Jen Hsun Huang says NO to Globalfoundries

[thilanliyan]

http://www.xbitlabs.com/news/other/...ys_No_to_Globalfoundries_Microprocessors.html

 

[Ben90]

Well hopefully TMSC can get their 40nm yields up, otherwise if Huang avoids GF because its an "AMD company", NVidia may not have the availability advantage over ATI if they switched.

Im sure if things came down to it, Huang would have no problem moving over to GF if TMSC was fecking up bad.

 

[IlllI]

still waiting for him to 'open a can of whoop-ass' on intel

 

[jvroig]

still waiting for him to 'open a can of whoop-ass' on intel

+1. Let's see what happens in the NV/Intel saga

 

[dflynchimp]

still waiting for him to 'open a can of whoop-ass' on intel

I wouldn't lose sleep on it. A rule of the thumb these companies seem to abide by is to treat PR with a refresh cycle. They expect, with a certain amount of correctness, that whatever their marketing team comes up with will hold an effective attention span for maybe a year or so, and if they contradict themselves somewhere down the road the majority of the people who count (aka consumers) won't remember.

It's techies like us who care enough about this stuff to keep a back log in our memory banks

 

[BenSkywalker]

still waiting for him to 'open a can of whoop-ass' on intel

Last time I checked, the GTX295 was utterly destroying every Intel graphics chip by at least an order of magnitude. I tend to consider that beyond a 'can of whoop ass' actually, quite a bit beyond. His comment was directed towards Intel's challenge on the graphics front and quite frankly, they are too scared to even enter the ring.

The real whoop ass that Intel should truly fear will be coming to a court room soon. Both AMD and nVidia stand to benefit if nV wants to play hard ball.

 

[nomagic]

Well hopefully TMSC can get their 40nm yields up, otherwise if Huang avoids GF because its an "AMD company", NVidia may not have the availability advantage over ATI if they switched.

Im sure if things came down to it, Huang would have no problem moving over to GF if TMSC was fecking up bad.

How many other foundries besides TSMC currently has 40nm bulk process?

Why do people just assume that GF's 28nm and 32nm process is destined to kick TSMC's ass?

There is got to be something that I overlooked. I mean it doesn't make sense at all.

 

[MrK6]

Why is this news?

 

[Acanthus]

How many other foundries besides TSMC currently has 40nm bulk process?

Why do people just assume that GF's 28nm and 32nm process is destined to kick TSMC's ass?

There is got to be something that I overlooked. I mean it doesn't make sense at all.

GF is pretty much the premier "for contract" fab out there now. If it remains managed like it was under AMD, it should continue to be that way for the foreseeable future.

They retain a lot of talent at that fab.

 

[Idontcare]

Why is this news?

It is relevant to any folks who are wondering how and when GF's high-performance SOI process technology (which TSMC does not have) might come to represent an advantage for any given GPU manufacturer to transition towards at the 32nm and beyond nodes.

Nvidia absolutely ruling out GF's also means NV is absolutely ruling out SOI.

That opens the door of opportunity for AMD to migrate their future discreet GPU IC's to GF's SOI processes and retain the performance advantage for themselves as Nvidia is apparently not interested in keeping the process technology playing field an even one by migrating their GPU's to the same process tech that AMD is.

 

[MrK6]

It is relevant to any folks who are wondering how and when GF's high-performance SOI process technology (which TSMC does not have) might come to represent an advantage for any given GPU manufacturer to transition towards at the 32nm and beyond nodes.

Nvidia absolutely ruling out GF's also means NV is absolutely ruling out SOI.

That opens the door of opportunity for AMD to migrate their future discreet GPU IC's to GF's SOI processes and retain the performance advantage for themselves as Nvidia is apparently not interested in keeping the process technology playing field an even one by migrating their GPU's to the same process tech that AMD is.

My comment was to point out this isn't surprising. I think it would have to be a cold day in Hell (insert your own idiom) before you'd see NVIDIA dumping large amounts of cash into its direct competitor. If I'm not mistaken, AMD still owns the majority of GF, yes? My assumption is GF could be producing a condensed-reincarnated god on a PCB and NVIDIA would still explore other options exhaustively. Thank you for the information on the tech though, that is very interesting.

 

[Fox5]

I wouldn't lose sleep on it. A rule of the thumb these companies seem to abide by is to treat PR with a refresh cycle. They expect, with a certain amount of correctness, that whatever their marketing team comes up with will hold an effective attention span for maybe a year or so, and if they contradict themselves somewhere down the road the majority of the people who count (aka consumers) won't remember.

It's techies like us who care enough about this stuff to keep a back log in our memory banks

Agreed, what nvidia says today usually isn't what they do tomorrow.

