(GURU3D) Valve reducing cost of VR for people who won't buy it anyway

[moonbogg]

At least that's my opinion. They are reducing the GPU requirements and want to make the system work well with older GPU's, such as KEPLER 680. They suggest a GTX 970, but if the $350 is too much for someone to pay for a GTX 970, what makes them think that person will throw down $800.00 anyway on the Vive?

http://www.guru3d.com/news-story/valve-working-on-solution-to-run-steamvr-on-weak-gpus.html

 

[tential]

At least that's my opinion. They are reducing the GPU requirements and want to make the system work well with older GPU's, such as KEPLER 680. They suggest a GTX 970, but if the $350 is too much for someone to pay for a GTX 970, what makes them think that person will throw down $800.00 anyway on the Vive?

http://www.guru3d.com/news-story/valve-working-on-solution-to-run-steamvr-on-weak-gpus.html

If you have a GTX 970 and you seriously think you'll be enjoying the VR future, you need help.

Not to mention Kepler? Seriously?

VR's largest hurdle is old hardware. It should be pushed as a premium option, not some budget thing that all hardware can do. That's the perfect way to kill VR on arrival. You should be getting Pascal/Polaris for VR at a minimum, and probably 2 cards if VR supports it at the time.

 

[moonbogg]

I agree, but VR graphics already suck so hard though. Who wants to be immersed in that crappy looking world from the early 90's anyway? lol. They need better graphics and they need to require way more powerful hardware.

 

[tential]

It should be focusing more powerful hardware and optimizing for that not on legacy hardware.

 

[Flapdrol1337]

If these methods make it more playable on modest gpu's they could be useful on big gpu's as well.

 

[moonbogg]

Yeah I don't know. They are the experts so lets see what they come up with.

 

[Headfoot]

Foveated VR seems like a promising research path that should help older GPUs as well as new ones

 

[moonbogg]

Foveated VR seems like a promising research path that should help older GPUs as well as new ones

Also, when you move your eyes, you are blind during the movement. You only see when your eyes stop moving. Your brain fills in the blank and makes you think you saw the whole time, but pay attention. You didn't see crap. Look in the mirror and move your eyes. Try to see your eyes move. You can't, but if you track something with your eyes or intentionally try to make them move slow and smooth in a circle, maybe you can do it then, but otherwise when just looking around, you are blind and only see when they stop.
They should add that crap to VR. That would save about 30% on GPU power. Imagine you look the the left and BOOM, screen goes black. It comes back on when you are looking at the thing you wanted to see.

 

[pj-]

Optimization is always a good thing. If they can run acceptably on hardware that old, it means anything above that is pretty much guaranteed a good experience.

Seriously, are you complaining that GPUs are being made relevant for too long? In the not too distant future your 980tis are going to be hunks of crap and I'm sure many people will be grateful that they don't need to shell out another $700 (x2) to play the latest VR games.

The fastest way to kill VR would be to cater only to the 1% of people with high end GPUs. If you are interested in VR you should be thanking valve for this. It doesn't affect me if someone wants to do crappy looking VR on a 680, aside from creating a larger pool of potential customers for developers to make living doing VR games.

 

[NTMBK]

Also, when you move your eyes, you are blind during the movement. You only see when your eyes stop moving. Your brain fills in the blank and makes you think you saw the whole time, but pay attention. You didn't see crap. Look in the mirror and move your eyes. Try to see your eyes move. You can't, but if you track something with your eyes or intentionally try to make them move slow and smooth in a circle, maybe you can do it then, but otherwise when just looking around, you are blind and only see when they stop.
They should add that crap to VR. That would save about 30% on GPU power. Imagine you look the the left and BOOM, screen goes black. It comes back on when you are looking at the thing you wanted to see.

Good luck getting the timing right on that. The latency to make that unnoticeable would be insanely low. You would end up seeing momentary flashes of black loads of the time.

The idea is to make sure that the image shown is as close to correct as possible, when the "blind moment" finishes. Without high framerate it increases the odds that you see a "bad frame", and start feeling nauseous.

 

[pj-]

Good luck getting the timing right on that. The latency to make that unnoticeable would be insanely low. You would end up seeing momentary flashes of black loads of the time.

The idea is to make sure that the image shown is as close to correct as possible, when the "blind moment" finishes. Without high framerate it increases the odds that you see a "bad frame", and start feeling nauseous.

The stuff about making the screen black while your eyes are moving is nonsense but foveated rendering is already working and will certainly be included on second generation VR headsets. It will be essential when we get to 4k per eye and beyond.

https://www.youtube.com/watch?v=2HS2p2BmVsk

 

[antihelten]

Also, when you move your eyes, you are blind during the movement. You only see when your eyes stop moving. Your brain fills in the blank and makes you think you saw the whole time, but pay attention. You didn't see crap.

This is a myth and it's false.

Whilst the brain does filter out quite a bit during saccadic eye movements, it doesn't filter out everything and you are most certainly not blind.

