|
Post by Pyroth309 on Oct 20, 2022 18:43:34 GMT
I don't have an issue with the refresh rate. I'm not sure it it's actually Vsync on hmds though since you can be at variable fps with it. With Vsync if would be dropping down to half fps or a bigger interval.
May be adaptive with a frame rate limit or just a frame rate limit. Either way... If it's disabled it allows the bench to generate as many fps as your system will allow. IE above 80,90, or 144 if your gpu is strong enough.
|
|
|
Post by dburne on Oct 21, 2022 0:22:38 GMT
Hmm I think on this one we will have to agree to disagree. Thanks for the feedback. Ok, but this is not about disagreeing or agreeing, this is how all modern benchmarking is done. BabelTechReviews also removes vsync in VR games - that's why you'll see some games with fps beyond 200, like in this article: babeltechreviews.com/vr-wars-the-red-devil-rtx-6900-xt-versus-the-rtx-3090-founders-edition-part-2/- and here the RTX 4090 is pulling nearly 500 fps in Doom Eternal - but you cannot get a 500 Hz monitor anywhere ;-) You always deactivate vsync to see the true performance: I think where you and I differ - and it may be I am not understanding you correctly on this - is it seems you were saying you could get the same performance fps wise at 144 Hz as you got at 90 Hz, or even 80 Hz, and I just do not see how that could be possible. Back when I was running a Reverb G2 I could lower the refresh rate from 90 Hz to 60 Hz and gain in performance. Similar with the Index. Perhaps there is something different about a benchmark that changes that , but I don't much get into benchmarking so I would not know on that.
|
|
|
Post by Pyroth309 on Oct 21, 2022 2:11:15 GMT
I think where you and I differ - and it may be I am not understanding you correctly on this - is it seems you were saying you could get the same performance fps wise at 144 Hz as you got at 90 Hz, or even 80 Hz, and I just do not see how that could be possible. Back when I was running a Reverb G2 I could lower the refresh rate from 90 Hz to 60 Hz and gain in performance. Similar with the Index. Perhaps there is something different about a benchmark that changes that , but I don't much get into benchmarking so I would not know on that. Yea, I think you just aren't understanding. So for example on Index you can set the Refresh Rate to 80, 90, or 144hz. Refresh rate is how many times your screen is refreshed per second. We measure that in Hertz. Frames per second is how many frames your GPU is rendering and sending to your screen per second. You can generate many more fps than your refresh rate but it can cause screen tearing and other artifacts from the two not being sychronized. So for example, With HMD's having a form of Vsync or Adaptive Sync...if I have it set to 80 hz... my video card can only generate up to 80 FPS to match because it's hard locked by the hardware/software of valve. This keeps the two synchronized somewhat so you don't see the excessive tearing...and it also saves performance and power because your GPU can throttle back and not have to run at 100% to maintain 80 fps.(provided you can generate more than 80 fps) With a synthetic benchmark, there's no reason to cap the FPS since we are just measuring performance of the video card rendering to the settings the HMD requires. So having the "vsync off" in effect uncaps the FPS even though the screen refresh rate is hard locked on the hardware to whatever it's set at. This allows your GPU to push as many fps as it's capable of and run to it's maximum...thus the refresh rate of the screens becomes irrelevant because we are measuring what the GPU can do while pushing this particular hardware to its maximum instead of being limited to only pushing to the locked refresh rate of the screens.
|
|
|
Post by dburne on Oct 21, 2022 12:19:28 GMT
Ok thanks for the info.
|
|
|
Post by Rune on Oct 22, 2022 15:12:40 GMT
Great explanation, and if you're not used to benchmarking gpus, surely it's strange how you suddenly can get more fps than the Hz.
To see real differences, it's very important not to restrict the fps to let's say 60 fps in 60 Hz, but to let the rig generate all the fps it can to see the true differences.
Like Pyroth wrote, vsync can work in several ways, but usually for VR in 90 Hz, you'll be capped at 90 fps (1:1), but if your rig only can do let's say 74 fps and not 90, you'll typically get 45 fps (1:2 - same frame repeated twice to match 90 Hz) - or maybe down to 3 x 30 fps to match 90 Hz. For 120 Hz you could get 120 fps (1:1 - optimal), 60 fps (1:2), 40 (1:3) etc. You can find those limits in VR games - but capping your gpu at 90 fps or 120 fps, when maybe it can render 250 fps would hide its true performance from you - so for real benchmarks it's very important not to limit your fps. An un-capped benchmark like the OpenVR Benchmark will register all those 250 fps
Btw, this is why I try to make 100+ % performance jumps when upgrading my rig, so I'm sure that all 45 fps in 90 Hz turn into true 90 fps. Because you are exposed to the capped fps in the games you play. But for real benchmarks it's important to see the true and uncapped performances.
Well, maybe enough of this, I'm sure there are many guides written about this already.
|
|
|
Post by dburne on Oct 22, 2022 16:53:46 GMT
Great explanation, and if you're not used to benchmarking gpus, surely it's strange how you suddenly can get more fps than the Hz. It was not so much how one can get more fps than Hz, I understand that. My confusion came from 144 Hz benchmarking the same or better than 90 hz.
|
|
|
Post by Rune on Oct 22, 2022 17:07:26 GMT
Btw, some dude posted results from the same rig here - but did not use same res for the different hmds. He did use a RTX 3090: imgur.com/a/VfENGOXHe got 102 fps with the CV1 - much the same as Nalex getting 105 fps with the 3080 Ti. He got 76 fps like Tom with the Index res 100% (Tom uses Vive Pro, but Vive Pro and Index use same res for SteamVR res 100% and both use native Steam drivers) - Tom also uses 3090. I usually get 80 fps, so his number do seem reliable. Interesting was the HP Reverb G2 (9.78 mill pixels): - and I tried to match same res (9.79 mill pixels): The difference was not as marked here, at least not 30% like what we saw with the CV1, but 23% (1% low was 42% higher with the Index). Still a substantial difference. Maybe Meta did not exactly fight for Rift CV1 to get top performance, while HP did... If I compare his CV1 results to my Index running same CV1 res, the difference was 33% - and 23% actually is a much lower difference, for what it's worth. It's also a lot better than the about 40% difference seen with Vive Pro 2. Btw, I got auto-updated from 471.11 to 512.15, but don't see any difference in performance - so no, I don't think 512.15 is key to anything here.
|
|
|
Post by Pyroth309 on Nov 16, 2022 1:14:43 GMT
I still plan to do this comparison.. Just haven't had time. Had another guy quit at work and have been working nonstop. By time I get home I'm too tired to bother setting up VR lol. I am scheduled to be off this weekend so maybe I'll be able to.
|
|