Nvidia in Discussions with Globalfoundries over Manufacturing

[LydianKnight]

Nvidia in Discussions with Globalfoundries over Manufacturing (http://www.xbitlabs.com/news/other/display/20090617070941_Nvidia_in_Discussions_with_Globalfo undries_over_Manufacturing__Chief_Executive_Office r.html)

Doesn't necessarily means NVIDIA is not going to deliver their GT300 this year, but maybe this means if negotiations go fine, we could see a 32nm refresh for next summer, courtesy of GlobalFoundries, as they're taking 32nm fabrication orders this late year

Sounds promising if this comes to a true end ^_^

[LydianKnight]

Hah, even a bit more of heat from BSN ^_^

Is NVIDIA really moving to GlobalFoundries? (http://www.brightsideofnews.com/news/2009/6/18/is-nvidia-really-moving-to-globalfoundries-.aspx)

So, as I suspected, NVIDIA is _really_ pissed off with the low results from TSMC's 40nm half-node process, so maybe they're already converting the GT300 design files to use GlobalFoundries libraries, just naive talk, of course, hyper-grain of salt recommended...

Things are becoming really interesting, although would be nice to have GT300 this year, not next one, but it all depends... let's hope for the best in any case

[a12ctic]

If Nvidia moved to Global Foundries wouldn't they theoretically be funding AMD/ATI indirectly?

[LydianKnight]

GlobalFoundries is a joint venture between ATIC and AMD, so in a sense, yes... but I wouldn't say ATI but plainly AMD, anyway, they could be benefiting from that, nanotechnology research, collaboration programs between AMD and IBM about SOI transistor technology and other interesting topics, advanced process nodes... you name it.

I see this possible migration interesting because current TSMC process seems to be quite leaky, and I personally don't want to end buying a GT300 if it's going to take a full 300W just for a single GPU card... (and they're targeting 225W, let's see...)

Better processes means better consumption levels, better clocks, smaller size, depends on the design, of course, but I think NVIDIA has spent too many money and resources and as I told in a previous post, business fidelity means nothing, you just need the best technology from the best vendor, so I think it's the time to leave TSMC behind

but it's just my opinion :P

[josiahsuarez]

it's interesting, GF certainly has en impetus to attract new customers. but if GT300 is switched over from TSMC, that would mean basically a respin. like BSN says Huang is possibly just looking for leverage with TSMC by making a public statement.

[LydianKnight]

Taking into account GlobalFoundries is not aiming at 40nm, but rather on mature 45nm, 32nm and 28nm for now, I'd say more like NVIDIA is going to deliver GT300 on TSMC's 40nm process, and future versions, including refreshes, could be 'courtesy' of GlobalFoundries, given that they have to send the orders late this year, so I think it could be more like that

[LydianKnight]

Huang is possibly just looking for leverage with TSMC by making a public statement.

Well.. this is just a thought, but... usually in the electronics manufacturing world (I've worked for a Fujitsu fab here in Malaga, so I more or less can depict you the kind of problems arise in production) pressing your engineers usually lead to bad results, and usually any threat is given privately, because if you go for it in the open it just leads to speculation and related phenomena.

What I mean is... if NVIDIA is really considering doing some serious business with GlobalFoundries, maybe it's just they're considering doing their future line of chips with them, although it could also serve as kind of a 'scare' technique for TSMC to awaken, but I'm more in the side of thinking it's a pre-announced migration, and it would be good.

From a practical perspective, we could all be benefited if NVIDIA makes the migration, because GlobalFoundries is supposed to take on 32nm orders late this year and risky orders for 28nm on summer 2010 or H2 2010, don't remember correctly, that means if NVIDIA takes some time to redesign or convert their existing designs with GlobalFoundries design libraries, we could have the chance to have a better product, less (to be supposed) leaks which would lead to lower power consumption, or even keeping their target 225W envelope for a GT300 refresh but with higher clocks, which in the end all of this turns to better performance for our to-be-acquired cards...

For me, however, it's really no difference whether they launch GT300 late this year on TSMC's 40nm process or H1 2010 with GlobalFoundries, because I'm determined to buy the GT300 for my next system (and SLI probably, as I'm explicitly saving money for that), so I think either way we will get more benefits from a transition

(Although I personally would like to buy it this xmas) ^_^

[Heinz68]

I don't think is going to happen.

I think NVIDIA would worry if ATI might not get some advance access to information about the new cards.

I think all this serves to put some pressure on TSMC.

Lastly, TSMC also announced the 28nm process and claim to have "very good yield".
http://translate.google.com/translate?u=http%3A%2F%2Fglobal.hkepc.com%2F3303&sl=zh-CN&tl=en&hl=EN&ie=UTF-8

[LydianKnight]

I did read about that, they're supposed to be focusing on full-node 28nm process, skipping 32nm, seems like an odd move, although not very strange if we take into account GlobalFoundries are skipping 40nm half-node, still...

