change of other technology than improved CPU performance. People need more storage, new interfaces to be compatible with new software tools. And you will certainly get change from Intel. I did say that. Just not the kind of change you'd take if you had a free choice.

Until Intel have monopoly on drives of all sorts and memory it seems reasonable to assume that things will progress at a rate which will have people needing new computing appliances at roughly the same intervals as today. Regardless of the progress of CPU performance/cost.
I doubt people will be direct buyers of CPUs. I think that option is gonna disappear. But they will buy things that contain Intel hardware, including CPUs.

 

[Nemesis 1]

Originally posted by: Accord99

Originally posted by: JimiP
I don't feel that to be true. What would we do without CPU's? No matter how costly they get, there will always be a market for them.

Yes, there will always be some who need the absolute cutting edge; but it's not enough to generate $40 billion in revenues and $9 billion in profits every year for Intel if it massively increases prices and greatly reduces product advancements. Intel has grown to where it is today because it sells to the entire market and has enough product advancement to get people and business to upgrade regularly. People used to spending $500 on a whole system are not going to spend $1000, especially if the new system is not much better.

Sorry posted wrong quote this reply meant to for Extelleron

Your forgeting the lawsuite . Intel has to show the courts the reason AMD has little market share is

1) Volumn shipping (Fab Capacity)

2) Performance . Since the lawsuite seems to be talking about pre K8 cpu's . It will be easy for intel to show that even after 3 years of having the performance crown . And losing it . That AMD is right back to were theywere pre lawsuite. I think we know Intel can show this . But intel needs Nehalem To really show this. Plus K10 sure doesn't help AMDs position . Than the delays . K5 was Late not sure about K6 but it did use nexgen tech K7 don't know if late or not . Hammer was late.

 

[Vee]

Originally posted by: Nemesis 1
I wanted to add this bit. How it relates I will leave to you.

I think I can handle that. It doesn't relate at all.

While I haven't posted at all in anand for a very long time, I will not hide from you that I hesitate to respond to you at all. Reason should be obvious to most. But possibly not to you. (sigh)

But anyway: whoever wrote the two paragraphs in your post doesn't fully understand what he's writing about either.
Compiler centric approaches (RISC, VLIW, EPIC) to CPU design for high performance CPUs have sofar proved to be rather lacklustre. And it is certainly not for lack of expenditure on research or development efforts!
RISC is superior for small CPUs but as performance requirements and CPU size increase things change a bit.
A few survive because it's the technical and market heritage of the companies that produce them. And they occupy a niche also featuring very expensive hardware environment for the CPU. Sofar noone has optimized a CISC architecture for such an environment (provided we don't consider Power6 CISC at which point the entire argument falls anyway).

There are very obvious reasons why Power6 performs well. It's big and expensive. And it's designed for and integrated into a very ambitious, expensive hardware environment.
Before one arrive at the conclusion that RISC is superior one should maybe ponder what could be accomplished with 800 million transistors and 4.6GHz with 300GB/s bandwidth with a X86-64 architecture and the same design team that did the Power6. Also one might ponder how much "RISC" Power6 really is.

And as RISC and EPIC has failed for general computing there is some suspicion that the opposite approach - to move complex dependencies, and complexity in general, downwards - is a better approach. This is also reflected in other developments. Consider Java and C#.

 

[Accord99]

Originally posted by: Vee
Intel has not always had products covering the entire market. And one of my points was that you're going to see large differences (virtually nonexistant today) and very major price markups (also pretty trivial today) as you move up the range. That is how they will make the most money.

Intel has always had expensive flagship processors, mainstream processors and cheap processors.

Also I don't think the CPU market is anywhere near saturated. In a decade or so you will find yourself being the owner of a rather large number of Intel 'platforms'. In various things and gadgets. Regardless if AMD survives or not.

The market is very saturated, there are very few new people in the developed world buying computers for the first time. And even 5 year old computers are already adequate for typical word processing, email, browsing tasks. And Intel's biggest competition, like Microsoft, is its own installed base. Computers, like OSes and applications, have reached a point where performance and capabilities are more than good enough.

And in the emerging markets, the demand is for cheap components, not for profitable $200 processors or OSes.

