NVIDIA Accelerates Volta to May 2017?

[btarunr]

Following the surprise TITAN X Pascal launch slated for 2nd August, it looks like NVIDIA product development cycle is running on steroids, with reports emerging of the company accelerating its next-generation "Volta" architecture debut to May 2017, along the sidelines of next year's GTC. The architecture was originally scheduled to make its debut in 2018.

Much like "Pascal," the "Volta" architecture could first debut with HPC products, before moving on to the consumer graphics segment. NVIDIA could also retain the 16 nm FinFET+ process at TSMC for Volta. Stacked on-package memory such as HBM2 could be more readily available by 2017, and could hit sizable volumes towards the end of the year, making it ripe for implementation in high-volume consumer products.



View at TechPowerUp Main Site

 

[techy1]

ok... great news... but why? what is financial reasons for this... now they have midrange cards (1070, 1080) for incredible high prices and sold out as soon as those hit the store.... it would be financial reasonable to keep it that way for as long as possible.. unless they know :O

 

[Frick]

ok... great news... but why? what is financial reasons for this... now they have midrange cards (1070, 1080) for incredible high prices and sold out as soon as those hit the store.... it would be financial reasonable to keep it that way for as long as possible.. unless they know :O

Punching AMD in the face I guess.

 

[ViperXTR]

HBM2 GTX 1180/1170 and "async compute?"
Welp, gotta sell this 1070 soon '__'

 

[P4-630]

I'm sure those HBM2 cards come with a premium price.

 

[qubit]

Awesome performance at a reassuringly expensive price.

 

[InVasMani]

Punching AMD in the face I guess.

Seems like a good enough reason. The other reason being it's better than remaining to stagnant and behind the 8-ball just look how well that's worked out for AMD.

 

[Prima.Vera]

ok... great news... but why?

To put AMD out of their misery for good?

 

[ensabrenoir]

HBM2 GTX 1180/1170 and "async compute?"
Welp, gotta sell this 1070 soon '__'

.......this

....this generation from both companies were just stop gaps......evil geniuses...

....going through a upgrade now on the "everyday machine"......so i'm still gonna get a pascal. My main "work" rig will be update when socket 2066 drops so volta will probably coincide...

 

[bug]

ok... great news... but why? what is financial reasons for this... now they have midrange cards (1070, 1080) for incredible high prices and sold out as soon as those hit the store.... it would be financial reasonable to keep it that way for as long as possible.. unless they know :O

This may be related to Nvidia's initial plans. The Pascal we have today was not meant to be, we were supposed to be getting Volta in 2016. See: http://www.theregister.co.uk/2013/03/19/nvidia_gpu_roadmap_computing_update/
With HBM production not moving forward as initially planned, Nvidia added Pascal to the roadmap. So Volta may be already done, just waiting for HBM availability for all we know. I'm speculating, of course, I have no inside information, but I think this could be an explanation.

 

[-The_Mask-]

ok... great news... but why?

Because in some aspects architecture wise Pascal's GP102 and GP104 is just a generation behind Polaris and Vega, even if the performance and performance/watt is better by nVidia. Volta will have HBM2 just like Vega and Volta will probably have better DX12 support like Polaris and Vega.

 

[btarunr]

Because in some aspects architecture wise Pascal's GP102 and GP104 is just a generation behind Polaris and Vega, even if the performance and performance/watt is better by nVidia. Volta will have HBM2 just like Vega and Volta will probably have better DX12 support like Polaris and Vega.

Pascal supports feature-level 12_1 but Polaris only supports 12_0. Maybe NVIDIA driver is exposing features hardware doesn't support.

 

[jabbadap]

ok... great news... but why? what is financial reasons for this... now they have midrange cards (1070, 1080) for incredible high prices and sold out as soon as those hit the store.... it would be financial reasonable to keep it that way for as long as possible.. unless they know :O

Summit, they have contract to fill. Don't expect to see consumer models though.

 

[P4-630]

True 4K with just a single card.

 

[-The_Mask-]

Pascal supports feature-level 12_1 but Polaris only supports 12_0. Maybe NVIDIA driver is exposing features hardware doesn't support.

Look at the individual support of the features and not the feature level.

 

[HumanSmoke]

Summit, they have contract to fill. Don't expect to see consumer models though.

Precisely. Once Nvidia decided to not wait around for 10nm and commit Volta to 16nmFFC production there probably wasn't any good reason not to go ahead with the HPC orders since it ties in with IBM's POWER9 schedule.

 

[ensabrenoir]

True 4K with just a single card.

Believe thats coming real soon....and with the rift and company gaining traction......and getting cheaper......the PC market should be in for a resurgence....

 

[ViperXTR]

Precisely. Once Nvidia decided to not wait around for 10nm and commit Volta to 16nmFFC production there probably wasn't any good reason not to go ahead with the HPC orders since it ties in with IBM's POWER9 schedule.

Oh i guess this this one then?

 

[HD64G]

ok... great news... but why? what is financial reasons for this... now they have midrange cards (1070, 1080) for incredible high prices and sold out as soon as those hit the store.... it would be financial reasonable to keep it that way for as long as possible.. unless they know :O

Just because DX12 and Vulcan will get adopted by many game devs sooner than they hoped (not) and Pascal will have a hard time vs Vega GPU. Another reason is becaus 10nm won't be ready as soon as they needed to.

 

[Parn]

I guess nvidia realises the threat from AMD in DX12/Vulkan titles, so they are speeding up the development of their own async capable architecture. AMD Vega will have a hard time if Volta's async performance is on par with the top end GCN cards.

 

[RejZoR]

Pascal supports feature-level 12_1 but Polaris only supports 12_0. Maybe NVIDIA driver is exposing features hardware doesn't support.

These feature levels are useless. NVIDIA supposedly supports higher levels and yet it still has no functional async. Stupid.

 

[bug]

These feature levels are useless. NVIDIA supposedly supports higher levels and yet it still has no functional async. Stupid.

Will you stop it already? Nvidia's async is fully capable: http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/9

 

[papupepo]

Source: WCCFTech

 

[basco]

quote from anandtech link from bug:
but on the whole I don’t expect NVIDIA to benefit from async by as much as we’ve seen AMD benefit. At least not with well-written code.

the whole article is very interesting :
Pascal has been optimized, it’s a slightly wider but mostly higher clocked successor to Maxwell 2
so for me the 1080 is not very spectacular vs 980\+ti

but i think we are going to see more power efficiency then moarrr power in future from any company
because i dont give a shit how much my hardware consumes then what i get fps in return

 

[RejZoR]

Will you stop it already? Nvidia's async is fully capable: http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/9

My GTX 980 ain't. And it's D3D12_1. Supposedly higher than 12_0 on Radeons. Also, that thing in GTX 1000 series is highly questionable. I don't believe any of the shit NVIDIA says. They've also been promising a "driver update" that would unlock async capability on Maxwell 2 which turned out to be complete bullshit. What makes you think I'll believe GTX 1080 has it? If I'd make a "best of the best" graphic card which actually now has async, I'd brag about it on all ends. And NVIDIA didn't even mention it anywhere, some people started talking about it just now, like 2 month after release of the Pascal series.