GlamFM
Banned
Runs super nice on my 1080, there must be something very wrong with your settings.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
HTC Vive Launch Thread -- Computer, activate holodeck
- Thread starter taco543
- Start date
- Status
Runs super nice on my 1080, there must be something very wrong with your settings.
Zalusithix
Member
Yes, but a 30Hz x+y update would be a worst case basis where it only has one lighthouse hitting the sensors. The majority of the time you'll catch both of them and have a 60Hz update for full positional data.
I haven't watched the videos with audio, but I believe the tracking is updated on the fly as it gets information. It'd correct X on an X sweep and Y on a Y sweep, and beyond that do it as the information streams in as the entire system is always in motion. A certain amount of estimation is required because of this even with the lighthouse "hard" data. The more stationary the object, the less fudgework required.
Rotational and locational drift are much the same thing. Rotation would be slower to drift theoretically due to less movement on that axis, but the correction is the same. The same algorithm that can place the unit within a 3D space can also determine the pose of the controller without an IMU at all. The IMU just helps fill in the blanks between lighthouse updates.
As for the move and headset for PSVR, the headset will not have rotational drift because the pattern of lights allows the camera to correct it. Drift comes from the IMU always basing its location/rotation based on the previous reading. The camera is an absolute. The move controllers don't have the ability to be rotationally corrected by optics due to only having one spherical tracking point. The only way they can manage rotation is a magnetometer update (compass). Or in the case of the PS3, with the additional assumption that you're always more or less facing the same direction (straight at the TV). Edit: I expect in PSVR they'll have a similar technique employed where rotation is partially corrected based on assumptions made by headset orientation and the relative position of the move controllers.
What settings are you using? If you want to maintain 90fps, you're looking at medium settings at best. Even then, with bad weather you will likely still get drops at the start of a race. Otherwise you're looking at reprojection.
It's not the most thorough, but see the graphs in the review below:
http://www.roadtovr.com/nvidia-gtx-...performance-head-to-head-against-the-980ti/3/
As you can see, at high settings, even without AA, neither a 980ti or 1080 reach 90fps very often.
Check the lighthouse. There is a good chance one your lasers died. Mine did. I had the exact same symptoms as you, and HTC replaced it.
I can't even see either of these on my sensors
Working fine, though
Zalusithix
Member
Sensors? Your camera? Or are you referring the the lighthouse as a sensor? The lighthouses are emitters, and the only way you're going to clearly see the laser sweeps is by taking a picture of it. The LED sync array is fairly visible to the naked eye however.
Use a phone camera, they're not visible to the naked eye
Do read the part about accuracy and anisotropic jitter though. While X/Y position on move should be pretty decent (within the confines of the camera frustum), I don't know about Z -- particularly in the absence of a pattern.
Yeah definitely, I just changed everything to low and no anti aliasing and still it only goes to 45 fps. Should I just try reinstalling PCars then? Don't know what's wrong with this game.
Yeah it's been pretty annoying but I'm already long past being able to refund with 6 hours played. Most of it is fuckiing around in menus.
low settings all no AA and low res. It just stays at 45 fps, actually when I changed to all low and recorded the fps it jumped to 90 for about 1 second in the entire race. So that's a little better.
Before you reinstall completely, try trashing the config files located here:
\Users\<username>\Documents\Project CARS\*.XML
There was an update in mid May where they recommended doing this after applying a patch that changed some of the VR config stuff. This will reset most of your preferences though, so keep that in mind.
Anybody aware of STALKER working with VR? Would love to find a method. It was a series I could never quite properly get into, but I think as a VR experience it would be perfect and I would be in it all the time.
Googling it all I can find is a lot of people discussing Oculus approaches
Haha, cool. Yes I was just eyeballing the things.
Googling it all I can find is a lot of people discussing Oculus approaches
Haha, cool. Yes I was just eyeballing the things.
Thanks man I'll definitely try that out.
It kinda worked, now with all low, I even lowered res a notch I can actually get 90 fps but i still feel like my gpu is not being used correctly for this. It still has frequent drops to the 70s and a few times it goes to 45. Shouldn't I get solid 90 fps with settings all low and/or off?
Your best bet is probably vorpX. STALKER is on the list of supported games. vorpX requires a lot of tinkering though, and the final result may still be somewhat underwhelming compared to games built with VR in mind.
I have one of these, could you spell out for dummies how I'd use it to help with the cable?
Even with VorpX, none of the menus/UI will be properly rendered in VR. Most likely the inventory screen will be unusable, and you'll need to take off the HMD and look at your screen to do any inventory stuff.
Anybody else played A Chair in A Room?
Fuck my shit. Me and my girlfriend trying to play it, got like 3 'levels' in, and it was just too much.
Very clever, though. The 'hub' and level/puzzle design is fantastic. So immersive, even if the graphics aren't great.
Hurrrr. Probably not worth it. Unless someone mods it to be VR I'm probably best leaving it.
Spent enough time tinkering Elite: Dangerous, and even that's not perfect yet.