Certain patterns are actually more easily perceived during saccadic eye movement (a phenomenon known as intrasaccadic perception):
http://www.ncbi.nlm.nih.gov/pubmed/11549741

Either way what you mentioned (blanking the screen), is actually already implemented on both the Rift and the Vive, but for very different reasons (i.e. it has nothing to do with saving GPU power and everything to do with reducing smearing/bluriness): http://blogs.valvesoftware.com/abrash/down-the-vr-rabbit-hole-fixing-judder/

 

[ShintaiDK]

Half done VR, just like the Sulon Q and other low end hardware is only going to drive people away if they try it. And they are not going to try a second time after the first experience.

It seems there is a desperation because the current possible consumer base is rather small (970/290 and up). But its a tactic that will backfire hard.

 

[tential]

Half done VR, just like the Sulon Q and other low end hardware is only going to drive people away if they try it. And they are not going to try a second time after the first experience.

It seems there is a desperation because the current possible consumer base is rather small (970/290 and up). But its a tactic that will backfire hard.

I doubt the 970 and 290 can handle it well. I think focusing on optimizing for old hardware so they can play on all low settings Will do exactly as you said. It will hurt vr.

People act like because the minimum specs say 290/970 that it will be a good experience and that's completely false.

The sets are 800 dollars I mean people with 290s and 970s aren't even that interested in this. It should have started optimizing for maxwell and Fiji, and really focused on Polaris and pascal.

Not try to bump Kepler and Tahiti to the bare minimum. Because a person still holding onto Kepler or Tahiti definitely should not be a focus of a premium product they have no intention of buying.

 

[pj-]

I doubt the 970 and 290 can handle it well. I think focusing on optimizing for old hardware so they can play on all low settings Will do exactly as you said. It will hurt vr.

People act like because the minimum specs say 290/970 that it will be a good experience and that's completely false.

The sets are 800 dollars I mean people with 290s and 970s aren't even that interested in this. It should have started optimizing for maxwell and Fiji, and really focused on Polaris and pascal.

Not try to bump Kepler and Tahiti to the bare minimum. Because a person still holding onto Kepler or Tahiti definitely should not be a focus of a premium product they have no intention of buying.

The developers aren't stupid, they know that VR is different and minimum spec is much more important than it ever has been.

I don't think any game developed for VR will ship with bad performance on a 970/290 in the next year at least. I would bet oculus play tests games and doesn't allow them on their store if they don't meet a certain performance requirement. Most of the "oculus ready" linked on oculus' website come with a 970.

Not everyone who is interested in VR can afford at $1500+ computer with high end specs. There are going to be a lot of people, most likely the majority, playing with <= gtx 980 level GPU.

That said, I don't think valve intends to market vive to people with 680s. The fact that it CAN run on a 680 is a good indication of the power of their dynamic performance scaling and other optimization efforts.

 

[bystander36]

If you have a GTX 970 and you seriously think you'll be enjoying the VR future, you need help.

Not to mention Kepler? Seriously?

VR's largest hurdle is old hardware. It should be pushed as a premium option, not some budget thing that all hardware can do. That's the perfect way to kill VR on arrival. You should be getting Pascal/Polaris for VR at a minimum, and probably 2 cards if VR supports it at the time.

You'd be surprised at how enjoyable 3D looks when properly setup, even with reduced settings. I'd rather play a mix of medium to high settings in proper 3D than Maxed in 2D. The biggest problem is support for it. Since DX11, support for 3D has gone way down.

 

[pj-]

I think people are overlooking the fact that these first generation headsets are not particularly high resolution, considering how few pixels you will be looking at in any given moment.

I've heard that a good comparison for the kind of detail to expect in a VR game is to watch non-vr footage of that game on youtube at 480p.

You can add all the fancy tech you want but much of it will be a waste until the screens improve.

 

[antihelten]

I doubt the 970 and 290 can handle it well. I think focusing on optimizing for old hardware so they can play on all low settings Will do exactly as you said. It will hurt vr.

People act like because the minimum specs say 290/970 that it will be a good experience and that's completely false.

The sets are 800 dollars I mean people with 290s and 970s aren't even that interested in this. It should have started optimizing for maxwell and Fiji, and really focused on Polaris and pascal.

Not try to bump Kepler and Tahiti to the bare minimum. Because a person still holding onto Kepler or Tahiti definitely should not be a focus of a premium product they have no intention of buying.

A 970 or 290 will be just fine for the first few years. People seem to forget that the vast majority of VR titles currently in the pipelines look like this:

Job Simulator:

Lucky's Tale:

Edge of Nowhere:

You will of course also find some games that are a bit closer to what we expect from todays PC games such as Eve Valkyrie, but even this game isn't really that demanding (low-res textures and low poly models), and seeing as it's also coming to the PS4 and PSVR, it should have no problem running on a 970/290:

The only really demanding games are a few tech demos (like Cryteks Climb), and non-VR games that have had VR bolted on (Project: Cars, Elite etc.), but those will be the minority of titles released (although they may very well also be some of the best ones).

More Oculus launch titles here

I think people are overlooking the fact that these first generation headsets are not particularly high resolution, considering how few pixels you will be looking at in any given moment.

I've heard that a good comparison for kind of detail to expect in a VR game is to watch non-vr footage of that game on youtube at 480p.

