The Index thread (please keep to subject)

Not easy remembering all these numbers - but here are the burdens the gpu have to carry when starting to increase the SteamVR resolution with the Index - I just did the math 😇 🤓 (Note that the Vive Pro uses same res as the Index, so the numbers below also represent the Vive Pro)

It does give a nice idea about what happens when going from res 100 % to 500 % - or from 9 mill pixels to 45 mill. 

Index typically needs about res 200% to shine - if temporal antialiasing (TAA) is used, you may need at least res 300% to significantly reduce the blur. With TAA, the higher the Index res, the lesser the blur. Just try Wolfenstein Cyberpilot - going from res 250% to res 375 % feels like getting a new and much more high-res hmd!

Note that ss with the CV1 is calculated in a different way - here ss 2.0 is about 17 mill pixels, thus similar to Index res 200%, but for the CV1 ss 2.5 is about 27 mill pixels and corresponds to Index res 300%, not 250%. Comparing different hmds is complicated 😉 

Not sure how interesting this is - but maybe better than nothing, lol:

" The latest update to SteamVR includes a curious addition… a new SteamVR Home destination that includes CT scans of dead bugs “found lying on the ground” outside of Valve’s office."

Source: https://www.roadtovr.com/valve-adds-bugs-latest-steamvr-1-23-update/

Direct link to the bugs:

https://steamcommunity.com/sharedfiles/filedetails/?id=2846608501

Since Valve considered The Lab a full game, maybe they consider the bugs some major release too? 🤔 😄 

Full info here:

https://store.steampowered.com/news/app/250820/view/3387289425146917362

So wireless is finally coming - but I don't plan on getting that, I don't mind the wire and don't need more batteries to charge...

Seems like the wireless Index kit is very popular - devs wanted about 200k on Kickstarter, seems they may have gotten 500k soon, it's already over 400k:

https://www.kickstarter.com/projects/nofio/nofio-wireless-adapter-for-valve-index/

It does look awesome - and the lowest latency you can get - it seems:

More info here:

https://www.roadtovr.com/valve-index-nofio-transmitter-kickstarter/

Btw, fun vid - for 5 years never had such issues with the cable, lol:

I keep hearing Index users say that going beyond SteamVR res 150% has no/little impact on image quality. Nothing could be more wrong. So I tried to take some shots showing differences in image quality when using Index res 100% (about 9 mill pixels per frame for both eyes), 200% (about 18 mill pixels) and 400% (36 mill pixels). Here are some examples, click on an image to enlarge or differences can be harder to see:

GNOMES & GOBLINS

Gnomes & Goblins with res 400% - full-size image

Gnomes & Goblins with res 100% - zoomed-in image

Gnomes & Goblins with res 200% - zoomed-in image

Gnomes & Goblins with res 400% - zoomed-in image - still 90 fps with a RTX 3090

OBDUCTION

Obduction with res 400% - full-size image

Obduction with res 100% - zoomed-in image

Obduction with res 200% - zoomed-in image

Obduction with res 400% - zoomed-in image. See how the lines in the windows are clear. The game now performs in 45 fps, but motion smoothing works wonders

WOLFENSTEIN: CYBERPILOT

Wolfenstein: Cyberpilot with res 400% - full-size image

Wolfenstein: Cyberpilot with res 100% - zoomed-in image

Wolfenstein: Cyberpilot with res 200% - zoomed-in image

Wolfenstein: Cyberpilot with res 400% - zoomed-in image. Now down to 45 fps, but again motion smoothing works wonders. In this game the difference between res 200% and 400% is like night and day...

Mona Lisa: Beyond the Glass

Mona Lisa: Beyond the Glass - with res 400% - full-size image

Mona Lisa: Beyond the Glass with res 100% - zoomed-in image

Mona Lisa: Beyond the Glass with res 250% - zoomed-in image

Mona Lisa: Beyond the Glass with res 500% - zoomed-in image

When I had a GTX 1080, I was thinking that Index res 200% was close to as good as it gets for image quality. It may depend on the game, Mona Lisa does win something going from res 250% to 500%, but not that much - while for a game like Wolfenstein: Cyberpilot the difference between res 200% and 400% is profound. Also thanks to pyroth309 for mentioning that using extreme res greatly improved image quality - also blurry TAA (temporal antialiasing) becomes very sharp using extreme res. 

The take-home message: try to increase res as much as you can, especially distant objects (see Obduction above) can become much more clear and sharp. This can greatly increase 3D depth perception (=how far you can see clearly). You also gain more antialiasing.