Fuck my shit. Me and my girlfriend trying to play it, got like 3 'levels' in, and it was just too much.
Very clever, though. The 'hub' and level/puzzle design is fantastic. So immersive, even if the graphics aren't great.
Hurrrr. Probably not worth it. Unless someone mods it to be VR I'm probably best leaving it.
Spent enough time tinkering Elite: Dangerous, and even that's not perfect yet.
The only consolation I can give you is that it's quite clearly psychological. It's not like Amnesia's proper monster-horror. It's clear that ACiaR is all in your character's head - more of a Jacob's Ladder kind of shindig. (Not gory or gross like that, just the psychological "this is all clearly imaginary but is still terrifying" aspect.)
It's a lot like playing a David Lynch movie. Down to the slightly pulpy voice acting. A lot of the time it's not scary, just interesting, and then it suddenly spikes you with these moments of pure terror. Sometimes they are truly horrifying, but other times they peter out into being a mundane sort of 'Oh, it wasn't scary at all'.
Really fucks with you. Well written and a great concept, though. Like a legit good puzzler, too.
For the price there seems to be a lot of content there. I recommend giving it a try at least. We've only played it for 81 minutes and that was worth it - enough to know whether to refund. And judging by the 'level select' screen it must be like at least 5 hours long.
It does have an "edge peek" option where you can zoom your view out to see everything, but yeah, still not ideal.
DavionStriker
Member
Did anyone actually try IL-2 yet? So tempted to jump in with the steam sale going on
Zalusithix
Member
Rotational drift is an entirely different beast than positional drift. For position, accelerometer measurements are integrated twice, and acceleration is biased by the (unknown) gravity vector. Rotation integrates gyroscope measurements only once, and there is very little bias.
As a result, orientation tracking works OK if it's uncorrected for many seconds, whereas positional tracking needs to be corrected tens of times per second.
PS Move can correct orientational drift even though it only has a single tracking point (the ball) by comparing instantaneous position changes measured by the IMU to those measured by the camera, but it is a rather indirect coupling. Having multiple tracking points on an object, such as Oculus Rift and the Vive and Vive controllers, and therefore being able to measure position and orientation of a tracked object directly, makes it easier. Here's an in-depth article about that: http://doc-ok.org/?p=1138
I have played with PS Move ( https://www.youtube.com/watch?v=0J5LaWykiIU ), and with a mono camera like the PS3 Eye, noise in Z is bad (noise in x/y is sub-millimeter at decent range). However, the PS4 Eye camera is a stereo camera, which should work much better.
Zalusithix
Member
How would you correct rotational (by this I mean the rotation along the axis formed by length of the Move) drift by comparing to something the camera can't even see. You could spin the Move like a top and the camera would nary see a difference in the bulb position. Error would continually build up over time.
qa_engineer
Member
Does PSVR use stereo cameras or only one? They're quite close together but may be enough combined with the large size of the move orb to derive Z?
crisdecuba
Member
I can't wait to jump into this ecosystem!!
qa_engineer
Member
http://www.psdevwiki.com/ps4/PlayStation_4_Camera
I believe its up to the dev because both the sensors are the same.
Features
- photo, video
- voice commands (available as well with an earset with microphone)
[*]depth calculation/imaging - pad, move, face, head and hand recognition/tracking
[*]one of the cameras can be used for generating the video image, while the other used for motion tracking.
Hardware Overview:
OV580: Back chip connected to Aux port is labeled OV00580-B21G-1C. The OmniVision 580 ASIC is not publicy documented (without signing a NDA with Omnivision), so it seems to have been made specially for Sony (other companies using it are Mantis Vision and LeapMotion) out of the 500/600 series of processors.
others Omnivision USB bridge solutions: PlayStation EyeToy(OV519) and PlayStation Eye (OV534).
OV9713: The 1-megapixel Image Sensors are two CMOS OV9713.
AK5703: The Front chip connected to microphone array, labeled 5703 324A, is a 4-channel 24-bit Analog-to-digital (A/D) converter packaged in a space-saving 28-pin QFN package. Datasheet is available here and here.
4g51a: EEPROM is labeled as 4g51a. It contains the initial configuration for PlayStation 4 Camera
Edit: Upon further review it sounds like both sensors are required for depth sensing.
Finally got afterburner working and showing me fps in game. What an eye opener, seems my rig isn't enough for 90fps pCars VR. Was running at 45fps constantly and dropping whenever there were a bunch of cars bunched up.
16 cars on track, high everything but shadows low, D2X, 8x AF, 1.5 render scale in steamvr.
First track monza, second sarth.
Turning up shadows made it unplayable at 30fps, any time I turned my head the whole image would jump.
16 cars on track, high everything but shadows low, D2X, 8x AF, 1.5 render scale in steamvr.
First track monza, second sarth.
Turning up shadows made it unplayable at 30fps, any time I turned my head the whole image would jump.
The game isn't built for VR, it's obvious. It's a shame, because I do have a lot of fun with it once I turn down the settings. But first and foremost I have to turn down the amount of cars because my 3770k can't keep up. The utilization % on my GTX1080 isn't even that high..
