Possible new stepping for Winchesters in January?

[trans am]

http://www.theinquirer.net/?article=20049

[RaptorRaider]

Old news... E0 was planned a long time ago.

?
Really, I thought the whole article is an example of a bad journalism. There are no facts there anywhere, only some vague "new chips" to show up and that AMD will be shouting something. Whatever.

You must be new to 'The Inquirer'...

LOL:

(1) According to our server logs, the majority of humanity still does not read The INQ. While this is reprehensible, it does mean that AMD has a chance at surprising a good many people with the news.

[joe2004]

?
Really, I thought the whole article is an example of a bad journalism. There are no facts there anywhere, only some vague "new chips" to show up and that AMD will be shouting something. Whatever.

[trans am]

?
Really, I thought the whole article is an example of a bad journalism. There are no facts there anywhere, only some vague "new chips" to show up and that AMD will be shouting something. Whatever.

That's the inquirer for you...

[mdzcpa]

?
Really, I thought the whole article is an example of a bad journalism. There are no facts there anywhere, only some vague "new chips" to show up and that AMD will be shouting something. Whatever.

I actually appreciate Mike's style of news. Once you get used to it, I find it very informative. Its not meant to be read like a traditional news source.

I'm sure there are plenty of industry buffs out there that know exactly what I'm talking about

That said, I wouldn't be surprised to find that AMD puts forth a large marketing buzz in January to surround the new steppings just as the Inq says to expect.

[Beenthere]

Charlie D. who writes some articles for The Inquirer is the one who tends to get carried away with "inside info." that isn't alway accurate...

It's been know for months that AMD was going to add SSE3 instructions to chips in Q1 '05. The main reason for doing this is so that the benchmarks Intel has paid for to show Pig 4 chips in a positive light, use SSE3, even though most software does not. By adding SSE3 to AMD chips then the bogus benches that Intel has paid to have created, will show AMD chips in a better light.

In addition, the 90 nano process AMD is using is their own, NOT IBM's and it ain't dumb by any stretch. AMD's big news for '05 is top notch dual core CPUs where they will kick Intel's azz even more than they did with the die shrink to 90 nano. Intel is still working on a hacked 32/64-bit X86 process as they have nothing worth buying for another two years.

[mdzcpa]

FYI: The current .09 process, and the improved new strained silicon SOI process on the upcoming stepping in Q1, were both devloped by joint venture between AMD and IBM. And most insiders are also well aware that the process "tuning" was done by IBM.

We'll be seeing more too. In December 02 these guys signed a 3 year deal to develop semiconductor manufacturing technologies for 65-nanometer and 45-nanometer. In September 2004 they tacked on another 3 years as well extending their partnership through December 31, 2008. We could be seeing these guys issuing joint product through the 32 nm process.

AMD has also now licensed IBM's C-4 "flip chip" packaging technology, which is used on IBMs server processors.

We'll see that both IBMs PowerPC chips and AMDs K8 will "improve" simultaneously in Q1...it will not be a mere coincidence

[Wolf]

overclockers.com also has an article on it on the front page. I'm holding out to upgrade to 939 until the new stepping hits and I can get a sli Nf4 board. It should be a nice improvement in the memoy controller and clock speeds as well as sse3 instuctions.

[WiCKeD]

Thanks trans, news to me. Doesn't help me decide things any better though.

[OC Detective]

FYI: The current .09 process, and the improved new strained silicon SOI process on the upcoming stepping in Q1, were both devloped by joint venture between AMD and IBM. And most insiders are also well aware that the process "tuning" was done by IBM.

We'll be seeing more too. In December 02 these guys signed a 3 year deal to develop semiconductor manufacturing technologies for 65-nanometer and 45-nanometer. In September 2004 they tacked on another 3 years as well extending their partnership through December 31, 2008. We could be seeing these guys issuing joint product through the 32 nm process.

AMD has also now licensed IBM's C-4 "flip chip" packaging technology, which is used on IBMs server processors.

We'll see that both IBMs PowerPC chips and AMDs K8 will "improve" simultaneously in Q1...it will not be a mere coincidence

The SSDOI process that AMD will be adopting is different from IBM's.
http://www.theregister.co.uk/2004/08...90nm_strained/

[reject]

ive been saying this for a while
why did they miss out 110 nm?

