Rumors indicate that NVidia might have pokered too high and now might have to use the Samsung instead of the 7 nm TSMC process for their high end GPUs and thus be one node behind AMD Big Navi with Ampere:
Not sure how much attention these rumors deserve. But if Nvidia has to go the Intel route and brute force performance with toasty power consumption levels it might become interesting to know whether an AMD RDNA 2 wouldn’t be another option to consider.
Question is: What would we miss particularly for VR with 8k(X) headsets?
As far as understood, AMD supports foveated rendering for ages. Question is: Does Pimax support their implementation of DFR?
DLSS 3.0 will probably be Nvidia exclusive? Or does AMD have something similar in the pipeline?
Does anybody use Pimax HMDs with AMD cards atm.? Does this work sufficiently well?
AMD has always been behind when it comes to support for “vr specific” features. For instance Asynchronous Space Warp on Oculus headsets had issues with amd cards and for a long time Valve’s ASW equivalent solution(can’t remember the name) straight up didnt work on AMD.
I think some caution should be had with going AMD for the highest level and newest features. Not to say their cards are bad or the new ones wont be monsters but I’d have to see amd be a little more involved in vr driver support before I’d consider.
Yes, interesting question. Also Raytracing might become interesting, no AMD numbers known there I think.
Perhaps the usage of RDNA 2 cards in the PS 5 and PSVR(2) will help regarding VR features trickling down to the PC. But no clue, my last AMD GPU was a Radeon 8500
Does anybody have hands on experience with Pimax HMDs and recent AMD cards?
Hm, difficult. Perhaps it might be best to keep the 1080 TI and wait for Q1(?) 2021 with 7 nm Ampere and 5 nm Big Navi? Would have no immediate use for the eye tracking module then though… Will see how well the 8kX works out of the box with the 1080 TI.
AFAIK Pimax DFR depends on a gpu features which is named NVIDIA Adaptive Shading (NAS) for nVidia and Variable Rate Shading (VRS) for AMD. The thing is, until the next gen of gpu, none of AMD’s ones have this feature. So there can be no DFR in its current implementation on AMD hardware until late this year, or maybe next year.
Even then it would require Pimax to do the necessary plumbing in Pitool. So, for now, no DFR for AMD and definitely not for old and current hardware.
Yeah, it is my understanding that RDNA 2 will be up to speed with 2xxx nvidia and 3xxx nvidia cards in terms of extra features. I’m sure it will not be a day 1 implementation especially with Pimax, but I do expect DLSS,VRS, and ray tracing to ‘just work’ on all cards in a year (no brand specific game settings, maybe wishful thinking but I can’t imagine we expect console devs to implement these features 3 separate times)
Thanks for that source. I had sth similar on my mind but since they “re introduce” VRS with RDNA2 it might not have made it in final product back then. Just like primitive shader didn’t happen on Vega at the end.
However, all VR hype started with Oculus and Vive on DX11 again. There was naturally all support for Nvidia cards because they were so much better in the old API.
Polaris/Vega n RDNA really need DX12 or Vulkan to shine and to make better use of async compute.
It was just that for AMD early developments got adapted later than for Nvidia. But for Oculus I know it was just a few months or weeks at times.
Liquid VR of AMD for DX11 is about Latest Data Latch, e.g. some QoS in the render pipeline for late user updates and less input/tracker latency and also some Dual GPU rendering techniques that were used for SeriousSamVR.
But basically this is all old stuff and hopefully VR developers use rather DX12 n Vulkan today for both Nvidia and AMD GPUs.
“There’s still no explicit word (yet) on how ray-tracing will work on next-gen Navi, but it will be a hardware solution, with support for DXR 1.1. Furthermore, RDNA 2 will join NVIDIA’ Turing and Intel’s Gen11 GPUs in supporting variable-rate shading (VRS)”.
My concern with buying ‘Big Navi’ for VR is Pimax/7invensun might have a gentleman’s agreement with Nvidia not to implement AMD version of VRS with their hardware/software either ever or on a delayed timescale, in order to maximise Ampere sales.
Actually, the 5xxx series is pretty much identical performance-per-watt wise to their nVidia counterparts, although one can buy a 390 and, indeed, enjoy toasty winters.
The question shouldn’t be, if AMD supports important VR features, but if HMD developers like Pimax and game developers support AMD. Pimax seems to be behind in that regard.
That’s unfortunate and not very forward-looking, because RDNA2 will be highly competitive, as we now know for half a year or so, since AMD spilled the beans. It’s also expected that RDNA2 will deliver higher performance per watt than Nvidia’s next gen Ampere GPUs, which is an important metric for total performance as well.
Nvidia will most likely have to compete with some tricks, like upsampling their images and comparing it to native rendered images on AMD GPUs, to remain competitive in most price ranges.
One thing, where we can expect Nvidia to remain on top, is ray-tracing.
by the numbers of GPUs not shipping with Virtual Link port I doubt VR is relevant for any marketing decision in boosting GPU sales.
I think it’s more like a me too checkbox atm for marketing ppl.
BTW. For AMD it is known they also support various types of BVH tree approaches and that the raytracing acceleration is done inline on a per shader unit basis.
Czerny basically explains you should use async compute combining tasks that are less bw heavy while doin intersection calculations on the PS5.
At the end it all comes down to AMD GPUs having quite the capabilities but AMD isn’t pushing alot efforts for VR specific implementations. It rather is betting on PlayStation VR developers to pave the way I assume.
Save yourself a headache and stick with nVidia. This goes for vr and non vr use cases.
nVidia is the most compatible GPU across the board. Any company programmmg bleeding edge features will target nVidia first and only patch amd support when people start complaining.
E.G. im in to emulation with Yuzu and Cemu and everytime there is some amazing new programming trick to increase performance or fix issues it almost always works on Nvidia first, and AMD support is TBD.
If you dont like being in for constant disappointmemt when you find out that for some random reason Pimax feature X has provlems with amd, just stick with nVidia no amount of mild performance increase, is worth the headach imo
I feel the same about intel. sure Im tempted to try Ryzen, because there are less compatibility issues on this front than with GPUs but they still exist.
Vulkan is derived from and built upon components of AMD’s Mantle API, which was donated by AMD to Khronos with the intent of giving Khronos a foundation on which to begin developing a low-level API that they could standardize across the industry.
Yeah, Nvidia’s lock-in strategy is unfortunately working well. E.g. using some machine learning libraries that rely on CUDA. AMD versions exist but are still more experimental and sometimes a little behind. As we have CUDA capable GPU machines at work and in the cloud this wouldn’t be a main factor for my personal machine. Having the option to run some training on the local machine here and there is handy though…
I just hope Nvidia doesn’t go crazy with power dissipation for Ampere - in which case I would most likely skip it. Not sure whether I’d buy an AMD card then or keep the 1080 TI. Let’s see how much faster it will be and how the 1080 TI copes with the 8kX…
Regarding Ryzen: Have a 1700 (as upgrade from a core i5 2500k) since several years, absolutely no regrets For people using the computer also for non-game applications that can utilize many cores, Ryzen is currently hard to beat, price and watt per performance is really good (particularly for the new 3xxx generation). Didn’t run into any compatibility issues yet.
For people using the computer also for non-game applications that can utilize many cores, Ryzen is currently hard to beat, price and watt per performance is really good (particularly for the new 3xxx generation). Didn’t run into any compatibility issues yet.