Accord99, I can't see you're making any viable argument at all. Everything I said still stands. And it seems a waste to write it all over again since it seems worded clear enough.
Just read it again.

Most of your "arguments" are weak opinions.

But to make some further effort, Let's state something obvious very clearly (you might be stupid, what do I know?): I'm talking about what the future will evolve into with Intel monopoly. Not what Intel will do from one day to the next when they achieve monopoly.
New processors are going to be put on market at progressively increased prices. Old lower level CPUs are going to be replaced by new ones at the same price and the same performance but cheaper to manufacture and maybe improved in other ways. By constantly building on the top with sharply increased prices, and roughly keeping the bottom, they are over time going to achieve the wide separation and market stratification they want for maximum profits.

I don't see this happening at all, this goes against what Intel has done in its lifetime, even in periods where it has virtually no competition. Nor has Microsoft, which is more dominant than Intel, done anything similar.

I think Intel would be just fine with people buying computers less often. They will earn more money from leaner manufacture, longer lasting generations and several times higher price on each unit anyway.

Even if people only buy a new computer every 10'th year instead of every 3'rd year Intel will still make much more money if they earn 4 times as much on each due to higher prices. Also they won't stay a CPU provider. They will increase their earnings by advancing their monopoly over other computer hardware.

I doubt it, Intel would not be so stupid as to give up their great manufacturing advantages and opens the door for competitors to sell lots of cheap, but still profitable CPUs for OEMs that will want to sell millions of $500 computers every year. Nor do I think Microsoft will want to lose the sale of hundreds of millions of copies of OS and applications every year thanks to a shrinking market.

However, people will still need new computers more often than every 10'th year . Major reason people buy new computers is rather change of other technology than improved CPU performance. People need more storage, new interfaces to be compatible with new software tools. And you will certainly get change from Intel. I did say that. Just not the kind of change you'd take if you had a free choice.

Of course, if few people are buying those new CPUs you will see those storage and software products remaining compatible with the very large existing market. Backwards compatibility has been one of the key feature points of the consumer computer market.

 

[DrMrLordX]

Oy. As important as Intel's future may be to AMD's future, this doesn't have all that much to do with the original post.

AMD is in some trouble, but their acquisition of ATI is working out better than previously thought. They'll be okay. Intel will do what they've always done: make money. How they do this is mostly academic for anyone outside the company.

Arguing with Vee is not a terribly good idea, either (Vee will probably win the argument). Good to see you posting here again, though, Vee.

 

[Pocatello]

Originally posted by: DrMrLordX
Oy. As important as Intel's future may be to AMD's future, this doesn't have all that much to do with the original post.

AMD is in some trouble, but their acquisition of ATI is working out better than previously thought. They'll be okay. Intel will do what they've always done: make money. How they do this is mostly academic for anyone outside the company.

Arguing with Vee is not a terribly good idea, either (Vee will probably win the argument). Good to see you posting here again, though, Vee.

Not really. Just a mixed bag of BS and nonsense. Too wordy, I might add.

 

[Nemesis 1]

Originally posted by: Vee

Originally posted by: Nemesis 1
I wanted to add this bit. How it relates I will leave to you.

I think I can handle that. It doesn't relate at all.

While I haven't posted at all in anand for a very long time, I will not hide from you that I hesitate to respond to you at all. Reason should be obvious to most. But possibly not to you. (sigh)

But anyway: whoever wrote the two paragraphs in your post doesn't fully understand what he's writing about either.
Compiler centric approaches (RISC, VLIW, EPIC) to CPU design for high performance CPUs have sofar proved to be rather lacklustre. And it is certainly not for lack of expenditure on research or development efforts!
RISC is superior for small CPUs but as performance requirements and CPU size increase things change a bit.
A few survive because it's the technical and market heritage of the companies that produce them. And they occupy a niche also featuring very expensive hardware environment for the CPU. Sofar noone has optimized a CISC architecture for such an environment (provided we don't consider Power6 CISC at which point the entire argument falls anyway).