But it takes years to design a product from the ground up on a new process, so nvidia wouldn't have an option to go GF until 2011/2012 at the earliest anyway, so what would be the point in talking about it? Also, will AMD get preferential treatment at GF?

What kind of capacity does GF have anyway? Back in the Athlon 64 days, spinning off the fab was considered a worthless move, since AMD was using all of the fab space anyway, and cpus are one of the highest profit items per die size so it didn't make sense for AMD to produce graphics chips on it. Sure they have more fabs now, with 2 or 3 up to date fabs, but it's not inconceivable that AMD by itself could use up all the initial 32nm fab space, leaving ATI and the rest for when the process becomes more mature.

 

[Idontcare]

How many other foundries besides TSMC currently has 40nm bulk process?

Why do people just assume that GF's 28nm and 32nm process is destined to kick TSMC's ass?

There is got to be something that I overlooked. I mean it doesn't make sense at all.

GF is pretty much the premier "for contract" fab out there now. If it remains managed like it was under AMD, it should continue to be that way for the foreseeable future.

They retain a lot of talent at that fab.

The relevant portion of nomagic's post is in the Bulk-Si CMOS domain and the confidence with which people appear to be willing to assume GF's process technology dominance in SOI somehow ensures it equal process technology dominance in Bulk-Si...yours speaks more to how well the fab runs (and why) regardless of the technology aspects of the underlying process technology.

Dresden has little to do with the development of the process technology that it will eventually be tasked with manufacturing. They do provide critical feedback regarding things that don't work or are discovered to be broken once the technology has transferred to Dresden (32nm BEOL dielectric cracking issues in packaging tests for example) but that is about the limit on the extent of the feedback loop.

The question of how well GF's bulk-Si process tech is going to turn out relative to TSMC's can perhaps best be answered by asking yourself who among IBM's bulk-Si development alliance is currently competing with TSMC's 40nm node? Whatever the gap is there between the best the IBM alliance has to offer for bulk-Si at 40nm and the best TSMC has to offer is probably good indication of the performance gap that will likely be present in GF's 32nm and 28nm processes.

 

[yh125d]

Not surprised at all. I know I wouldn't be champing at the bit to get fabbed by a competitors joint venture like that



Although I hope Nv moves to SOI sooner or later, whether at TSMC or elsewhere. ARM has already demonstrated the power performance which seems so much better than current techs, and I think GPUs from now on will need it, since they seem to be growing in power (heat) faster than node shrinks lower the die size and thermal characteristics

 

[T2k]

still waiting for him to 'open a can of whoop-ass' on intel

Rrrrrrrrrriiiiiiiight... :biggrin:

 

[Idontcare]

My comment was to point out this isn't surprising. I think it would have to be a cold day in Hell (insert your own idiom) before you'd see NVIDIA dumping large amounts of cash into its direct competitor. If I'm not mistaken, AMD still owns the majority of GF, yes? My assumption is GF could be producing a condensed-reincarnated god on a PCB and NVIDIA would still explore other options exhaustively. Thank you for the information on the tech though, that is very interesting.

True that, you make a good point regarding the cash-angle...I had previously responded to another poster regarding their concerns over "information transfer" between NV and AMD being the reason NV would not migrate to GF (my response was that the info transfers regardless the foundry selection) but I had not contemplated the fact that the last thing Jensen wants to do is have his company validate AMD's fabless model by adding cash to GF's revenue.

We know Jensen to be a passionate person, some might call it ego, and I doubt he wants to be part of the validation of AMD's business model after the failed merger negotiations with Ruiz. That might change with Ruiz out of the picture though, bruised ego's heal in different ways.

 

[Idontcare]

Not surprised at all. I know I wouldn't be champing at the bit to get fabbed by a competitors joint venture like that



Although I hope Nv moves to SOI sooner or later, whether at TSMC or elsewhere. ARM has already demonstrated the power performance which seems so much better than current techs, and I think GPUs from now on will need it, since they seem to be growing in power (heat) faster than node shrinks lower the die size and thermal characteristics

ARM reports 45-nm SOI test chip with 40% power-saving

The results show that 45-nm high-performance SOI technology can provide up to 40 percent power savings and a 7 percent circuit area reduction compared to bulk CMOS low-power technology, operating at the same speed. This same implementation also demonstrated 20 percent higher operating frequency capability over bulk while saving 30 percent in total power in specific test applications.

http://www.eetimes.com/showArticle.jhtml?articleID=220301622

Agreed.

 

[SlowSpyder]

I would think this potentially gives AMD an advantage. AMD can use whoever has the better process at the time, or even multiple fab companies. Nvidia is banking a lot on TSMC, if their process has problems then so does Nvidia. Though I can understand their decision as using GF ultimately could also strengthen AMD.

