NVIDIA GeForce 20 Series (Volta) to be released later this year - GV100 announced

[Ajay]

Planck, Schrodinger, Heisenberg. Mach would at sound cooler though

 

[Head1985]

Two things wrong with Fudzilla here. One is GV104 in HBM2 and GDDRX. Nonsense, as you point out.

The other thing is GV104 with Titan versions. That's what GV102 will be for.

But wait.

What if, 4 flagships were not enough? What if they need more? Envision if you will:

GTX 2080 flagship (cut GV104) followed by
Titan XV flagship (full GV104) followed by
Titan Xv flagship (cut GV102) followed by
GTX 2080 Ti (cut GV102 but faster) followed by
Titan XXX flagship (full GV102)

Don't say they wouldn't do it. =P

Or full GV106 as new 1500USD TITAN.Bet people will buy it if its 5% faster than full GP102 with half TDP.
Btw i really dont care what nv doing.Since they have monopoly its hilarious what they are doing.

 

[alcoholbob]

Two things wrong with Fudzilla here. One is GV104 in HBM2 and GDDRX. Nonsense, as you point out.

The other thing is GV104 with Titan versions. That's what GV102 will be for.

But wait.

What if, 4 flagships were not enough? What if they need more? Envision if you will:

GTX 2080 flagship (cut GV104) followed by
Titan XV flagship (full GV104) followed by
Titan Xv flagship (cut GV102) followed by
GTX 2080 Ti (cut GV102 but faster) followed by
Titan XXX flagship (full GV102)

Don't say they wouldn't do it. =P

But it's not that likely, mainly due to costs.

I think most of the people who buy Titans are typically looking for a cut down Tesla card for productivity work. So they certainly could sell a full GV104 as a Titan, but it would need more RAM than the current Titan to be a sell, so it would have to be 16GB GDDR6X if sticking to 256 bit memory, and going with the Titan tradition would need faster RAM chips binned.

To be honest I don't think that's likely because it sounds really expensive. A GPU die honestly isn't that expensive (maybe $35 for a 600mm2 core) for Nvidia, a large amount (> 12GB) of GDDR6X or HBM2/3 costs way more than the GPU core, which is a small portion of the manufacturing cost of a video card. It would actually cost Nvidia way more to sell puffed up GV104 as a Titan to the public rather than just waiting a few months for memory and die yields to improve and sell a cut down big die with a 384 bit bus and sell it with 12GB of RAM later instead of 16GB of RAM right when it launches.

 

[lobz]

Or full GV106 as new 1500USD TITAN.Bet people will buy it if its 5% faster than full GP102 with half TDP.
Btw i really dont care what nv doing.Since they have monopoly its hilarious what they are doing.

yes, with a 106 being faster than a 102

 

[Jackie60]

I'm suspecting that Vega may be a monster what with Nvidia releasing the Titan Xxp and bringing forward Volta.
Clearly they want to milk the high end for all it's worth while they can and bring in something very soon to continue the milking process
if AMD retake the performance crown. I do love this potential renewed competition in the GPU space. Here's hoping we see at least another 20-30%
performance on top of Titan Xpp before the end of the year!

 

[Sweepr]

Edit: never mind.

 

[lobz]

Edit: never mind.

Way to lure me in here....

 

[xpea]

Volta announced in few hours, it's nearly official:

During GTC 2017, Inspur and NVIDIA and will jointly present a cutting-edge AI supercomputer, AGX-2, with the NVLink™ 2.0 enabled. The AGX-2 is designed to provide maximum throughput for superior application performance for science and engineering computing - taking AI computing to the next level.

Note: NVLINK 2 is only present in Volta (Pascal is only NVLink 1)
Source: http://www.prnewswire.com/news-rele...uter-and-aistation-at-gtc-2017-300452957.html

and Nvidia is already stock pilling Volta, announced by JHH him self yesterday during Q1 earning QA session:

C.J. Muse - Evercore Group LLC
Very helpful. I guess as my follow-up, on the inventory side, that grew I think 3% sequentially. Can you walk through the moving parts there? What's driving that, and is foundry diversification part of that? Thank you.

Jen-Hsun Huang - NVIDIA Corp.
The driving reasons for inventory growth is new products, and that's probably all I ought to say for now. I would come to GTC. Come to the keynote tomorrow. I think it will be fun.

 

[Krteq]

How they can stock pilling Volta cards without GDDR6 chips soldered to PCB?

 

[tamz_msc]

Bleh, this is going to be the formal unveiling of their ORNL collaboration.

