http://wccftech.com/nvidia-gm200-core-titan-ii-soon/
Hmm, is CP really Consumer Sample? Could it be Customer Sample and therefore be for the HPC market? 2 months sounds too early for a consumer GM200 doesn't it?
Hmm, is CP really Consumer Sample? Could it be Customer Sample and therefore be for the HPC market? 2 months sounds too early for a consumer GM200 doesn't it?
This kind of shipping data with detailed contents is available in the open?
I find that more fascinating than the contents of the shipments.
I find that more fascinating than the contents of the shipments.
https://www.zauba.com/import-gm200-hs-code.html
It seems yes, something i find too really fascinating. Now, i dont see why Nvidia, or AMD or anyone want to write the codename of the sku on the box...
Is really Nvidia want to sold GTX880 and say to their client. ok guys, buy it, but in 2 months we launch the real thing. Im sure the strategy have largely work with Kepler, im not sure peoples will follow it the next times.
This said, you can find report of GM200 parts send in July, and at other date, with different codename.
I think that's specific to India.
Yeah. The GK110 was first launched in November 2012 as a Tesla K20 part, and later as Titan in February 2013. We will probably see that same pattern here.
If memory serves well they started serving HPC contracts somewhere in late September 12' already. If they follow a similar trend this time I'd expect the first desktop GM200 based SKU not earlier than Q2 15'.
If you want what you are doing to stay secret, better not send anything there or from there lol.
wccftech.com/nvidia-gtx-980m-gtx-970m-maxwell-mobility-queued-week-august/
How many chances that Maxwell desktop gpus are at 20nm? (1%?
)
How many chances that Maxwell desktop gpus are at 20nm? (1%?
)
I don't think I'm far off saying this year 99% of TSMC's 20nm revenue comes from Apple and Qualcomm, not NVIDIA (nor AMD)
A1xLLcqAgt0qc2RyMz0y
Veteran
Last edited by a moderator:
This is the most likely scenario.
Any work that was done on a 20nm ASIC should be easily ported to 16FinFet or 16FinFet+, according to TSMC.
Though 16FinFet+ is out of the picture until late 2015/early 2016.
I think a XT PE/Ultra Extreme is coming.
Last edited by a moderator:
A1xLLcqAgt0qc2RyMz0y
Veteran
What makes 16FF+ more difficult than 16FF that it will be available 9-12 months behind 16FF?
This post that quotes from TSMC's Quarterly Report and Q&A shows the delay between 16FF and 16FF+ to be 6 months or less:
http://forums.anandtech.com/showpost.php?p=36487476&postcount=38
Last edited by a moderator:
I keep hearing this kind of information but it's a total non-sens for TSMC as a business point of view.
First, Nvidia is an historical partner.
Second JHH has very long term and personal relationship with TSMC CEO.
Third Nvidia is officially a leading customer for 20 and 16 nodes.
Then, I don't know a single company that relies on only 2 customers for their business.
Finally, you must also consider that Nvidia is a loyal customer, sourcing 100% of their products at TSMC. On the other side, Qualcomm has more than one supplier and Apple is a volatile bird that will also use Samsung 16nm.
So why on earth TSMC will piss off Nvidia with no 20nm allocation ? It will be a big mistake that will force Nvidia to look elsewhere for next node...
The truth is that Nvidia will get 20nm wafers from TSMC. Will they get everything they have allocated ? maybe not because of Apple deal, but I'm sure Nvidia will have a reasonable pie, at least enough to offer 20nm products if they have planned to.
Last edited by a moderator:
To see an actual product...
NVIDIA is small potatoes compared to Apple and Qualcomm, whose needs naturally come first. If they need a lot of wafers, TSMC will sell them everything they want.
NVIDIA comes second.
I knew that the 20nm process was not suitable for complex chips like gpus, just for the Soc like apple's ones
