So, in all other games when I use VR (AC, iRacing and Dirt Rally) my EVGA 1080FTW card boosts to 1924MHz. However, in RaceRoom when using VR I noticed the other day it was only showing the stock 1700MHz. However, if I run the EVGA tuner and force "K Mode" which be default operates in turbo mode at all times, this appears to work. Any idea why RaceRoom isn't pulling my graphics card into turbo mode by default? I do notice that my GPU usage never goes above 70% even though with Oculus Debug Tool i can start dropping below 45FPS under certain settings. Is my CPU limiting the cards ability to get to boost mode?
There's clearly a bug here somewhere. No matter what i do, my CPU and GPU don't go above 60% usage when using VR. Hopefully this gets fixed soon.
Without sharing a single setting you are using, or what your PC specs are, it is incredibly difficult to offer any advice. Namely, it's impossible. Please share your PC specs, all of your in game video settings, your Nvidia CP settings, and whatever settings you are using for VR when launching the game. Screenshots work well for that stuff. Edit: Also, did you read this from the header on the forum? Steam VR Beta is broken if you are using that.
Not using Steam VR Beta, although it looks like its fixed now. i5-4690K EVGA 1080 FTW upon start: -vr 2.0 I then use MSI Afterburner to watch GPU and CPU percentage. Never goes above ~60% no matter how high I go. I can max out all the settings below and still the same thing. The frame rate does drop though. Right now, my frame rate is around 45FPS occassionally getting to 90FPS.
Already tried that. It's set at "Prefer Maximum Performance" in both the generic settings as well as the raceroom specific settings.
its the second pass on a beta vr in DX9 ....... its not going to be optimised like a vr dx11 game Andi
Yeah. Kind of what I thought but wanted to make sure. Sounded like a couple folks thought they were getter me full usage. Looking forward to further improvements or dx11. This game has so much potential.
Interestingly enough, after today's update the CPU usage now goes up to ~95% while the GPU hovers around 60%. Looks like I'm getting the most out of the system which is CPU limited at the moment.
Could be the case. I know my 4670k that is overclocked bottlenecks my gtx 1070 in a fair number of games. (I play at 144 fps though, or try to.) I can imagine the gtx 1080 being further bottlenecked by a similar cpu.
I thought a i5-4690K would be enough, but i guess not. Guess, i'll try to minimize anything that may be running in the background.
R3E graphics engine is obviously very CPU limited and VR just highlights the fact. You get a high CPU load on the main thread core while GPU usage stays at a low level. Low GPU usage is also the reason why GPU doesn't easily boost: it simply doesn't need to. I can get my GTX980TI to reasonable boost clocks by using 2.0+ oversampling values. Because of the CPU limited situation there is very little negative impact on FPS compared to most games where such overampling values are mostly unusable. This is of course not a good thing overall, the load should be more on the GPU because it is just far more effective way of doing things.
If I could chuck my 2 cents into the mix regarding this issue. I have done a little testing to try and figure out why my GPU is not outputting 100%. I set ASW off, 3.0 PD (can comfortably run 2.0, so set 3.0 to stress GPU), and went out for a few laps. The FPS were constantly hovering around the 70-90 FPS mark, only settling at 90 a few times. So 'technically' the GPU should have been on max 100% effort trying to maintain 90 FPS - but it wasn't. As you can see from the attached pic max figures, the best it did was 85%. CPU was also not an issue, as you can see all the cores were barely taxed. Temps also not an issue - max 67 deg on the GPU (clearly the CPU would not be O/heating at such low load either). This is a very strange situation - I have never seen such discrepancy in the operation of hardware during gameplay before, with no 'actual' reson for it. There are similar issues with DCS in so far as there are constant 'blips' of negative headroom and FPS dips for no apparent reason at all. I honestly believe no one has the faintest idea how VR actually works, or how to implement it successfully in modern day PC architecture. i7 6700K @ 4200, nVidia GTX 1080ti Strix OC, 16GB G.Skill 3200 CL14, R3E installed on a Samsung 950 Pro M.2 512GB drive, Win 10.
I spent last weekend trying to get my gpu to reach above 82% while running race room. I cannot make this happen but obviously get stutters and poor performance. Jumping back to assetto Corsa and yep i get 95-100% before I get stuttering. Same with Iracing. CPU gets a small tax (67%) for a brief second at the start of the race with 18 AI.. then settles.. yet I get stuttering. This is with a titan (pascal) nvidia card and a i7 extreme over clocked. In fact I get no performance benefits by over clocking. Testing this weekend I over-clocked the cpu to 39% on all 20 threads and over clocked the gpu by +20%. Worse I get the same performance in race room by turning both my video card and cpu into green mode (under-clocking or cool mode). This is running steam vr beta, not beta, various nvidia drivers from the last 6 months and turning adaptive power on/off and telling the video card max performance under that power option. This is on win10 but it only seems to be in raceroom. Sure wish I knew what was going on. For now I turn the shadows to low, reflections to low, contact shadows to 12, visible cars to 15, LOD to medium, AA4, tire tracks off.. This still gets me stutters in the first 2-3 seconds of a race but then it goes away. Mirror medium. No msaa, bloom, post etc. Running PD at 1.5 via the “vr 1.5” argument. Sadly in assetto Corsa I have everything maxed but reflections. Plus post processing and 1.6 pixel density with a field of 20 cars. I reach high 90’s% gpu load at the start. Hope this gets some attention, but glad it’s not just me. Nathan
The GPU load can be reported lower, depending on the bottleneck situation. Say issues on data movement or gpu's frontend processor can yield low usage of the GPU's shading cores. Unfortunately there is not much in terms of decent profiling tools for dx9 anymore. Check your mainboards pcie configuration that x16 is working as expected.
So the thought here is dx9 is creating a bottleneck somewhere besides the cpu or gpu and thus not utilizing the computers main processors? As I mentioned other titles do give the correct gpu useage but to your point are not using dx9 (I don’t know what Iracing uses honestly). Pcie express slots are indeed working as intended, I’ve even played with the auto, x2, x3 and x8 settings to see benchmark differences. I find this hard to swallow - however, I’ll run a older 3dmark that uses dx9 and see what happens with the gpu/cpu load. See if they break 80%.
No, the use of dx9 means, as developer you won't have it easy to do full GPU profiling to find the bottlenecks, as the hardware vendor's tools do not support the old APIs anymore. The other point was that the some parts of the GPU can be bottlenecked, but the reported GPU load is still "low", given the shading processors aren't utilized to full extent. To find out in detail what's going on you need the tools... Or differently put, using dx9 makes it hard to find out in detail where performance is lost today. However, a good tool for stutter analysis is called GPUView and would work with any API https://developer.nvidia.com/conten...ing-video-memory-overcommitment-using-gpuview there is also another guide how to capture here http://nvidia.custhelp.com/app/answers/detail/a_id/3507/~/generating-an-event-trace-log-for-gpuview