There are very obvious reasons why Power6 performs well. It's big and expensive. And it's designed for and integrated into a very ambitious, expensive hardware environment.
Before one arrive at the conclusion that RISC is superior one should maybe ponder what could be accomplished with 800 million transistors and 4.6GHz with 300GB/s bandwidth with a X86-64 architecture and the same design team that did the Power6. Also one might ponder how much "RISC" Power6 really is.

And as RISC and EPIC has failed for general computing there is some suspicion that the opposite approach - to move complex dependencies, and complexity in general, downwards - is a better approach. This is also reflected in other developments. Consider Java and C#.

You make a good case based on old thinking. In tomorrows Cpus. lets say 16 cores and up . X86 low order cpu's will fall by the wayside. Simpler inorder high level cores without the X86 baggage will be the way forward. We could dedate this till were blue in the face. Intel is lucky they picked up Apple as a Vender. As Apple contiues to take market share Intel will focus more on their platform . Which in the long run will benefit us all.

Lucky for us Larrabee will be out end of 08 start of 09 . But we will know a lot more than that come spring /fall IDF. 16 inorder 2threads per core with ability to run x86 apps. Its not that far off. Than we will know were intel is heading . AMD can't take this course as they don't have the software. This time Intel doesn't have to share this tech with AMD thank GOD. Intel will retain the ability to run X86 apps. while at the same time going to a much simpler design . It works for me. In the not to distant future IDF more and more web masters will come to same conclusion . SPRING IDF.

 

[Fox5]

Simpler inorder high level cores without the X86 baggage will be the way forward. We could dedate this till were blue in the face. Intel is lucky they picked up Apple as a Vender. As Apple contiues to take market share Intel will focus more on their platform . Which in the long run will benefit us all.

If this was true, why didn't Apple go with Cell? Apple didn't want asymmetric processors.

 

[mruffin75]

Originally posted by: Nemesis 1
You make a good case based on old thinking. In tomorrows Cpus. lets say 16 cores and up . X86 low order cpu's will fall by the wayside. Simpler inorder high level cores without the X86 baggage will be the way forward.

People have been saying the same thing for at *least* the last 20 years..and it never happens...

 

[Nemesis 1]

Originally posted by: Fox5

Simpler inorder high level cores without the X86 baggage will be the way forward. We could dedate this till were blue in the face. Intel is lucky they picked up Apple as a Vender. As Apple contiues to take market share Intel will focus more on their platform . Which in the long run will benefit us all.

If this was true, why didn't Apple go with Cell? Apple didn't want asymmetric processors.

One reason could be that cell can't and never will beable to do X86 apps. . It could also be that Apple knows more about what Intel has coming than we do . Besides if one looks at Apples PC market share gains it looks like Jobs hit a home run. All those years with IBM and Apple lost market Share.


mruffin75
Won't be long wait now . Larrabee will show the direction intel is moving. We already know what intel has told us about Tera Scale processors . Larrabee is the first. Coming to a resellernear you in late 08 early 09. Not long to wait.

I am also very interested in Bulldozer could be great . Man I love this stuff.

 

[Nemesis 1]

Back on topic.

http://www.xbitlabs.com/news/c..._AMD_Phenom_Chips.html

 

[DrMrLordX]

That article certainly is interesting:

As those who wanted to have high-speed AMD Phenom processors ask why the chipmaker decided to delay their desired microprocessors, the maker of central processing units just claims that hardly anyone expects high-performance chips by AMD now.

Words escape me. But hey, at least it's an honest assessment.

The world?s largest x86 chip manufacturer did not unveil whether its customers did not want to get higher-speed Phenom chips, or the company had to delay the release of its higher-speed chips, or it could not get the new steppings of chips in time.

"world's largest x86 chip manufacturer"? Huh? Did I miss something?

 

[Nemesis 1]

I believe they were referring to die size.

 

[Idontcare]

Originally posted by: Nemesis 1
I believe they were referring to die size.

LMAO, that was good.

Seriously though this move by AMD is not so much unlike Intel's move to not release a 4GHz P4 chip after all that grandiose demo'ing "netburst FTW" 10GHz crap they put out at IDF.

AMD is following a script written by Intel here from my perspective, only losing a buttload of money in the process. Intel did what for their worst quarter? $1B net profit at worst to recollection.