1080 is not 100% useless in DX12 async they just get less gain than AMD, but are still faster. In timspy DX12 with async compute the 1080 is about 40% faster than a fury X already.
Last edited:
-
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 Potential Roadmap Update for 2017: Volta Architecture Could Be Landing As Early As 2H 2017
- Thread starter Gregster
- Start date
1080 is not 100% useless in DX12 async they just get less gain than AMD, but are still faster. In timspy DX12 with async compute the 1080 is about 40% faster than a fury X already.
LOL at all the async garbage being posted!
Anyway, as for Volta it might well appear in 2017. The 10nm process is looking to be the same as he 20nm and thus Nvidia and AMD may skip 10nm and release a 2nd generation on 16nm, just like they did on 28nm. In which case we will see a slightly cut-down Volta next year
Nvidia have now announced the full line up pascal GPUs: GP100,102,104 and 106. A few smaller and mobile part to come soon. Next year would make perfect sense to release the next generation.
Anyway, as for Volta it might well appear in 2017. The 10nm process is looking to be the same as he 20nm and thus Nvidia and AMD may skip 10nm and release a 2nd generation on 16nm, just like they did on 28nm. In which case we will see a slightly cut-down Volta next year
Nvidia have now announced the full line up pascal GPUs: GP100,102,104 and 106. A few smaller and mobile part to come soon. Next year would make perfect sense to release the next generation.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,073
What is the async garbage?
They'll use it for small mobile designated chips, much better yields plus Apple pays good money for big volume.
I wouldn't be at all surprised if Volta is pushed forward. Pascal was just a node-shrink of Maxwell with 'emergency hardware patches' (<100us context switch / dynamic load balancing) so that they can get any sort of improvement in DX12. But it's still quite restricted compared to GCN. Volta surely will address this shortcoming once and for all.
Sentiment can change very quickly. A few more DX12/Vulkan games and Vega out in 1H 2017 and I can already see headlines with ecstatic AMD fanboys claiming NVidia is exiting gaming, and upset NVidia fanboys whining about how NVidia failed them and all that.
As others pointed out, NVidia are very good at making money, so it makes absolute sense they'd list Volta as 2018 in order to push those Pascal sales. But surely NVidia have the capacity to make this happen (push Volta forward) and they really should do it if you ask me.
It'll all depend on the progress of DX12/Vulkan adoption. So far it seems to be coming sooner than NVidia expected so we may see them 'adjusting accordingly' depending on how things go.
Sentiment can change very quickly. A few more DX12/Vulkan games and Vega out in 1H 2017 and I can already see headlines with ecstatic AMD fanboys claiming NVidia is exiting gaming, and upset NVidia fanboys whining about how NVidia failed them and all that.
As others pointed out, NVidia are very good at making money, so it makes absolute sense they'd list Volta as 2018 in order to push those Pascal sales. But surely NVidia have the capacity to make this happen (push Volta forward) and they really should do it if you ask me.
It'll all depend on the progress of DX12/Vulkan adoption. So far it seems to be coming sooner than NVidia expected so we may see them 'adjusting accordingly' depending on how things go.
It will be interesting to see what architectural changes come with Volta. I agree that Pascal is more akin to Maxwell and not really a change up in architecture so to speak and will they go big with ACE? I think it is a given this time.
Volta should be very interesting. They won't necessarily mimic AMD and they may very well come up with something even better. It's not like NVidia don't have excellent engineers.
Yer, I can't see them missing the boat on what seems to give some good performance gains.
Agreed. I'd like for this rumour to be true.
Always nice to see a speed up of progress after our long time with 28nm. Also I'm a big fan of VR so I'd like to see graphics performance advance as fast as possible, in order to support much better VR resolutions.
Now where's the AMD Vega news. The big one with HBM2 please!
Last edited:
Agreed. I'd like for this rumour to be true.
Always nice to see a speed up of progress after our long time with 28nm. Also I'm a big fan of VR so I'd like to see graphics performance advance as fast as possible, in order to support much better VR resolutions.
Now where's the AMD Vega news. The big one with HBM2 please!
I think 4k samsung phone screens will be out by next year so could have 4k VR headsets and Volta. Sounds good to me.
oh as long as it does not cost £1000 for a 1180 by then. 
I think 4k samsung phone screens will be out by next year so could have 4k VR headsets and Volta. Sounds good to me.
I wish they would get the 4K VR headsets out. Whilst I really enjoy my Rift, I can't help but think what the future will bring and I want that now.
Being able to drive 4K at the hz needed for VR is unlikely with the slowdown in progress now. Pascal/Polaris were supposed to be twice as fast as they ended up being and NV have no reason to put out any major uplifts now that they know AMD are pretty much finished.
I think 4k samsung phone screens will be out by next year so could have 4k VR headsets and Volta. Sounds good to me.