[Beenthere]

For those who don't know about AMD and IBM's 90 nano processes whick are different, AMD helped IBM resolve their problems and that is why IBM is keen on fabbing AMD processors. You'll note that AMD has scaled quickly while IBM is just getting up to speed... The current AMD products is an all AMD 90 nano process. It is not IBMs work at all.

[Geforce4ti4200]

"3 year deal to develop semiconductor manufacturing technologies for 65-nanometer and 45-nanometer. In September 2004 they tacked on another 3 years as well extending their partnership through December 31, 2008. We could be seeing these guys issuing joint product through the 32 nm process."


How does one know what comes after 90nm? it could be 75nm and where did 45nm come from? Not that it even matters, the limits of deep ultraviolent litography is supposedly 70nm, 65nm would be a real stretch but we would need an entirely new process for any smaller and also the leakage would be so great. 90nm already has problems, 45nm would quad those problems!

[conrad.maranan]

After a lengthy cionversation with AMD earlier today, I still have my doubts that it'll be a January release, regardless of how much assurance the man on the other end gave me.

[saaya]

the release date of early in 2005 is new, there was no release date known before...

[Geforce4ti4200]

I shall mark my calandar, my winchester is begging to be upgraded pretty soon. will any of those new steppings trickle down to 3500+ or even 3200+ and will we see over 3GHz air? finally if prices are $200 or less, I am sold always hungry for more power, albet must be affordable to me. NO fx chips

[Super strokey]

i thought you were still waiting to get the winnie? If not maybe you should upgrade the vidcard first

[mdzcpa]

The SSDOI process that AMD will be adopting is different from IBM's.
http://www.theregister.co.uk/2004/08...90nm_strained/

They will be the same by first quarter 05. That was the point of the original article.

For those who don't know about AMD and IBM's 90 nano processes whick are different, AMD helped IBM resolve their problems and that is why IBM is keen on fabbing AMD processors. You'll note that AMD has scaled quickly while IBM is just getting up to speed... The current AMD products is an all AMD 90 nano process. It is not IBMs work at all.

This is hilarious. I'd suggest you go back and do some research on the joint partnership. It is AMD who paid cash to IBM for development process technology ealy in 2003....not fabbing.

Furthermore, AMD 09 has NOT scaled well at all. Where are the high frequency 09 chips? That's right...they can't make em.

How does one know what comes after 90nm? it could be 75nm and where did 45nm come from?

The engineering concepts for these shrinks are already on the table. The preliminary enginnering happens years before you actually see any chips.


Wow...all you guys live under a rock or something? Here's some reading material to enlighten all of you:

The orignial agreement - IBM hlps AMD with SOI technology

Integrating SOI has been a problem for AMD. The company was going to include SOI in an upcoming chip called Barton, but it removed the technology and delayed the chip. Until now, AMD was obtaining its SOI technology from Motorola. A Motorola representative said Motorola and AMD mutually agreed to end their pact in early 2002.

IBM, of course, will also benefit. The company rivals Intel in semiconductor research, but its chip sales are only one-eighth as large. Licensing its technology, and manufacturing chips for other companies, opens revenue streams for Big Blue

IBM's services and technology don't come cheap, though. Companies typically hand over several million dollars--even hundreds of millions of dollars, depending on the deal--to IBM

The partnership extends

IBM (Profile, Products, Articles) Corp. and Advanced Micro Devices Inc. (AMD (Profile, Products, Articles)) have tacked a three-year extension onto their December 2002 collaboration agreement that could see the two companies work on chip-making technologies based on a 32 nanometer process. The agreement was set to expire at the end of 2005, but now extends until Dec. 31, 2008.

Under the revised agreement, dated Sept. 15 and included in documents filed with the U.S. Securities and Exchange Commission Tuesday, AMD will pay IBM between $250 million and $280 million over the next four years. In exchange, IBM will continue to host the research and development work at its East Fishkill, New York, facility.