 

[krumme]

Volta announced in few hours, it's nearly official:

Note: NVLINK 2 is only present in Volta (Pascal is only NVLink 1)
Source: http://www.prnewswire.com/news-rele...uter-and-aistation-at-gtc-2017-300452957.html

and Nvidia is already stock pilling Volta, announced by JHH him self yesterday during Q1 earning QA session:

That is great news. We need new tech from nv. Compettition.
Its always great when stuff like fermi and gcn1 arives as its the foundation for years. No more facelifts. No more old dx11 stuff from nv.

 

[xpea]

How they can stock pilling Volta cards without GDDR6 chips soldered to PCB?

First Volta is for HPC and with HBM2
I have no idea about the first consumer Volta but last rumors say September

 

[Krteq]

OK, makes sense

 

[raghu78]

https://www.nextplatform.com/2016/11/20/details-emerge-summit-power-tesla-ai-supercomputer/

I think we will see an unveil of Volta GV100 with HBM2 which will go into IBM Power 9 / Volta supercomputers by late 2017 and also GV104 for a Q3 2017 launch.

 

[Krteq]

Hmm, shouldn't be Volta manufactured on 12nm process?

 

[tamz_msc]

Hmm, shouldn't be Volta manufactured on 12nm process?

TSMC's 12nm FFC is for low-power stuff.http://www.anandtech.com/show/11337/samsung-and-tsmc-roadmaps-12-nm-8-nm-and-6-nm-added/4

 

[Krteq]

Yep, but it's planned for 2018. That's why I'm asking.

 

[tamz_msc]

Given there is not roadmap for anything "12nm" in 2017, I assume that it will still be 16nm FF+.

 

[Head1985]

https://www.nextplatform.com/2016/11/20/details-emerge-summit-power-tesla-ai-supercomputer/
and also GV104 for a Q3 2017 launch.

No way.Pascal is not even 1 year old.Volta gv104 will launch when GTX1070/1080 will be 2 years old.

 

[Karnak]

No way.Pascal is not even 1 year old.Volta gv104 will launch when GTX1070/1080 will be 2 years old.

Maybe not 2 years, but I'm expecting Consumer Volta (GV104) to launch in Q1 2018.

 

[raghu78]

No way.Pascal is not even 1 year old.Volta gv104 will launch when GTX1070/1080 will be 2 years old.

Hynix have already confirmed with the press release of GDDR6 production for a high end graphics card in early 2018 with 768 GB/s bandwidth.

https://www.skhynix.com/eng/pr/pressReleaseView.do?seq=2086&offset=1
http://www.anandtech.com/show/11297...gddr6-memory-for-graphics-cards-in-early-2018

Micron is also confirming GDDR6 production in late 2017 or early 2018.

http://www.anandtech.com/show/11100...ce-roadmap-updates-forecasts-and-ceo-retiring

There are 2 scenarios possible.

1.) If GV102 is releasing in early 2018 , GV104 will release earlier given thats how Nvidia launch new GPU generations 104 first and then the 102 high end chip. In this case GV104 might use GDDR5X.
2. The second case is Nvidia launches Volta top down like they used to do back in the Fermi days. GDDR6 production is slated for late 2017 at Micron and Hynix. So if GV104 has to also sport GDDR6 then this option is possible.

 

[Head1985]

Nv have monopoly they dont need volta.99% gv104 will launch first and it will be close to 2years from GP104 launch.Its always like that since kepler.

 

[Arachnotronic]

How they can stock pilling Volta cards without GDDR6 chips soldered to PCB?

GV100 will use HBM2.

 

[Arachnotronic]

Volta announced in few hours, it's nearly official:

Note: NVLINK 2 is only present in Volta (Pascal is only NVLink 1)
Source: http://www.prnewswire.com/news-rele...uter-and-aistation-at-gtc-2017-300452957.html

and Nvidia is already stock pilling Volta, announced by JHH him self yesterday during Q1 earning QA session:

Yep, NVLink2 is GV100. I think we will see GV100 announced at GTC 2017, this is going to be awesome!

 

[Despoiler]

The Volta launch will look just like Pascal. It's going to lead with the very top end HPC/data center/deep learning chip. Consumer will be several months after. Nvidia is doing this because Radeon Instinct(Vega architecture) is AMD's first legit attempt at a card in that segment. AMD has already laid the tools ground work with the Boltzmann Initiative, MIOpen, and ROCm. They just needed a card to go with it. Radeon Instinct is that card. Nvidia actually has to start competing against someone other than itself. That means if they have an opportunity to launch early they will.