Crashing Down The Mountainside: 'Dirt Rally VR' Hands-On

[kcarbotte]

Codemasters added Oculus Rift support for "Dirt Rally" and released it on the Oculus Home Store a couple of weeks ago. We took a long look at the game, and now we think you should, too.

Crashing Down The Mountainside: 'Dirt Rally VR' Hands-On : Read more

 

[-Fran-]

Thanks a lot for this review!

I have a question though. Do you *need* VR support for a game? Is there a problem if you want to use any HMD as a "monitor slapped to your face"? I mean, I was thinking you could add the movement tracking thingy to the axis's that move the camera and be done with it. I don't know if there's an additional thing you might need, like correcting the image for each eye or something.

Cheers!

 

[kcarbotte]

Thanks a lot for this review!

I have a question though. Do you *need* VR support for a game? Is there a problem if you want to use any HMD as a "monitor slapped to your face"? I mean, I was thinking you could add the movement tracking thingy to the axis's that move the camera and be done with it. I don't know if there's an additional thing you might need, like correcting the image for each eye or something.

Cheers!

Glad you enjoyed it.
As to your question, yes and no.
You don't necessarily need VR support, but its not nearly as simple as stapping a screen to your face.
The image has to be warped to fit the view, the game's scaling has to be corrected for VR too.

There are some third party apps that will convert standard games, but not all games are supported by that.

The Rift and the Vive are not treated as monitors by your graphics card, so you can't even "just do it anyways."
Look up Vireio Perception and VorpX (both covered on the website) to see what I'm talking about.

 

[Rhinofart]

So, the VR aspect of this game requires way more graphics horsepower? I play this game on the highest settings at 2560 x 1440 with my i7 980x OCed to 4.25Ghz, and my XFX R9 290x stock with zero problems at all. I have no lag, I don't know the FPS I'm getting, but I do know it's smooth, and doesn't lag what so ever.

 

[rhysiam]

So, the VR aspect of this game requires way more graphics horsepower? I play this game on the highest settings at 2560 x 1440 with my i7 980x OCed to 4.25Ghz, and my XFX R9 290x stock with zero problems at all. I have no lag, I don't know the FPS I'm getting, but I do know it's smooth, and doesn't lag what so ever.

Yeah I was surprised by the performance figures reported here too. For years Codemasters have done a fantastic job of getting seriously nice look games out that don't require ultimate hardware. In terms of raw display the rift is only 2160x1200 @ 90hz, which from a raw pixel throughput perspective is significantly fewer pixels per second than a standard 1440p @ 60hz monitor (156M pps vs 221M pps).

I'd be really interested to know why the VR setup taxes the system so much more.

 

[kcarbotte]

So, the VR aspect of this game requires way more graphics horsepower? I play this game on the highest settings at 2560 x 1440 with my i7 980x OCed to 4.25Ghz, and my XFX R9 290x stock with zero problems at all. I have no lag, I don't know the FPS I'm getting, but I do know it's smooth, and doesn't lag what so ever.

Yeah I was surprised by the performance figures reported here too. For years Codemasters have done a fantastic job of getting seriously nice look games out that don't require ultimate hardware. In terms of raw display the rift is only 2160x1200 @ 90hz, which from a raw pixel throughput perspective is significantly fewer pixels per second than a standard 1440p @ 60hz monitor (156M pps vs 221M pps).

I'd be really interested to know why the VR setup taxes the system so much more.

VR doesn't need an average of 90fps, it requires a minumum of 90fps.
There's big difference there. When the game dips to 60fps, even for a moment, its very discernable in a VR HMD.

You also got your math wrong.
2160x1200 @ 60Hz = 156M pps
2560*1440 @ 60Hz = 221M pps

2160x1200 @ 90Hz = 233M pps
But you also have to consider that the GPU can't let the frame rate drop, so it would actually average well above 90fps if frame synchronization wasn't in the picture.
VR is just hard on your GPU.

 

[GR33kFR34k]

I’ll give you a few hints. "One" through "six" represent how serious the bend ahead is. The higher the value, the more aggressive the turn will be."

I'm pretty sure this is the other way around.

 

[rhysiam]

VR doesn't need an average of 90fps, it requires a minumum of 90fps.
There's big difference there. When the game dips to 60fps, even for a moment, its very discernable in a VR HMD.

You also got your math wrong.
2160x1200 @ 60Hz = 156M pps
2560*1440 @ 60Hz = 221M pps

2160x1200 @ 90Hz = 233M pps
But you also have to consider that the GPU can't let the frame rate drop, so it would actually average well above 90fps if frame synchronization wasn't in the picture.
VR is just hard on your GPU.

Thanks for the response. Epic maths fail there on my part... that's embarrassing!

I hear you about the minimums and I can't seem to find any benchmark list that reports minimum fps, but according to hardware.info even the lowly gtx 960 can average 120fps at medium settings at 1440p. Maybe there are frequent dips below 90fps, I don't know. Irrespective, it seems fair to conclude that a 960 provides a fantastic experience at medium settings on a 1440P @ 60hz monitor, yet the roughly 3 x faster 1080 has to settle on the same medium detail settings for VR, despite the similar pixel-per-second demands of each display.

