The Index thread (please keep to subject)

Deleted.

Deleted.

Mark Poppin from BabelTechReviews just wrote to me that he for personal reasons has retired, so we may not get any more VR benchmarks from BabelTechReviews, which is very saddening. Due to their special setup, BabelTechReviews for years was able to provide the most precise and elaborate benchmarks results for VR I have ever seen. 

The latest results from BabelTechReviews covered the currently fastest video cards - using the Valve Index hmd at SteamVR res 100%:

Source: https://babeltechreviews.com/hellhound-rx-7900-xtx-vs-rtx-4080-50-games-vr/

These results may be somewhat hard to follow - but you can see that the RTX 4090 always is the fastest. If we set the RTX 4090 performance to 100%, and scale the other cards to the RTX 4090 we get (I didn't bother including the 6900 XT):

- corresponding to these specific percentages:

As I do not like mixing games and more synthetic benchmarks, we can now calculate how fast other video cards are against the RTX 4090 - for average VR gaming performance:

Average VR Gaming performance relative to the RTX 4090

RTX 4090 = 100%

7900 XTX = 59%

RTX 4080 = 75%

RTX 3080 Ti = 60%

These gaming results are based on 10 games. From these results you can see that the RTX 4090 is about 67% faster than a 3080 Ti (100/60 = 1.67), and about 33% faster than an RTX 4080 (100/75 = 1.33). And so on. 

It may also be noted that the benchmarks: VR Mark Blue Room and OpenVR Benchmark are very close to the average gaming results. So you can use these benchmarks to nicely predict your real VR gaming performance - actually the benchmarks are slightly more demanding than real games, so expect about 5 - 10% better performance in real games than these VR benchmarks predict. 

Are VR games different and more demanding than 2D games - or do we see similar trends? Luckily TechPowerUp also tested the same cards in 25 different 2D games and found (scaled after the RTX 4090, 4K res to ensure that the results are gpu bottlenecked): 

Average 2D Gaming performance relative to the RTX 4090

RTX 4090 = 100%

7900 XTX = 82%

RTX 4080 = 79%

RTX 3080 Ti = 61% (here counted as a RTX 3090)

Source: https://www.techpowerup.com/review/amd-radeon-rx-7900-xtx/32.html

Interestingly, we can see that for the RTX 4080 and RTX 3080 Ti (or 3090 as these two cards perform similarly) that results from 2D games are much similar to VR games. It seems we can trust that if a new Nvidia video card is twice as fast in 2D games, it's likely to be twice as fast in VR games too. That's good to know. 

For the AMD 7900 XTX many have complained about bad VR performance, and this is also what we see here. In 2D games, RTX 4090 is 22% faster than the 7900 XTX, but in VR games the RTX 4090 is a massive 69% faster! This is why you want to avoid the new AMD flagships for VR games. Some said that AMD has fixed the issue with newer drivers - but that's not true. On Reddit, last week a 7900 XTX owner told me, he's getting the same bad performance as always - 80 fps in the OpenVR Benchmark with Index res 100%. Now that isn't bad, it's similar to the RTX 3080 Ti or 3090, but for a 7900 XTX we would expect at least RTX 4080 performance in VR, and that's not what we're seeing. In short, 7900 XTX works great in 2D games, but VR performance still is way too slow. 

Also, different hmds have different performance due to drivers. Using the results from BabelTechReviews, we can see that using an Index hmd res 100% with a RTX 4090 in the OpenVR Benchmark should result in about 155 fps. And you can see performance for RTX 4080, RTX 3080 Ti and 7900 XTX too. If you have one of those video cards and want to test you performance in native SteamVR games, set a custom res for you hmd corresponding to Index res 100% at 2016x2240 (90 Hz) and run the benchmark. This may also be a good way to test if Link or AirLink - with or without Virtual Desktop - perform the best etc. 

Happy VR Gaming 🙂

I deleted two posts in this thread a few days ago due to too premature results. Results have now been checked several times. Also in the deleted posts, two different but very similar rigs were used, which may per definition create some bias. 

Using the exact same rig (and drivers) and similar res for 1) Quest 3 with Link, 2) Quest 3 with Virtual Desktop (VD) and 3) for the Index, these are the current results from the OpenVR Benchmark. The OpenVR Benchmark only supports native SteamVR drivers (so no OpenXR or native Meta driver support here - but most SteamVR games also require native SteamVR drivers, including Alyx and Vertigo 2, and only few support OpenXR even if the OpenXR trend seems to increase):

Quest 3 with Link (tethered) - res 9.70 mill pixels per eye

AirLink provided the exact same result using 2880 for encoding width, which is the max for AirLink. An encoding width of 2880 was used for both Link (tethered) and for AirLink.  

- and here's the Index - res 9.44 mill pixels per eye 

Note that the Index is running a 3% lower res, so we have to take that into account. The difference is still profound - Index is 37% faster (or about 34% taking the 3% res difference into account), and for the 1% Low the difference is even higher at 46% (about 43% taking the 3% res difference into account).

Virtual Desktop does help to improve the Quest 3 performance:

Quest 3 with Virtual Desktop (VD) - res 9.88 mill pixels per eye

The results may seem much similar to using Link, but note that here the res is 5% higher than the Index, which we will need to adjust for. The difference here to the Index is 25%, but the res is 4.5% higher, so it's more safe to assume that Index is about 20% faster, which also better fits results with Quest 2 using VD. For the 1% Low, Index is here 31% faster - but probably more like 25% faster when adjusting for the 5% higher res.

Now, comparing Quest 3 with tethered Link (30.0 fps) to VD (32.8 fps), VD is using a 2% higher res and is 9 % faster - and 11% faster for the 1% Low. This indicates that Quest 3 users can win about 12% in performance using VD instead of pure (Air-)Link. Maybe 12% does not seem like much - but it's close to the difference between a 3080 and a 3090.

I just use a simple spreadsheet to keep track of games I’ve finished, ones I want to try next, and a score out of 10 so I remember how much I liked them.