What really matters for VR screens is the PPI they can be manufactured at, and the ability to run at 90+ Hz and low persistence.
Both headsets use a custom size/shape screen, so just having 4K phone screens available isn't enough in of itself.
The current headsets are 463 PPI. So assuming Samsung made a 4K screen the same size as the Galaxy S7 edge, that would be 801 PPI.
Assuming that was the best they could do, that would then mean they could manage a maximum of 3736x2076 (1868x2076 per eye) at the same FOV as the current headsets.
They could also increase the FOV and do something like 4484x2491. That would be 120 degrees vs the current 100. However the resolution (as in fine detail) would remain the same, ~3x the current.
No?
They're pretty much spot on where they're meant to be. Remember 16nm/14nm is 1 node from 28nm, not 2 (because it's actually 20nm finfet). So you should only expect 2x the performance per mm2. And that's exactly what Pascal and Polaris deliver.
Additionally a GTX 1070 (or 980 Ti) could run a 4K VR headset right now. The headsets already use 1.4x supersampling as standard (anything you do over '1.0x' is actually over 1.4x). So any card that can run 1.5x supersampling in SteamVR is actually rendering at 4536x2520. And, barring something very fancy with everything turned up to full (like Elite dangerous), both those cards can run 1.5x supersampling today. AND that's without Nvidia's simultaneous reprojection, OR full foveated rendering.
The ability to jump to ~5K VR screens should actually be very fast. It'll be going on from there that'll be the tricky part.
Last edited:
You've explained it better than I could have.
The new titan X could do 4K VR very well and Volta even better. As has been mentioned a 4k VR headset will only be using what 70 percent of the display area and only the middle 50 degree's of each eye will need high quality rendering then a falloff to the periphery.
I'm wondering if the drive for higher VR resolutions is whats driving progress nvidia at the moment or maybe the future as they know people will buy the cards so what AMD is doing maybe irrelevant to some extent?
Yeah I can't wait too. 4K VR headsets are really needed. The resolution at the moment although decent is nowhere near what it needs to be.
Indeed.
If they combine full foveated rendering with simultaneous multiprojection, I wouldn't be surprised if something as low as the 4GB RX 480 could run a 4K VR headset.
4K only requires ~61% more processing power than the current VR resolution (since the actual render resolution is 3024x1680). And that would be if you used no software tricks to make it easier at all.
I'm hoping they go for 3x the current PPI/resolution, along with a 30% increase in FOV. The way screen manufacturing is looking, they should be able to pull that off relatively easily within ~12 months.
It would result in a resolution of 4858x2699 (or thereabouts). But would only require a screen PPI of 801, the screens would be physically larger to make the FOV bigger.
And even though that resolution may look too big to render, it is actually only ~2.6x the number of pixels already being rendered with VR headsets (so comparable jump as going from 1920x1080 to 3440x1440), and a GTX 970 can do the current tech. With a couple of software tricks (like foveated rendering) it should definitely be doable on a GTX 1070, and if they can do a lot of optimisation, even an RX480. Also since they've already got the dynamic supersampling in place they could subsample to make it easier to render.
The important thing is to get more PPI and more FOV into the headsets. You wouldn't actually have to render at the full screen resolution to make it gorgeous. You just need it to have very high clarity (can't see the physical pixels on the screens) and have enough FOV to be hugely immersive (they're already 95% of the way there anyway, so a 20-30% larger FOV would be more than enough).
Last edited:
Which is why, despsite the current VR sets been impressive, I am waiting for gen 2
Yeah ALLBODIES the more PPI the better but unfortunately I think phone screens are the only driving factor for samsung. Hence I think they'll stop at a 4K screen which is the same physical size as say an S7. I can't see anyone going above that anytime soon unless the VR market gets much bigger and such screens can be specifically made.
You are right we just need the PPI. The screen door is a huge issue..
Having a 4k screen and rendering a 1440p type image will still look much clearer as the screen door wont be as pronounced.
I'm waiting for Gen2 aswell Greebo. Got a feeling it'll they'll be released next year contrary to what everyone else is saying.
As soon as a 4k screen is mass produced Oculus and HTC will have it in a VR headset very quickly. As both will want to be first to 4k.
Tracking has been largely sorted as have controllers. 4k is the next thing they'll be working on getting.
You are right we just need the PPI. The screen door is a huge issue..
Having a 4k screen and rendering a 1440p type image will still look much clearer as the screen door wont be as pronounced.
I'm waiting for Gen2 aswell Greebo. Got a feeling it'll they'll be released next year contrary to what everyone else is saying.
As soon as a 4k screen is mass produced Oculus and HTC will have it in a VR headset very quickly. As both will want to be first to 4k.
Tracking has been largely sorted as have controllers. 4k is the next thing they'll be working on getting.
Last edited:
