01-08-2022 03:19 PM - edited 01-11-2022 06:27 AM
I've seen people claiming lately that using a Quest 2 for PCVR comes with a massive performance cost compared to PC-only headsets. Some posted benchmarks in the Index thread implied that the Index ran 30% faster than the Quest 2 when using similar settings. This didn't seem comparable to my own experience, so I thought I would do my own benchmarks to see how my Quest 2 compares to my Rift CV1.
I tested both headsets using OpenVR Benchmark through SteamVR. I did my best to match pixel count between the two headsets, and ran the test at low res (Rift 100% in SteamVR), and high res (Quest 100% in SteamVR, using 1.7x resolution in the Oculus app). I tested the Quest with AirLink (dynamic bitrate up to 200 Mbps), tethered Link at default bitrate, and also with bitrate raised to 350 Mbps (the point at which I don't notice any compression artifacts).
EDIT: I ran the tests with Virtual Desktop, and those results exceeded both Link and Rift. I don't know why VD performs better in SteamVR, but I've added the results below. In VD I used the Ultra setting, which gave me similar resolution to Link at 1.7x, and a bitrate of 150 Mbps (the highest setting VD).
Here are my low res results:
Rift at 100% (2.15 million pixels per eye):
Quest 2 at 30% (2.21 million pixels per eye)
AirLink:
Tethered Link, default bitrate:
Tethered Link, increased bitrate (350 Mbps):
The Quest 2 is pushing slightly more pixels (3% more), but it was the closest I could get with the way SteamVR adjusts resolution. So looking at these results, the Rift does run a little faster, but not anything near 30%. Airlink runs at 94.5% of Rift's framerate, Tethered Link is at 93.0%, and increasing the bitrate dropped it to 92.6%. So overall, a 6 to 7% performance hit when running the Quest 2 at default Rift settings. That doesn't seem too bad.
Edit: Virtual Desktop at 30% (2.24 million pixels per eye):
Virtual Desktop ran 8.6% faster than Rift, and 14.9% faster than AirLink!
Here are the high res results:
Rift at 344% (7.40 million pixels per eye):
Quest 2 at 100% (7.40 million pixels per eye)
AirLink:
Tethered Link, default bitrate:
Tethered Link, increased bitrate (350 Mbps):
The Quest 2 is again pushing slightly more pixels, but there was less difference than the low res tests. Looking at these results, the Rift still runs a little faster, but with even less advantage. Airlink runs at 95.1% of Rift's framerate, Tethered Link is at 98.1%, and increasing the bitrate dropped it to 95.2%. So overall, a 2 to 5% performance hit when running the Quest 2 at maxed settings.
Edit: Virtual Desktop at 100% (7.48 million pixels per eye):
Virtual Desktop ran 15.5% faster than Rift, and 21.4% faster than AirLink!
Obviously results will vary depending on many factors--I'm using a current-generation GPU that may be more efficient at encoding than older cards, and I have more RAM in my system than the PC used for tests posted in the Index thread. Be that as it may, I started using Quest and Quest 2 for PCVR when I was using a GTX 1080, and although I don't have benchmarks for that setup, I never noticed any performance hit that felt like 30% loss.
I would be interested in seeing other people's results with different hardware, if anyone cares to do some comparative testing. OpenVR Benchmark is free utility on Steam.
On a side note, I haven't used my Rift in quite a while, and OH MY GOD does it look blurry and low-res compared to my Quest 2. Whether comparing them at low or high resolution, the Quest 2 image looks so much cleaner and sharper, while the Rift looks so blurry that I have a hard time reading text in menus, even at 344% super-sampling. It was a great headset in its time, but I could never go back to using it now.
TLDR: On my PC, Quest 2 performance overhead ranges from 2% to 7% compared to my Rift CV1, with less performance loss at higher resolution. AirLink versus tethered Link made little difference.
EDIT: Virtual Desktop ran 9% to 16% faster than Rift, and 14% to 21% faster than AirLink in SteamVR! I will try to do some testing in a game that uses native Oculus drivers to see how Rift, Link, and Virtual Desktop compare in that environment.
