Nvidia in negotiations with Global Foundries
- Thread starter Viditor
- Start date
I see no other option for them, nvidia has everything to lose if they gamble on TSMC keeping pace with GF.
By using the same foundry as ATI they can just focus on 'out engineering' them rather than risk losing the process lead.
Which does beg the interesting question, was this really a smart move for AMD as a whole? .. if they could have used their tech to boost ATI ahead of nvidia before the spun off GF.
By using the same foundry as ATI they can just focus on 'out engineering' them rather than risk losing the process lead.
Which does beg the interesting question, was this really a smart move for AMD as a whole? .. if they could have used their tech to boost ATI ahead of nvidia before the spun off GF.
Just as the cost of developing the tech outstripped the ROI needed to justify developing it inhouse, so too did the cost of manufacturing with the tech.
For AMD there really was little choice. To be sure their GPU's would have had a sizable process tech advantage had they kept the fabs and their process tech solely to themselves.
The process tech advantage could have been combined with lackluster design to mask the lacklusterness or combined with a comparable design to deliver some real knock-out advantages performance/Watt wise.
Consider Intel's situation with Larrabee, the design approach could suck but leading edge process tech could make all the difference once drivers have been ironed out.
So yeah AMD "gave away" what could have been a great competitive advantage over Nvidia when it comes to GPU's, but unfortunately it probably is that very ATI purchase they put them into the ER and ICU (financially speaking) which made it impractical for them to keep their fabs to begin with. $5.4B is a lot of coin, had they invested it differently back in 2006.
As it stands they are no better or worse off than Nvidia, as you so rightly put it, ATI has to focus on "out engineering" (although really we should call it designing) Nvidia...which is the same whether they were fabbing at UMC or TSMC anyways.
And really if you think about it, if AMD can't "out design" Nvidia then they got real fundamental problems with their business model considering they are just a design house now.
I'll be surprised if GF doesn't pick up SUN too. TSMC is too used to being top-dog in the foundry world with their only real competition being UMC. GF, with their node development heritage, jumping into the pool at the top of the node tree like they are is going to show this. AMD was second in the world when it came to process tech performance and delivery timeline for high-performance logic, TSMC doesn't even know how to compete in that marketspace (their xtor development model is more mobile/low-performance and general purpose oriented...anyone who needed high-performance xtors already had their fabs or dedicated foundries up until now).
It is a win-win for both AMD and Nvidia. AMD is no longer saddled with the extra expense of trying to run it's own foundries and Nvidia has access to advanced chip making capabilities of the competitor.
i see this as being a win/win situation for ati and nvidia, as well as GF. if one or the other chip designer doesnt do so well, the foundry technology driving both companies chips wont slack behind due to decreased sales nearly as much, since buying chips from the competition will just increase demand for a different chip being fabbed on the same manufacturing lines, and with 2 major developers putting resources into GF now, it will help in the long run in competing with intel's process tech lead, which could show up as benefits in AMDs CPU space as well
They have HMK tech almost ready for prime time, so the answer is most definitely yes on that one.
For high-performance logic on SOI, yes.
For low-power (mobile stuff) and general purpose (cost sensitive) on bulk-Si cmos, no.
Come 32nm (2010) GF intends to change that "no" to a "yes". Time will tell.
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,570
- 10,199
- 126
uh it would make sense, considering it's probably cheaper for AMD to fab their GPUs at GF since they have like a 49% share in the company
They have already stated they are going with GF
Not knowing anything about it, I would actually expect them to be well into designed for GFs node, and have products out on it before nvidia.
Not knowing anything about it, I would actually expect them to be well into designed for GFs node, and have products out on it before nvidia.
It stands to reason to believe AMD had plans to fab gpu's in-house, whether those plans were actual commitments or just options on the table is something we'll never know unless Anand can score another insider story like the RV770 scoop.
But currently neither AMD nor NV fab their gpu's at GF, so we can't say definitively who is following who here if and when one side or the other releases a gpu fabbed by GF's. (since product release timeline does not provide concrete engagement dates either)
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 18K
-
-
-
Facebook
Twitter