I suppose what I'm actually wondering is whether VR is inherently more demanding than a 1440P @ 60hz display, or is it just that we need to shift from average to minimum fps to gauge VR performance? If you measure raw FPS, is VR similar to 1440P?

I realise this is a news post not a full review, so not expecting concrete testing/answers here. But given the rise of VR, I'd certainly be interested in this featuring in future articles, particularly alongside the CPU aspects (which you mention as a possible culprit for the 1080 results). What's the impact of positional tracking on the CPU? Do more cores help? All interesting questions which I don't believe have been explored much as yet... if they have, post me some links!

 

[rhysiam]

I’ll give you a few hints. "One" through "six" represent how serious the bend ahead is. The higher the value, the more aggressive the turn will be."

I'm pretty sure this is the other way around.

Yeah you're correct.

I actually always thought it was the recommended gear to take the corner in assuming a 6 speed gearbox, but it turns out that's not quite right: http://blog.codemasters.com/dirt/04/co-driver-calls-explained/

In any case, 6 = fast corner, 1 = slow

 

[kcarbotte]

I’ll give you a few hints. "One" through "six" represent how serious the bend ahead is. The higher the value, the more aggressive the turn will be."

I'm pretty sure this is the other way around.

You're right. it's been amended.

 

[kcarbotte]

VR doesn't need an average of 90fps, it requires a minumum of 90fps.
There's big difference there. When the game dips to 60fps, even for a moment, its very discernable in a VR HMD.

You also got your math wrong.
2160x1200 @ 60Hz = 156M pps
2560*1440 @ 60Hz = 221M pps

2160x1200 @ 90Hz = 233M pps
But you also have to consider that the GPU can't let the frame rate drop, so it would actually average well above 90fps if frame synchronization wasn't in the picture.
VR is just hard on your GPU.

Thanks for the response. Epic maths fail there on my part... that's embarrassing!

I hear you about the minimums and I can't seem to find any benchmark list that reports minimum fps, but according to hardware.info even the lowly gtx 960 can average 120fps at medium settings at 1440p. Maybe there are frequent dips below 90fps, I don't know. Irrespective, it seems fair to conclude that a 960 provides a fantastic experience at medium settings on a 1440P @ 60hz monitor, yet the roughly 3 x faster 1080 has to settle on the same medium detail settings for VR, despite the similar pixel-per-second demands of each display.

I suppose what I'm actually wondering is whether VR is inherently more demanding than a 1440P @ 60hz display, or is it just that we need to shift from average to minimum fps to gauge VR performance? If you measure raw FPS, is VR similar to 1440P?

I realise this is a news post not a full review, so not expecting concrete testing/answers here. But given the rise of VR, I'd certainly be interested in this featuring in future articles, particularly alongside the CPU aspects (which you mention as a possible culprit for the 1080 results). What's the impact of positional tracking on the CPU? Do more cores help? All interesting questions which I don't believe have been explored much as yet... if they have, post me some links!

VR games are generally more demanding, and it has a lot to do with correcting for lens distortion. VR games are often rendered at 1.4x the output resolution to help with image clarity. We touch on this in the Oculus Rift review. http://www.tomshardware.com/reviews/oculus-rift-virtual-reality-hmd,4506-6.html

Your suggestions for future articles are all questions I'm desperate to answer for you guys. At this time there aren't any proper tools to run benchmarks like this in VR that record the data. We're working on it, though. Keep your eyes peeled for more on this topic. We'll have more in the near future.

 

[rhysiam]

VR games are generally more demanding, and it has a lot to do with correcting for lens distortion. VR games are often rendered at 1.4x the output resolution to help with image clarity. We touch on this in the Oculus Rift review. http://www.tomshardware.com/reviews/oculus-rift-virtual-reality-hmd,4506-6.html

Your suggestions for future articles are all questions I'm desperate to answer for you guys. At this time there aren't any proper tools to run benchmarks like this in VR that record the data. We're working on it, though. Keep your eyes peeled for more on this topic. We'll have more in the near future.

Thanks Kevin - that page you linked is a relevant read. Looks like there's quite a bit of ground work required before we can start accurately quantifying the VR experience. I appreciate the response and look forward to seeing how this all plays out in future articles.

 

[-Fran-]

Glad you enjoyed it.
As to your question, yes and no.
You don't necessarily need VR support, but its not nearly as simple as stapping a screen to your face.
The image has to be warped to fit the view, the game's scaling has to be corrected for VR too.

There are some third party apps that will convert standard games, but not all games are supported by that.

The Rift and the Vive are not treated as monitors by your graphics card, so you can't even "just do it anyways."
Look up Vireio Perception and VorpX (both covered on the website) to see what I'm talking about.

Thanks for the response and I will check them out.

Cheers!