Last time I checked, the GTX295 was utterly destroying every Intel graphics chip by at least an order of magnitude. I tend to consider that beyond a 'can of whoop ass' actually, quite a bit beyond. His comment was directed towards Intel's challenge on the graphics front and quite frankly, they are too scared to even enter the ring.

The real whoop ass that Intel should truly fear will be coming to a court room soon. Both AMD and nVidia stand to benefit if nV wants to play hard ball.

So Nvidia wants to open a 'can of whoop ass' where Intel really isn't even (yet?) competing? I guess in that case Intel has opened a can of whoop ass on Nvidia in the SSD market.

 

[yh125d]

Agreed.

They continue to keep adding power connectors and bigger coolers, but they're going to reach a maximum soon! Hopefully SOI will reverse that


6800GT - top end card circa 2004

GTX280 - top end card circa 2008

I don't like this trend!

 

[MrK6]

True that, you make a good point regarding the cash-angle...I had previously responded to another poster regarding their concerns over "information transfer" between NV and AMD being the reason NV would not migrate to GF (my response was that the info transfers regardless the foundry selection) but I had not contemplated the fact that the last thing Jensen wants to do is have his company validate AMD's fabless model by adding cash to GF's revenue.

We know Jensen to be a passionate person, some might call it ego, and I doubt he wants to be part of the validation of AMD's business model after the failed merger negotiations with Ruiz. That might change with Ruiz out of the picture though, bruised ego's heal in different ways.

Well, from my limited knowledge of the whole situation, Ruiz was incompetent and dragged AMD down, it's a good thing he's in hot water now (and should be all but exiled from the company). But in general I think we're saying the same thing - the competition between AMD and NVIDIA will prevent NVIDIA from considering GF as an alternative solution for some time. And thank you for the link on the SOI test chip - 40% power savings is downright amazing.

I would think this potentially gives AMD an advantage. AMD can use whoever has the better process at the time, or even multiple fab companies. Nvidia is banking a lot on TSMC, if their process has problems then so does Nvidia. Though I can understand their decision as using GF ultimately could also strengthen AMD.

I definitely agree. Having options is never a bad thing.

So Nvidia wants to open a 'can of whoop ass' where Intel really isn't even (yet?) competing? I guess in that case Intel has opened a can of whoop ass on Nvidia in the SSD market.

I consider Intel on a completely different level than the rest of these companies. In all honesty, I see Intel as competing only with itself. The size, depth, and resources of the company are outright amazing. I mean, I don't really see Intel partaking in the kind of shit talking that both NVIDIA and AMD do, and my assumption is they don't need to. So NVIDIA can say all it wants and it probably barely registers with Intel.

 

[T2k]

I would think this potentially gives AMD an advantage. AMD can use whoever has the better process at the time, or even multiple fab companies. Nvidia is banking a lot on TSMC, if their process has problems then so does Nvidia. Though I can understand their decision as using GF ultimately could also strengthen AMD.

I agree - top dogs at AMD have learned their lesson from chasing the now-infamous .09 micron bug in their libraries that held up their R520 launch for several quarters.
Aside of making money on customers on the long run GF can save tremendous amount of money for AMD anyway.

So Nvidia wants to open a 'can of whoop ass' where Intel really isn't even (yet?) competing? I guess in that case Intel has opened a can of whoop ass on Nvidia in the SSD market.

Yeah, makes no sense. The only market they WERE competing was the integrated mobile chipsets - and Nvidia got whooped by Intel pretty badly, let alone its full of crap buggy chips and other tricks that pretty much killed the mobile GF brand for while or at least drove away most of the OEMs (google up Apple showed the door Nvidia or Dell drops Nvidia etc.)

 

[shangshang]

....

It's techies like us who care enough about this stuff to keep a back log in our memory banks

nah.. who cares about a pissing contest between 2 super rich CEOs. They come and go

 

[SunnyD]

The only thing that can run a company into the ground faster than government regulations: Foolish pride.

 

[dflynchimp]

Most of us don't give a rats ass about CEO catfights, true, but when I say "cares about" I mean caring about the technology itself enough that we keep up with the tech news and have a good memory of things related to it.

And to be fair to Nvidia, it's not like ATI hasn't said things that it later swallowed. IRCC Nvidia was bashed by ATI for being the first to use bulky dual slot cooling in the FX5800's, yet by the next generation (Radeon X850XT) ATI had followed suit. The same goes for extra power connectors.

I could be talking out of my ass here since this bit of memory comes from a second hand recollection from some other tech forum two-three years back, so someone correct me if I'm wrong.

The point is don't anyone be specifically going "Nvidia's a lying conniving whore" when the truth is Nvidia is just like any other company who has dough to spend on PR and anti-competitive practices like vendor locks. (see: Apple). Badmouthing and shady business practices aren't unique to Nvidia, it's just the way things are.

 

[ZimZum]

My question is why is ATI even using TSMC at all?