Is VR dead?

[RCHeliguy]

Waiting for the "Take my F'ing money" comments from the latest MRTV video at Pimax Headquarters.

So he is trying out a new prototype lens that they've apparently just created and they've improved the local dimming since his test just a few days ago?

I'm glad they are continuing to improve it, and I hope they do get everything sorted before they ship it.

I guess this only looks like a train wreck to me because of the rapid fire nature of this. "This is broke", "We just fixed it", "this is broke", "We just fixed it". Why are they sending these things out in their current form if they are still rapidly iterating on them?

They aren't done yet! But I guess they like the click bait aspect of this and if they just keep changing things and sending things out so they stay in the headlines?

Maybe some people like that. I don't know. But all I keep thinking, is can you just wait until you finish the product before sending out any more headsets with known issues to be reviewed?

I can't imagine a company worth their salt would use reviewers as alpha testers, but it seems like they are doing just that.

 

[RCHeliguy]

Oh come on!

If this isn't some kind of hype machine, I don't know what is.

How many even better versions are they going to release?

Reading into this from the "worst motivations" side.

If I were trying to manipulate a target audience, I could send out bad copies of something and then seem like a super hero quickly addressing each problem. And if I did this just right the product would even look better at release because of the comparisons with the earlier versions with known issues.

Some might see this as transparency into their inner operations, but typically you wouldn't want to do that unless the inner workings were more impressive than this.

 

[Deleted member 197115]

A better one!

Pirates of the Caribbean: The Curse of the Black Pearl (2003) clip with quote A better one! Yarn is the best search for video clips by quote. Find the exact moment in a TV show, movie, or music video you want to share. Easily move forward or backward to get to the perfect clip.

getyarn.io

 

[blekenbleu]

If this isn't some kind of hype machine, I don't know what is.

The sort of "reality distortion field" that reminds me of a young Steve Jobs...

 

[Spinelli]

Exactly that. Single core speed is very important in VR.

Typically one or two cores if you are lucky are dedicated to feeding VR. Those cores can be completely overwhelmed and impact your frame rates in very meaningful ways.

Interesting. I did some research and can't find much regarding VR taking a noticeable amount more CPU power to run than non-VR. Even if it uses a thread or 2, does it really use a thread or 2 at 90%+ usage? And if it does, are the other threads being used up so much by the game that the 1 or 2 being used for VR results in a performance loss? That would imply a game has near-perfect thread-scaling and near-perfect usage of all a CPU's threads. Most games barely load up more than a few threads. Even then, you'll almost never see something like 8 threads all loaded up to 90%+ simultaneously (unlike, for example, many productivity & benchmarking programs).

Not saying you're wrong. I'm just trying to truly understand and figure this out. GPU is easy. All things being equal, VR obviously takes way more GPU performance for the same framerate than non-VR, but CPU? How bad and hard must all of a CPU's threads be saturated from the game if it has no threads / performance remaining to run whatever additional CPU processing VR needs?

I can understand the importance of CPU in VR to try to have higher lows and less-often lows because dipping under the targeted framerate (according to refresh rate) can be more noticeable and bothersome in VR...but that's a different topic.

DCS is particularly bad about this and is very much bottlenecked by your CPU single core speed because of it's older engine. They just recently released a new beta that is a step towards better multi-threading, but you can google CPU bottleneck and DCS and get a pile of hits.

I thought the problem with DCS was bad, or pretty-much lack of, multithreading regardless of whether running in VR or non-VR...

 

[Deleted member 197115]

It's in NVidia whitepaper

Why does VR have to render twice?

The high-performance requirements are: The need to render the scene twice since we have two different points of view to represent. This will double the amount of vertex processing executed and the time spent executing draw calls in the CPU.

 

[Spinelli]

It's in NVidia whitepaper

Ohhhh, that makes sense. I thought all that stuff, especially the "have to do twice because rendered twice" stuff was almost purely from the GPU.

 

[Gelly916]

Very much looking forward to more sim racers/pilots getting their hands on the Pimax Crystal. I've never owned a Pimax product before, would I be able to use my Index Knuckle controllers with it if I leave my base stations up?

 

[Spinelli]

Very much looking forward to more sim racers/pilots getting their hands on the Pimax Crystal. I've never owned a Pimax product before, would I be able to use my Index Knuckle controllers with it if I leave my base stations up?

You should be able to. They work with other Pimax headsets like the 8K X and 5K Super.

 

[RCHeliguy]

Very much looking forward to more sim racers/pilots getting their hands on the Pimax Crystal. I've never owned a Pimax product before, would I be able to use my Index Knuckle controllers with it if I leave my base stations up?

They are selling a snap on module for lighthouse tracking. It makes the headset a bit larger and is not integral to the headset. It requires a separate rechargeable battey to run off of, and I haven't seen the additional weight of this listed yet or the price.

 

[Leynad777]

Just pulled the trigger on a refurbed Pimax 8K X even my 3080 (with i5 13600k) doesn't sound appropriate, but the "potato"-FOV seems an option and still wider than the Index people say.

