• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA Volta with GDDR6 in early 2018?

Associate
Joined
26 Mar 2016
Posts
150
Having said that apart from mixed precision,do we know what a gaming Volta GPU will have over Pascal yet??

New Cache subsystem as i pointed out above, which eases developing on volta. Hopefully all DX12.1 Features.

On Timeframe of Volta, i heard from a guy who has some contacts to nv, that 1080Ti is already EOL for partners. Partners can't order anymore GPUs of this type. The last batches are in production and will be delivered till end of this year to them. Of course take that with a truckload of salt, but it would definately indicate a launch in Q1 of GV104.
 
Soldato
Joined
9 Nov 2009
Posts
24,838
Location
Planet Earth
New Cache subsystem as i pointed out above, which eases developing on volta. Hopefully all DX12.1 Features.

On Timeframe of Volta, i heard from a guy who has some contacts to nv, that 1080Ti is already EOL for partners. Partners can't order anymore GPUs of this type. The last batches are in production and will be delivered till end of this year to them. Of course take that with a truckload of salt, but it would definately indicate a launch in Q1 of GV104.

Just realised March is Q1(not Q2),so that would kind of fit in with historical precedent too,and would give time for Pascal stock to be run down too.

What I do hope is that we start to see proper penetration of Vulkan and DX12 now,once Pascal drops,as both AMD and Nvidia will now have more DX12/Vulkan orientated GPUs.

With Intel also now joining MOAR cores too,hopefully we will start to see games start to scale better with cores over the next few years,and hopefully that means we can have some more leaps on the technical side too.
 
Man of Honour
Joined
13 Oct 2006
Posts
91,096
Yeh but like 1 person only had the card then, the other 99% of us had to wait until sept for pre ordered cards, that launch went down like alead balloon.

1080? only a couple of models where they've vastly over-estimated how few cards would hit the clocks that model was advertised at were delayed that long. Contrary to the perception a lot of people were getting cards granted production was a bit pedestrian but demand was very high.
 
Soldato
Joined
18 May 2010
Posts
22,376
Location
London
If Volta was already done and dusted for gaming cards I can't see Nvidia just leaving it on the shelf! :p

They would simply tail off production of Pascal cards to run down inventory and start building up Volta inventory. Having said that,they might not want to do that during Christmas as that would be a good period for sales,surely??

TBH,if you look at the cadence from GTX680 to GTX980 and GTX980 to GTX1080,it was around 28 months and 20 months respectively.

The GTX1080 was launched in May 2016,so a 20 month cadence would give you January 2018,and a 28 month cadence,would give you September 2018.

If you look at Nvidia card launches - they seem to generally either following a Q2 or Q3 launch period - so either March~ May or roughly around September(looking at the GTX480,GTX580,GTX680,GTX780,GTX980 and GTX1080 launches).

Having said that apart from mixed precision,do we know what a gaming Volta GPU will have over Pascal yet??

MOAR SPEED! :cool:

Other than that nope.
 
Last edited:
Caporegime
Joined
18 Oct 2002
Posts
32,618
GV100 has a completely knew scheduler that is a lot more efficient and flexible. Not clear if gaming Volta will see this but I expect so. this could improve efficiency and multi-engine synchronization performance. Otherwise it is not clear hwat Volta will bring because Nvidia hasn't talked about Gerforce parts. There is some interesting things, e.g. Gv100 apparently doesn't have double Fp-16 support unlike GP100. FP16 main use today is Deep learning, and Nvidia instead made dedicated tensor cores with much higher performance. This differentiation would allow Nvidia to add FP16 support to consumer hardware for example without affecting sales of GV100. Although it is not clear whther the benefits are worth the silicon cost, e.g. you could just put in more Fp3 cores instead.
 
Back
Top Bottom