GlobalFoundries courts Nvidia
GLOBALFOUNDRIES PLANS to build up capacity "fairly aggressively" within the next three to four years and invest "very aggressively" in 32nm and 28nm process technology whilst reaching out to multiple potential customers, according to the firm's head of communications, Jon Carvill.
Carvill told the INQ the firm would target the top 20 fabless semiconductor manufacturers and try to entice them with "the most advanced process technology of any foundry", a wide "global reach" and Automated Precision Manufacturing (APM).
APM, which Carvill reckons "no other fabs in the world have" is a three step process offering looped feedback at every step, so one can know how well a chip is yielding at every stage. This means if corrections need to be made, manufacturers don't need to wait until the very end of the process, instead they can be made during the early phase, the middle, or the end as required.
Carvill said there has been a lot of interest from potential customers already, noting, "thus far, the response has been very positive," with potential punters apparently "very interested to learn more". How does this sit with AMD, though?
"Our priority is giving them [AMD] outstanding customer service," gushed Carvill, adding that his former employer gave GlobalFoundries "a competitive advantage" the firm could unlock for other customers in the market by lowering the barrier of entry to advanced process technology nodes.
Having a customer like AMD with "really sophisticated microprocessors" is a competitive advantage for his fab fab, explained Carvill. "By the fact that we ramp AMDs products so quickly and then by the time another customer would adopt the process it's very robust," he said, adding: "at that point it's very much battle tested and ready to go."
Boasting that GF's Dresden fab was consistently ranked as amongst the most efficient in the world and that new manufacturing capabilities would soon be available in upstate New York, Carvill told the INQ: "If you're lagging edge maybe we wouldn't be your best option, but if you're leading edge, you would absolutely want to take a look at us."
So, is Nvidia absolutely wanting to take a look at GlobalFoundries then? Carvill wouldn't say, but noted, "As long as it didn't compromise our relationship with AMD and customer service to them, we would happily have those discussions with Nvidia and anybody else."
Here at the INQ we've heard persistent rumours that Nvidia's Jen-Hsung Huang has already met with reps from GlobalFoundries to "kick the tyres" and see what's what. The move would make sense as the Green Goblin hardly has the best relationship with its current manufacturer, TSMC.
A particularly bad 40nm ramp combined with a plethora of reliability problems purportedly to do with bad DFM models from the Taiwanese manufacturer means GlobalFoundries could offer Nvidia leverage in a relationship where they desperately need some.
Also, Huang is apparently increasingly worried that ATI could now end up a generation ahead on process technology and could get better performance and economies of scale in production.
This is not unrealistic and Nvidia's current pain and concern over AMD's potential win in the next generation (DX11) GPU war, will, if it hasn't already, in all likelihood drive Nvidia's queen bee into Lost and Foundry's open arms.
Oh, and another reason Nvidia may be interested in pursuing the GF route...
"We have HUGE experience with the most sophisticated processors in the world, X86 microprocessors,", Carvill told us.
Hmmm... no wonder Intel is throwing a wobbly. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
I doubt GlobalFoundaries will pose a major threat to TSMC or UMC in the near future. The infrastructure those guys have allows them to be a lot more flexible than AMD fabs have been in the past. Though some serious competition to TSMC could be good for all..
( posting more pics of sylvie after having inq-fabois gush over her hasn't gone unnoticed! )
It makes no sense for globalfoundries to even consider working with nvidia. It's like shooting itself in a leg. Or maybe they can do 90nm stuff for nvidia ;)
AMD's fabs have been some of the most flexible fabs in the world on the count that they needed to supply a lot of different CPU's and other products from the same single factory AND be able to upgrade it to a new process while it was still producing CPU's
flexibility is one of GF fab's main trump cards.
and it makes perfect sense for GF to want Nvidia as a customer. that way they make money no matter who won that generation's GFX war.
and it makes perfect sense for nvidia as they'll have 2 suppliers so they can go to the one that comes out best while at the same time getting the best price.
ATI will probably keep its relation to TSMC aswell.
that is, as long as TSMC can keep up to GF.
TSMC use to be the only fab you could go to as a fabless company to get cutting edge production processes. but GF is set to 'one up' TSMC with their extremely aggressive schedule for new process technology.
Perfect sense for Nvidia to work with 2 foundries and pick the best. Yeah if you want to double your mask costs, your design work, your engineering needs, your qualification efforts, increase forecasting complexity and then drop one or split the business (and probably get less of a volume discount?). Nvidia will have to choose in advance (though it could be on a product by product basis) or they will get snowed under by the engineering and logistic efforts.
Also, if Nvidia is courting GF because of x86 possibilities that is only STRENGTHENING Intel's case about AMD license issues and potential transfer to GF. I would doubt GF (or AMD) would want to touch that anytime soon. AMD has said they will divest from GF over time, so even if it is still considered a technical subsidiary right now, it will certainly not be one in the future (and therefore will lose any x86 license rights).
Where in the story is it indicated that GF is COURTING Nvidia... I see plenty of Nvidia kicking the tires, and AMD saying they will talk to anyone.. how does that go to GF courting Nvidia?
We've heard this baloney from AMD before. They have a legacy of over-promising and under-delivery, which is why they are in the predicament they're in. Put a shovel in the ground, build the fab, produce one wafer, THEN start running your mouth. These guys never change.
See this is what is wrong with Nvidia - they are desperate beyond desperate for partners and ATI are gracious enough to offer them assistance via a 3rd party...but Nvidia just wont take it up.
Talks with Nvidia are just like talks with North Korea. They do a lot of chest beating to go along with their failed launches and the bare minimum useful products. They don't even realise how small time they are because they are too busy pretending to be bigger than what they are.
People who are saying that should have a look at the benchmark charts around the web, and tell me what is what.
I would have bought two 260 GTX today instead of having two 4870 but I am happy with my setup.
IBM is no slouch in the process field; they are as good as, perhaps even better than Satan Clara's process tech.
GloFo's technology alliance with IBM, coupled with GF's new APM 3.0 rev is a highly compelling story if you're a vendor looking for state of the art chippery. I would have to call Jensun completely bonkers if he stayed with TSMC any longer than contractually obligated.