 

[kyle382]

That moment when your navigator yells "4 over crest, jump maybe" and you are already moving at 120mph. 0_0 We definitely jumping homie, hold on to your codpiece.

this game is great.

 

[mac_angel]

lol, 1 through 6 is the suggested gear. I'm pretty sure it explained that in the very first of Collin McRae (spelling?) games.

 

[Arcaias]

The thought of someone getting into dirt rally VR with 100% ignorance makes me very warm inside. This game is so challenging, and the VR aspect is promising. Can't wait to try it in VR.

 

[Joshua_106]

http://chilp.it/c09316a

 

[Joshua_106]

The link above is a permanent fix to windows 10 from breaking the old logitech drivers for the g27. Install 510 drivers the run the WheelDriverCleaner.exe this is straight from Logitech but hard to find on there website.

 

[Joshua_106]

This link is the fix to keep windows 10 from breaking the logitech 5.10 drivers for logitech g27. http://chilp.it/c09316a it is straight from logitech. Without it everytime you unplug the g27 and restart windows and plug the wheel back in, you get the driving force conudrum.

 

[Poul Wrist]

So a Radeon 390 and Fury handle the game as well as a Geforce 1080? Or is there something amiss here? :S

 

[mitch074]

AFAIK, most VR conversions these days go the "render upscale then warp and downscale" way - a mobile game maker decided to warp the geometry and render at native and got impressive performance. Too bad no-one else use it...

 

[kcarbotte]

lol, 1 through 6 is the suggested gear. I'm pretty sure it explained that in the very first of Collin McRae (spelling?) games.

The first car only has 4 gears, but they still call 5 and 6 all the time. Gear doesn't add up.

 

[srmojuze]

So, the VR aspect of this game requires way more graphics horsepower? I play this game on the highest settings at 2560 x 1440 with my i7 980x OCed to 4.25Ghz, and my XFX R9 290x stock with zero problems at all. I have no lag, I don't know the FPS I'm getting, but I do know it's smooth, and doesn't lag what so ever.

Yeah I was surprised by the performance figures reported here too. For years Codemasters have done a fantastic job of getting seriously nice look games out that don't require ultimate hardware. In terms of raw display the rift is only 2160x1200 @ 90hz, which from a raw pixel throughput perspective is significantly fewer pixels per second than a standard 1440p @ 60hz monitor (156M pps vs 221M pps).

I'd be really interested to know why the VR setup taxes the system so much more.

VR doesn't need an average of 90fps, it requires a minumum of 90fps.
There's big difference there. When the game dips to 60fps, even for a moment, its very discernable in a VR HMD.

You also got your math wrong.
2160x1200 @ 60Hz = 156M pps
2560*1440 @ 60Hz = 221M pps

2160x1200 @ 90Hz = 233M pps
But you also have to consider that the GPU can't let the frame rate drop, so it would actually average well above 90fps if frame synchronization wasn't in the picture.
VR is just hard on your GPU.

Hi all, don't forget for VR you need to render two separate cameras at the/ of the exact same time.

A simplification would be that at 2160x1200 two scenes have to be rendered - 1080x1200 @ 90hz ~minimum~ for left eye and 1080x1200 @ 90hz ~minimum~ for right eye.

You can imagine the demands are high as Kevin mentions.

Because of parallax one can't render a big scene and "slice" it in two for the left and right eye, your brain will compensate.

Also, there are other things as Kevin and posters have mentioned where there's a lot of pixel-mangling because rendering to a HMD requires "tweaks" to the normal rendering of one single straightforward 16:9 screen at regular FOV.

So, while each eye is needs only half of 1080p you need to do this render "twice" with a variety of pixel manipulation, wider or different FOV, while ensuring both scene cameras are exactly in sync, both scene cameras apply post effects accordingly, and both scene cameras are in sync and following the head tracking suitably, all at 90hz minimum.

Nvidia VRWorks and AMD LiquidVR SDK apparently try and solve this, not sure what's being used in the shipping titles, whether it's these, game engine or custom or a combination of those... which will certainly make benchmarking VR games and the Red vs Green rivalry even more... intense.

Nvidia VRWorks especially with Nvidia Pascal promises: "Lens Matched Shading" (decreases graphics quality of parts of the scene which will be "chucked" or "squashed" by the VR headset output stage), "Single Pass Stereo" (draws geometry once then simultaneously projects left-eye and right-eye views, instead of current 2-camera-pass render per frame), and "VR SLI".

AMD LiquidVR says it can do "Efficient low-latency GPU head tracking" (GPU-powered head tracking?), "Affinity multi-GPU", "Asynchronous shaders" and "Direct-to-display". ( http://www.pcgamer.com/amd-liquidvr-vs-nvidia-vrworks-the-sdk-wars/ )

Things will certainly be interesting now that consumer VR is here.

 

[realtrisk]

Wow. I used to run this game on DK2 with settings maxed on a GTX 980, and I'd get silky-smooth 75FPS. The extra resolution and extra 15FPS of the CV1 must really add a lot of stress to the GPU.