For those new to super sampling, Nvidia explains the mechanisms quite nicely here (focusing on a 2D monitor, but it works similarly for VR):

Getting and understanding a new hmd is not easy these days, as both panel res, software res and refresh rates may vary a lot. You could end up buying a HTC Vive Pro 2 and getting very bad performance - and then you may think it's because the panel and software res is much higher with Vive Pro 2 than other hmds. I would be able to accept that - fair enough, you have to render more pixels and performance shrinks. Unfortunately this is not the whole story. Having to use an extra set of drivers easily costs tons of performance. Therefore it's very important to test performances of two hmds by comparing the same pixel burden. See what happens when we do that with Vive Pro 2 and Index:

Using same res, the Index is 40% faster than Vive Pro 2 - and 40% (ok 38%) is a massive difference. Sure getting that RTX 4090 will make the Vive Pro 2 run better, and you may not notice the 40% performance hit - but it will still be there in games needing native SteamVR drivers.

Looking at the above image, Reddit newbies often say "hey, you're using different cpus - the difference is because the i9 10900K is faster than the i7 8700K!" - and then I have to argue that the test is not cpu bottlenecked at all, which many can't seem to comprehend.

But how many fps can the i9 10900K do in this test before it does get bottlenecked - well, at least 300:

The 8700K is only 11% slower than the 10900K and should be able to do at least 270 fps before it may get bottlenecked:

So no, the 40% difference between Vive Pro 2 and Index has nothing to do with the cpus, lol. 

In short - do be careful getting a new hmd, if you do not know the performance impact of an extra layer of drivers. 

Such considerations have kept me waiting for the Deckard - if such a hmd will arrive and is great in other areas too. 

Btw, just to confirm - found another example, here Index is 43% faster than Vive Pro 2 when using same res (9.6 mill pixels, RTX 3090):

While we're at it - some dude tested several hmds using the same rig and a RTX 3090, and his numbers for CV1 and Index match what Nalex (3080 Ti) and I got (RTX 3090):

https://imgur.com/a/VfENGOX

- so numbers seem reliable. My interest was the HP Reverb G2 though - he got (9.78 mill pixels):

- and with the Index pushing 9.79 mill pixels I got:

Note that I get the same performance with 471.11 and 512.15, so I don't think Nvidia drivers help here - and the G2 dude having a faster cpu does not help either. The difference was not as marked here, at least not 30%, but 23%. Still a substantial difference. 

According to Road To VR, the Index still is the best overall hmd, and they did consider both Varjo Aero and Quest Pro too:

https://www.roadtovr.com/best-vr-headsets-in-2020-buying-guide/

Quote: "If you’re looking for the very best overall PC VR headset, Valve Index is our pick. It’s pricey compared to the rest, but has an excellent balance of quality, performance, and comfort. That’s why we called it “the enthusiast’s choice” in our full review of the headset."

Personally I feel that the RTX 3090 is a great match for the Index - so no, Index is not old, it's first with current high-end gpus we can't start to fully enjoy the Index. Especially if you're into 144 Hz with high levels of super sampling (=high res). Well, actually I'm down to only 80 or 90 Hz in most games with the ultra-high res I prefer - so I seriously will need at least a RTX 5090 to start pushing Index to the hmd's limits.

Btw, BabelTechReviews measured performance using the Index in many different games and benchmark apps using different video cards - note that Index and Vive Pro have similar performance due to same res and both use native Steam drivers, and RTX 3080 Ti and 3090 usually have same performance too:

Source: https://babeltechreviews.com/hellhound-rx-7900-xtx-vs-rtx-4080-50-games-vr/5/

Note the relatively poor performance with the 7900 XTX, might be a driver issue, not sure it's fixed yet. Also several Index users say that 144 Hz does not work with the high-end AMD cards. I'd be very cautious buying any video card from AMD these days. 

Results from RTX 3090 and RTX 4090 are also found here:

Source: https://www.reddit.com/r/ValveIndex/comments/y3bf0s/valve_index_from_rtx3090_to_rtx4090_85_benchmark/

This is a difference of 85% when comparing apples to apples in the OpenVR Benchmark. Comparing RTX 3080 Ti and 4090 at BabelTechReviews, these 85% translate to about 67% in games (range 33% to 79%). Sniper Elite got only 33% faster (cpu bottleneck?), while Moss 2 got nearly 80% faster.

When I upgraded from GTX 1080 to RTX 3090, my OpenVR Benchmark score got about 130% faster - this did translate nicely into all 45 fps ASW 2.0-activating games changing to solid 90 fps with the Rift CV1. So personally going from 3090 to 4090 will not be a similar jump in performance - maybe the RTX 5090 will be 🙂
Do expect that OpenVR Benchmark changes are very close to average changes in gpu-bound games, and games are likely to be gpu-bound when pushing high levels of res (=super sampling). 2c.