Sounds to me like something in the style of 'yeah yeah don't worry, I'll have this and this and this for everyone', although if they really deliver what they're promising, it could still be nice, early 2010 production for 28nm full-node

But anyway, seems like a bit strange... they're focusing on an advanced process and they're not paying enough attention to the current desired process (40nm)? Their engineers should focus on what their client needs plus their own future investments but sounds a bit fishy to me...

Either way, we'll see sooner than later, I guess :P

[josiahsuarez]

I don't think putting pressure on engineers would work either, but threatening TSMC could have other benefits. like better pricing, or giving higher priority to Nvidias orders. that would be particularly relevant on a new process where capacity is limited.

[LydianKnight]

Well... you do have a point there

If it's a threating move rather than a migration, yes, I think it's perfectly reasonable to think about added benefits for NVIDIA in the medium-to-long run, but that's taking into account TSMC fulfill their promises on advanced processes, for example, that announcement about early 2010 availability of 28nm full-node process, if that comes true, it could be a big win for NVIDIA if they're given priority in regards of the current situation, if not... story could be a completely different story...

[Bman212121]

I don't think putting pressure on engineers would work either, but threatening TSMC could have other benefits. like better pricing, or giving higher priority to Nvidias orders. that would be particularly relevant on a new process where capacity is limited.

I think you're probably on the right track. At work we did the same thing with our computer supplier a few years ago. The price difference wasn't that much but we wanted to check out our other options. We went back to our original supplier but I'm sure losing our business for a year had an impact on how they worked with us.

One example that we had was we would get a price from them, and let's say we found another company that would give us a lower price, they would beat that price by 10%. That sounds good in one way but obviously they weren't giving us the best deal they can. We basically told them we aren't playing that game. You give us one price and then we'll decide if we want to buy from you or not.


Well... you do have a point there

If it's a threating move rather than a migration, yes, I think it's perfectly reasonable to think about added benefits for NVIDIA in the medium-to-long run, but that's taking into account TSMC fulfill their promises on advanced processes, for example, that announcement about early 2010 availability of 28nm full-node process, if that comes true, it could be a big win for NVIDIA if they're given priority in regards of the current situation, if not... story could be a completely different story...

Well I don't know if they want to up and leave TSMC, but explore their options. If GF has a better process node than TSMC they might decide to make a switch, but they need to try them out first and see if they do have the better company to make their product. I'm sure GF is probably catering to Nvidia pretty well right now since they just split from ATI and can make chips for other companies.

Actually another little thing I just thought of is it could almost be like a chess match with ATI. If Nvidia is working hard with GF for a low end part that might come out in the next year or two, there might be scheduling conflicts for GF to get both companies orders made. This could possibly cause a slight delay in ATI's flagship launch. Obviously that's a lot of speculation, but if I can think it up the thought might have crossed someone elses mind. Now is as good of a time as any to disrupt ATI's flow.

[Heinz68]

Well I don't know if they want to up and leave TSMC, but explore their options. If GF has a better process node than TSMC they might decide to make a switch, but they need to try them out first and see if they do have the better company to make their product. I'm sure GF is probably catering to Nvidia pretty well right now since they just split from ATI and can make chips for other companies.

Actually another little thing I just thought of is it could almost be like a chess match with ATI. If Nvidia is working hard with GF for a low end part that might come out in the next year or two, there might be scheduling conflicts for GF to get both companies orders made. This could possibly cause a slight delay in ATI's flagship launch. Obviously that's a lot of speculation, but if I can think it up the thought might have crossed someone elses mind. Now is as good of a time as any to disrupt ATI's flow.
GlobalFoundry didn't split from ATI. ATI is part of AMD. Global Foundry is joint venture by AMD and ATIC of Abu Dhabi. (http://www.globalfoundries.com/about_us/history)
GLOBALFOUNDRIES (http://www.globalfoundries.com/news/globalfoundries_opens), a new leading-edge semiconductor manufacturing company formed by a joint venture between AMD [NYSE: AMD] and the Advanced Technology Investment Company (ATIC), today announced its official launch and outlined plans to drive profound change and expand opportunities in the semiconductor industry. GLOBALFOUNDRIES is led by an experienced semiconductor management team, including CEO Doug Grose, formerly senior vice president of manufacturing operations at AMD, and Chairman of the Board Hector Ruiz, formerly executive chairman and chairman of the board at AMD

While AMD has only 34.2% shares in the GF but AMD has 50% voting rights.

Now based on all the above why would you speculate that NVIDIA would get any preferential treatment from GF..?

Or why would you think "Now is as good of a time as any to disrupt ATI's flow"?
When you say it like that it almost sounds as a wish but probably I''m reading it wrong.

[noko]

I think in the end it is very smart for AMD and Global Foundaries to support Nvidia as much as possible. Kinda funny thinking that Nvidia may use GF before ATI and I hope if they do they have great success with it. Nvidia has helped AMD tremendously in the past with chipsets supporting AMD processors starting I say with NF2 while NF1 was good as well. Anyways the competition between AMD ATI/Nvidia is great and really pushing the technology by leaps and bounds with now some unbelievably cheap cards for the performance one gets. No fanboy of either but really like both companies alot.