I do get native 90fps on a few tracks at around 8-10 cars though. And once it works, it's a great experience.
Can't wait for racing games built from the ground up for VR and using Nvidias VR hardware accelerated features.
Zepp Twofist
Member
Specialized components and tech that have never been in consumer products before in combination with a relatively low production volume. Tech is guaranteed to become cheaper and better in a few years, but something has to get the ball rolling.
As I said, you do this by comparing instantaneous motion detected by the IMU with motion detected by the camera. Say the IMU detects a motion by a displacement vector of (0, 1, 0) between two camera images, and the camera detects a motion by (0.1, 0.95, 0) over the same time frame. Then the sensor fusion algorithm can deduce that the IMU's orientation has drifted a few degrees around the Z axis, and can counter that drift.
It's not just the HMD that carry the cost though. Besides the HMD there are two great quality motion controls, and not the least the lighthouses themselves. And in that context it's worth noting that one of the best thing about the Vive VR system is that the motion control system doesn't need to be upgraded for the next generation HMDs, it's kinda future-proof.
I thnk the point was, you can twist and turn the move controller whe keeping the ball effectively static in space. How does it handle that?
Pharaoh 117
Member
Thanks. I'm gonna think a little more about it before I decide.
Charles Alan Ratliff
Member
I am so sad right now. I just opened and set up my replacement Vive and it has worse dead pixels than the original headset I sent in. It took me several weeks to get a replacement because they made me send everything back, get a refund, and then reorder one. And the only reason I even got any help is because I contacted Julia Herd, the VP of Global Communications at HTC on Twitter, and today is her last day at the company.
I have been busy moving and getting situated in my new home to follow this stuff. Are there any updates regarding dead pixels? Anything I can actually do on my own? Is customer support better? Thanks
I have been busy moving and getting situated in my new home to follow this stuff. Are there any updates regarding dead pixels? Anything I can actually do on my own? Is customer support better? Thanks
Zalusithix
Member
I am so sad right now. I just opened and set up my replacement Vive and it has worse dead pixels than the original headset I sent in. It took me several weeks to get a replacement because they made me send everything back, get a refund, and then reorder one. And the only reason I even got any help is because I contacted Julia Herd, the VP of Global Communications at HTC on Twitter, and today is her last day at the company.
I have been busy moving and getting situated in my new home to follow this stuff. Are there any updates regarding dead pixels? Anything I can actually do on my own? Is customer support better? Thanks
If you're talking about stuck green pixels on dark grey backgrounds, a SteamVR beta update pretty much eliminated that issue some time ago. No clue if it has been pushed out of beta yet.
Charles Alan Ratliff
Member
Holy shit. I can't believe that worked! Thank you so much.
They were appearing on all dark backgrounds, and not just loading screens, but in games like Audioshield and such. I opted into the SteamVR beta and downloaded the update and tried my hardest to find something wrong, loading up all kinds of stuff, but they are all gone now!
I wonder how many RMAs would have been avoided if this update was out near the beginning. I look forward to becoming an active member in here again. I have a lot of catching up to do!
I bet pretty much all RMAs would have been avoided
My judder/skipping/whatever it is, like a quick side-to-side motion every so often as I move my head, is still around and seemingly unaffected altogether with a fucking GTX 1080. Now I'm back to eliminating variables and I'm trying to string up a tarp to cover all the photos in the room in case something is reflecting. What a pain in the ass.
Have you changed front camera to 30 hz? Not saying thats the solution but it fixed my tracking problems I had early on but I was just losing tracking completely.
My front camera is disabled completely as it was one of the things people advised me to try regarding this issue. I appreciate the suggestion though, would be happy to get any others anyone has :/
Dreams-Visions
Member
Multiplier of 1.5. VR High setting in game. MSAA if you're feeling sassy.
Zalusithix
Member
I guess I'm missing something here. In my control case, the camera isn't going to detect any motion as far as I can see. The tracking sphere would be rotating around its central point. Movement is happening, but invisible to the camera since a uniformly lit sphere is going to look the same no matter which side you observe. As such, there should only be the IMU data to pull from.
Even if movement of the sphere did occur, I don't see how it could be really correlated with the rotation and accurately corrected. If the sphere has physically rotated 5 degrees and translated 2 mm to the left between frames, and the camera sees the 2 mm shift, how would it know whether the rotation calculated by the IMU (say 6 degrees) is correct or not from that? I can see other means of attempting to correct drift error that might be there, but not via the camera.
Glad it worked out for you. That update was certainly one of the best updates for the Vive.
Damn!
That still really sounds like a framerate issue, but it's hard to diagnose this stuff over the 'net, especially for VR.
I still wonder about conflicting software. How consistent is this judder? Does it have like a predictable cadence if you continuously rotate around?
Zalusithix
Member
Out of curiosity, are your controllers stable? No visible jitter? Have you ever run the jitter test on your setup? Tried intentionally disabling one lighthouse and getting closer to the other?
- Status