You can add all the fancy tech you want but much of it will be a waste until the screens improve.

The screens are actually decently high resolution (2160x1200), plus the rendering buffer is even bigger (by a factor of 1.4), so the actual resolution is 3024x1680. This is then reduced a bit thanks to intelligent cropping (by using a stencil mesh), but even so the performance will probably be roughly comparable to what you have with a 1600P screen.

 

[Snarf Snarf]

The screens are actually decently high resolution (2160x1200), plus the rendering buffer is even bigger (by a factor of 1.4), so the actual resolution is 3024x1680. This is then reduced a bit thanks to intelligent cropping (by using a stencil mesh), but even so the performance will probably be roughly comparable to what you have with a 1600P screen.

Not sure how many of you guys have actually tried the consumer version of the Rift but this about sums it up. A lot of the titles are going to be aimed directly at the 970/290 segment, but every demo I have tried at conventions the systems had much more powerful cards running. AMD was at PAX South running Fury X systems with EVE Valkyrie and I have to admit the experience was excellent. So excellent that if I didn't already put the money down for a summer vacation this year I would have pre-ordered the Rift after the demo.

 

[pj-]

The screens are actually decently high resolution (2160x1200), plus the rendering buffer is even bigger (by a factor of 1.4), so the actual resolution is 3024x1680. This is then reduced a bit thanks to intelligent cropping (by using a stencil mesh), but even so the performance will probably be roughly comparable to what you have with a 1600P screen.

Yes, they are high resolution, but they are crammed up in your face and cover a very large portion of your FOV. If you play a VR game that simulates your desktop, the amount of physical pixels your virtual monitor spans would probably be 640x480 or worse.

My point was that you can crank up the settings in VR all you want, but that isn't necessarily going to translate into a significantly better looking experience.

If you were playing The Division at 640x480 do you think you could tell the difference between Medium and Ultra settings? (That is hopefully an exaggeration but should demonstrate what I'm talking about)

 

[Raduque]

I'm almost completely uninterested in VR except for one application: Racing and flight games. Being surrounded by the fully-rendered cockpit of your chosen machine, able to fully look around to see more of the road/track or looking around to track targets in a flight/space game. That's what makes VR exciting for me. I think my 970m will do just fine with those games in VR.

Maybe Mech games will make a comeback.

Edit: Though at $800 for a headset, well, I can think of a lot better use for that $800.

 

[antihelten]

Yes, they are high resolution, but they are crammed up in your face and cover a very large portion of your FOV. If you play a VR game that simulates your desktop, the amount of physical pixels your virtual monitor spans would probably be 640x480 or worse.

My point was that you can crank up the settings in VR all you want, but that isn't necessarily going to translate into a significantly better looking experience.

If you were playing The Division at 640x480 do you think you could tell the difference between Medium and Ultra settings? (That is hopefully an exaggeration but should demonstrate what I'm talking about)

You are of course correct that the angular pixel resolution (or pixel per degree PPD) is quite low. Oculus Rift has a PPD of roughly 15, whereas a 27 inch 1080p monitor is about 30 PPD, and a modern smartphone is at about 70 PPD. In other words the Rift and Vive are roughly comparable to a 27 inch 960x540 monitor, or alternatively a 54 inch 1080P monitor viewed at a distance of 20 inches (not that anyone would ever do that I imagine).

Generally speaking we definitely need to get something higher than the 2160x1200 resolution to match what we're currently used to. We would probably need somewhere around 8K though before we really hit the sweet spot, so it will take some time.

 

[moonbogg]

This is a myth and it's false.

Whilst the brain does filter out quite a bit during saccadic eye movements, it doesn't filter out everything and you are most certainly not blind.

Certain patterns are actually more easily perceived during saccadic eye movement (a phenomenon known as intrasaccadic perception):
http://www.ncbi.nlm.nih.gov/pubmed/11549741

Either way what you mentioned (blanking the screen), is actually already implemented on both the Rift and the Vive, but for very different reasons (i.e. it has nothing to do with saving GPU power and everything to do with reducing smearing/bluriness): http://blogs.valvesoftware.com/abrash/down-the-vr-rabbit-hole-fixing-judder/

In other words, I was perfectly correct for anyone who uses plain language. You did the typical internet thing where you found some nit picked exception, naturally.

saccadic masking is what its called, and yes, when you move your eyes, you don't see anything between movements.

http://nowiknow.com/temporary-blindness/

 

[pj-]

In other words, I was perfectly correct for anyone who uses plain language. You did the typical internet thing where you found some nit picked exception, naturally.

saccadic masking is what its called, and yes, when you move your eyes, you don't see anything between movements.

http://nowiknow.com/temporary-blindness/

Even if you are completely blind while your eye is moving, the idea you suggested is nonsense.

"Imagine you look the the left and BOOM, screen goes black. It comes back on when you are looking at the thing you wanted to see."

No part of this fits with how computer graphics work, nor would it be a performance gain over normal foveated rendering.

 

[TeknoBug]

Every company and their fricking mother are jumping on the VR bandwagon, the VR market is going to crash.