AMD struggles with high frequency 90 chips

What we’re looking at here is that the 90nm SOI process developed by AMD and IBM runs into a similar set of problems as Intel has. At some point the processor needs to be fed with a higher voltage to enable higher clockspeeds but due to the higher voltage starts leaking power which significantly increases its heat production. That is exactly why IBM fails to ship 3GHz PowerPC processors to Apple, the heat production would simply be too much to properly manage. But there’s more, at these higher voltages the maximum POH, power on hours, of the processor is cut by a significant margin, at a 1.3-volts core voltage it is cut to <50.000 hours, whereas at 1.2-volts it is 100.000 hours. Of course PowerPC processors are not to be compared directly to Athlons, but the process technology used is the same, so the same can be applied to them, explaining exactly why AMD has yet to introduce faster processors based on the 90nm process.

So what does this tell us? Simply that AMD is having its own set of problems with 90nm, it is hitting a clockspeed wall, much like Intel has, which prevents them from introducing faster processors. Unlike Intel, that basically launched their Prescott core Pentium 4 processors regardless of their power consumption and heat production, AMD only released processors that run at a, relatively speaking, low clockspeed, so heat production is manageable. One of the reasons why they have not been promoting their 90nm processors that much is to simply avoid the obvious question why they don’t ship faster processors at 90nm, which would be more economical and also result in higher volume shipments. The answer, as explained, is that they simply can’t yet as they’ll run into exactly the same problems as Intel, with processors generating over a 100-watts of heat. In a few months time however, when the second generation 90nm process debuts, again developed in a joint venture with IBM, it should alleviate some of these problems and make 3GHz or even faster AMD processors a reality. Until then don’t expect AMD to shift it’s full product line to 90nm just yet, once they do issue a press release about their top-of-the-line processor being manufactured in 90nm, you’ll know they pulled it off.

Wow....I cannot beleive there is anyone here who denies that AMD has been helped by IBM to develop the SOI (and now strained silocon) technology. Seeing that AMD never had a process of thier own to begin with

An "all AMD 90 nano process" ...holy smokes the AMD fanboyism is in full swing. You all need to calm down. Its perfectly "okay" that AMD is licensing process technology from others. Its the chip design that really shines.

[Beenthere]

Ya just have to laugh at people without a clue who believe they are "experts" and they will tell ya so.

[joe2004]

Ya just have to laugh at people without a clue who believe they are "experts" and they will tell ya so.

[IvanAndreevich]

Geforce4ti4200
Yes these chips will go 3 GHz on air. Happy?

[mdzcpa]

Ya just have to laugh at people without a clue who believe they are "experts" and they will tell ya so.

LOL...based on your lack of any information to the contrary its pretty obvious who doesn't have the clue here...but that's okay, I won't laugh at you. Ignorance is to be pitied not made fun of

[texuspete00]

I know who I believe, and it starts with an M. I mean why the hell not, that's what it says when you look around a bit. Not to mention they have partnered and IBM has used both SOI and strained silicon before AMD. That said the egos are flying... at least we are straightening it out though. AMD has talented engineers too and I'm sure Big Blue isnt doing all this just to be nice, or even for just $280 million. As it is right now though, it seems reasonable for the AMD fans to be thankful of IBM. I for one am glad. WHen intel successfully shrunk to 130nm first, AMD was on the sidelines for awhile. Any time AMD seems to have the upper hand, it mostly seems to come from smart chip engineering, not fab technology and shrinks. Now with some help, you are hard pressed to find a review suggesting you buy a prescott over an A64.

Same thing with GF.... yes, if you look... the next chips are 65nm. Thats just the way it is. Not to put you down, but there is a lot of interesting things outside of 3dmark01. I think it has something to do with the engineering as well. Notice the chip makers usually settle on some common denominator. AMD/Intel/IBM then nV/ATi on 110nm. Cpu makers probably just skipped 110nm(like they did 150nm) because in my not so expert opinion, thats not really that much of a shrink. From 180 to 130 to 110? One thing seems certain GF, you can't just throw out why not 70 etc.

[OC Detective]

AMD's SSDOI technology on 90nm will differ from that currently used on the FX55's. The SSDOI used at present is similar (read identical!) to that used by IBM however for 90nm (second generation) it will differ as it will be localised to certain components and not involve the removal of silicon germanium.

[OC Detective]

The progress from 180 to 130 to 90 to 65 to 45 roughly represents a halfing of the area required for the same number of transistors (or a doubling of transistors for the same area!) - pretty simple really.