01-08-2022 04:09 PM - edited 01-08-2022 04:10 PM
Fantastic breakdown and information! Honestly, it's nice to see some real and unbiased stats on the forum. I haven't done any benchmarking yet (I have never had the patience for it), but considering I played Lone Echo 1 on my Rift CV1, and Lone Echo 2 on my Quest 2, it was painfully obvious to me that there was no real difference in performance at all. Certainly nothing near the fictitious 30% being claimed in the hyperbolic Index thread.
A difference of 6-7% sounds about right, considering that this wouldn't be noticeable to the user without the use of a benchmark tool. And like I said, I never noticed a performance hit going through Lone Echo between Rift and Quest 2.
On top of which, Quest 2 is just so amazingly clear and not hindered by the noticeable SDE in both the Rift and Index.
01-08-2022 04:53 PM - edited 01-08-2022 05:21 PM
Indeed interesting results - compared to these, where Index was 30% faster compared to Quest 2 (similar res, same refresh):
Quest 2 - 2032x2032 - 90 Hz
Note that compared to the OpenVR database these results are normal for a RTX 2060 Super and Quest 2
Index - 2016x2240 - 90 Hz
Note that compared to the OpenVR database these results are normal for a RTX 2060 Super and Index
Index has about 10% higher res here and is about 30% faster - if it's not the streaming impact, then it might indicate native Steam drivers working wonders with the Index compared to Oculus hmds?
My result using Index 2016x2240 (res 100%) is 82 fps, but can't compare with the Quest 2. These results were published earlier - but it's higher res so hard to compare - but least it's a RTX 3090:
@nalex66 it would be interesting to see your Quest 2 results if you could hit about 2016x2240 res in 90 Hz! At least to compare with the dude with the RTX 2060 Super. Very interesting if high-end gpus have less streaming impact - or if native Steam drivers may have a part to play...
Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"
01-08-2022 06:16 PM
@RuneSR2 wrote:! At least to compare with the dude with the RTX 2060 Super. Very interesting if high-end gpus have less streaming impact
The problem is that the example uses Virtual Desktop's streaming technology and nothing was shown for Air Link. Plus, we don't know anything about this person's internet and wifi settings, nor their distance from the PCVR machine when streaming; all of which are highly important to get max streaming capabilities.
At best, the information you provided merely speaks to Virtual Desktop's capabilities, not Quest 2's functionality.
This is why simply doing a copy/paste of 2nd hand information is so misleading.
01-08-2022 06:40 PM - edited 01-11-2022 06:32 AM
I ran my Q2 over AirLink with the resolution as close to 4.5 million pixels per eye as I could--Steam only increments by 2%, so I did one test at 60% (4.43 million pixels) and one at 62% (4.58 million pixels):
Q2 60%:
Q2 62%:
Essentially the same result for both of those runs. Since these tests can't be meaningfully compared to those from a different PC, I ran the Rift too:
Rift at 210% (4.51 million pixels per eye):
Interestingly, the Quest did better than the Rift this time. I'm sure the results vary a bit from run to run, but in any event, there seems to be negligible impact from encoding/streaming in this test.
Edit: I ran this test with Virtual Desktop, too:
Virtual Desktop at 60% (4.48 million pixels):
Virtual Desktop ran 16.0% faster than Rift, and 14.4% faster than AirLink!
As far as different GPU generations go, I've heard that the encoding chip is significantly better on 2000-series and above, compared to 1000-series and older Nvidia cards. That said, I never noticed any big performance differences with my GTX 1080 when I went from Rift to Quest.
I agree that some of the difference in those Q2 vs Index tests may be down to native drivers versus non, which was partly why I thought my benchmarks might be useful--they're as close to an apples-to-apples test as possible, with both headsets running through the same driver/API stack. Apart from minor differences in resolution due to different render target aspect ratios, the encoding process should be the only different factor between them. (Edit: Obviously, the Virtual Desktop results are not apples-to-apples with the Rift and Link results, as VD bypasses the Oculus API and SteamVR wrapper, but it's definitely interesting to see how much better it fares when running a SteamVR app.)