For people aiming for a wireless VR experience in Sim Racing: Forget it! I've tried my Pico 4 with Virtual Desktop and tethered. The latency is okay in HL:Alyx and even didn't notice it in Beat Sabber (just tried once), but with rF2 or AMS2 the latency easily triples and performance is bad. And the procedure to run it (turn on VD, headset, controller, pick your boundary, launch VD inside the Pico and connect, G9 switch to 1080p etc.) is so unacceptable with bad performance, it's not an option for me.

I've started using my old Samsung Odyssey for sim-racing instead despite the horrible long-term comfort and low pixel density. I hope I have a similar good experience running the Pimax just with the gyro-sensors like I have with the Odyssey. 3 DOF is fine for me in sim-racing and not having to deal with tracking-issues is a good trade-off. I'm also surprised about the much better performance of the Odyssey compare to my former Vive Cosmos running similar resolutions. The Cosmos is a pain in the a** and not so much the headset itself, the software is just horrible. It refuses to work if it doesn't recognize your environment and it won't recognize it if you made your setup during daylight and want to use it with artificial lights vice versa. You can't just set it up for seating position or 3 DOF and will ask for a new roomscale-setup and only at the end of the setup-process (after both controllers needs to be paired again etc.) you can pick the option to use a simple seating position cirle But even with the seating position it asks for a bright room or refuse to work, certainly the next day.. Other than 2nd hand base-station I will never buy anything from this company again. My 74 mm IPD should be inside the range, but felt off, tiny sweetspot and FOV and useless for full roomscale VR anyway 'thanks' to the bad controller-tracking. And it was jittering like crazy. I've ditched VR because of this

 

[BrunoBæ]

It's in NVidia whitepaper

Why does VR have to render twice?
The high-performance requirements are: The need to render the scene twice since we have two different points of view to represent. This will double the amount of vertex processing executed and the time spent executing draw calls in the CPU.

I just searched for the more efficient Single Pass Stereo rendering.
Because I could remember that iRacing (if I recall right) changed to this rendering method instead of the old or standard one where L/R pic was rendered independent (as in the nVidia quote).

Single Pass Stereo rendering allows the GPU to share culling for both eyes. The GPU only needs to iterate through all the GameObjects in the Scene once for culling purposes, and then renders the GameObjects that survived the culling process.

Unity - Manual: Stereo rendering

docs.unity3d.com

 

[anton_Chez]

Ah yes, the wonderful SRS that Nvidia used to sell their 10 series cards. Then proceeded to include it in ZERO games other than iRacing who did it themselves within their own update.

Amazing company.

 

[Deleted member 197115]

Nvidia Foveated rendering shares the same fate.

 

[BrunoBæ]

Ah yes, the wonderful SRS that Nvidia used to sell their 10 series cards. Then proceeded to include it in ZERO games other than iRacing

Hehe yes I can see my argumentation have a lil problem here

CatsAreTheWorstDogs: But is it actually my problem that no game dev exept iRacing have realized (and included) the most effective rendering method in VR?

 

[ltnordb]

Custom shaders patch for AC also has a single pass stereo setting. Reduces gpu load but introduces some visual bugs from time to time.

 

[RCHeliguy]

Nvidia Foveated rendering shares the same fate.

We just had someone add dynamic foveated rendering support for DCS at the XRToolkit level. It would appear that the VR users are having to do this for themselves. I don't know if what they have created is compatible with the Pimax eye tracking. Time will tell once again.

So far that makes 2 of my big 3 ( iRacing and DCS ) MSFS is also supported, but I don't use that title much.

BTW, they are claiming 30-50% improved frame rates in DCS. Woo Hoo! That's HUGE!

Regarding eye tracking, it's yet one more thing to keep up with. Windows updates can impact the privacy settings so your VR headset's camera suddenly stops working until you allow it access again.

 

[RCHeliguy]

Another person who will be reviewing the Crystal soon said that Seb's recent video also covered his very first experience with a 4090 and that likely impacted his perceptions as much or more than the Crystal did.

Interesting if true, but it also reinforces how important a 4090 is to the performance of a headset at this level. This same guy did have a 3090 before his 4090 and considers the performance improvement night and day.

This guy used specific comments from Seb's video like the details he called out in MSFS as being amazing and yet those same details are perfectly clear in the Aero and have been for a while, if you have a 4090 to drive it at the highest resolution it's capable of.

I'm trying not to read too much into any comments at this point, but I think Seb had an emotional response and would take his comments with a serious grain of salt.

To be clear, I'm not trash talking, I'm just suggesting waiting a little longer to get some more realistic reviews by some more level headed people.

 

[ltnordb]

I trust MRTV as much as I trust Pimax

 

[Deleted member 197115]

We just had someone add dynamic foveated rendering support for DCS at the XRToolkit level. It would appear that the VR users are having to do this for themselves. I don't know if what they have created is compatible with the Pimax eye tracking. Time will tell once again.

So far that makes 2 of my big 3 ( iRacing and DCS ) MSFS is also supported, but I don't use that title much.

BTW, they are claiming 30-50% improved frame rates in DCS. Woo Hoo! That's HUGE!

Regarding eye tracking, it's yet one more thing to keep up with. Windows updates can impact the privacy settings so your VR headset's camera suddenly stops working until you allow it access again.

Foveated rendering is a great thing and works brilliantly in OpenXR, the problem is that NVidia driver implementation was buggy and supported only for handful of games and then they just abandoned it.
But hardware specific implementation will always have adoption problem.