It might be interesting to test your Rift at a similar resolution to your Index, in order to see how much difference there is between Oculus and Steam headsets with this benchmark tool. That could possibly shed some light on the differences observed.
01-08-2022 07:45 PM
The thing to remember with an Index vs Quest 2 comparison in an OpenVR benchmark is that the Index is running natively in OpenVR, while the Quest 2 is being forced through an OpenVR to Oculus wrapper written by Valve. Let's see how both do in an OpenXR benchmark (if one existed), or benchmarked in a game with Oculus and OpenVR support.
Anyway, here's my Quest 2 with link cable against my Reverb G2 (which is also not native to OpenVR, it's running through an OpenVR to WMR wrapper), both at roughly 7.4MP per eye (well, 7.3MP for the Reverb, the SteamVR resolution scale sucks).
A difference of 2.7% (with the Reverb doing about 1.3% less pixels).
01-08-2022 10:27 PM
Thanks kojack, that’s helpful. So we’ve got a 2000-series comparison between Quest 2 and a PC-only headset, with a similarly minor impact from the encoding process.
It’s certainly possible that encoding could have a greater performance cost on low and mid-tier cards. If anyone has a 60 or 70 level Nvidia GPU to test, that might shed more light. Likewise, I’d be interested to see how AMD cards compare.
01-08-2022 10:30 PM - edited 01-08-2022 10:32 PM
@nalex66 I actually wrote a longish note about all this but it seems to have disappeared. Maybe it was too long or something, lol! In any case, no way I'm going to re-write it. I'll just give the main points I was trying to make.
I agree that based on my benchmarking in the past that streamed PCVR with a Q2 will have ~5% lower performance than dedicated PCVR like my Rift cv1 or Vive Pro1. This is because with streamed PCVR the gpu needs to use about 10% of its resources to encode/compress data while this is not necessary with dedicated DP/HDMI PCVR headsets. The overall performance differences will probably vary depending on how good your gpu is. Newer rtx3000 series gpu's have much faster encoders than older gtx1000 and 1600 series gpu's. (and faster vram) as well as any current AMD gpu's.
Dedicated PCVR headsets will also avoid compression artifacts and with higher bitrates distant objects will have better clarity. Likewise, Link with 500mbps bitrate will do a little better than Air Link (or Virtual Desktop) with 100-200mbps bitrates. How important this is will depend a lot on the specific app, the user's pc specs, router/wifi setup, and settings imho.
Benchmark results are useful to help determine if you system is performing in the right ball park. I find that it's much better to then finetune settings based on things like latency, fps, and stuttering/stability. Using them to compare PCVR headset is nonsense and way to subjective to be useful imho.
I refuse to get into any headset vs headset slinging match. I would just say that people need to find out what works best for them, say why, and move on. In fact, I don't believe that there is any one-best PCVR headset out there right now. I use a dual PCVR headset with am AMOLED Vive Pro (with Etsy lens mod and Index controllers) and a Q2 (mainly wireless with Air Link). I believe that this allows me to use the best headset for the job.
I'm very happy with my dual PCVR setup and current pc/router/wifi setup and I doubt that I'll be upgrading anything very soon (18-24 months). One thing I will say is that I doubt that my next PCVR headset is going to be wire-only, lol!
01-08-2022 10:50 PM
I’m not arguing that one headset is better than another; I just wanted to test the claim that Quest 2 suffers a massive overhead from encoding compared to headsets that use DisplayPort or HDMI. Numbers like 30% have been thrown around, and these sorts of declarations convince people that they need to spend many hundreds of dollars to get an acceptable PCVR experience. Maybe the performance deficit is more severe with some systems, but I haven’t found that to be the case with the setups I’ve used.
01-08-2022 11:24 PM
Fair enough mate. Cheers.