CPU constrained quality settings

[dmulligan]

I'll get this out of the way up front. My CPU is underpowered for Elite Dangerous VR. My 1080 GPU should be able to keep up, though always being ready for the next frame will increase the load on the CPU. I won't be able to upgrade my CPU/MB/RAM for several months and I would still like the best Elite Dangerous experience that is possible with my configuration in at least in solo mode. I'm also hoping this is possible since Oculus lists lower minimum requirements for Windows 8.1+.

What quality settings are most CPU constrained or which ones cause the most CPU usage? Which settings can I turn up that will only add GPU load?

What tools are best for monitoring CPU vs GPU bottlenecks? I assume that FCAT VR is the best tool for detecting FPS and dropped frames but what tools will break down GPU, overall CPU and single core use? Is there any way to see FPS or other stats in my HMD?

Is there any way I can play through exactly the same in game sequence? Is there a best practice for benchmarking for comparing quality settings?

I'm running the latest Oculus Rift software and Windows 10. I'm launching ED using EDProfiler and Steam. ASW should be on by default right? Does it matter that I'm launching ED via Steam? I'd like the option to play in 2d occasionally too.

My system:
i5-3450 3.1GHz-3.5GHz CPU
H66 based MB
EVGA GTX1080 FTW Gaming GPU
Oculus Rift CV1
16GB DDR3 1600 RAM
500GB SanDisk Extreme SSD with plenty of slack space
Windows 10
ED Launched from Steam

I've searched through this forum and Googled for answers already. It's very hard to tell which information is up to date and what has since changed.

Thanks,
David

 

[Morbad]

Shadows, particles, and terrain settings seem to influence CPU the most, and the settings in the audio config can make a difference with a slower CPU in larger battles, but any powerful GPU will need a solid CPU to support the render threads.

All you really need to see is per-core CPU load and the GPU load at the same time. Task manager should be able to do this well enough, though MSI afterburner or Sysinternal's Process Explorer may do it better.

As for testing similar loads, I'd recommend the tutorial missions.

 

[heretowatch]

Shadows, particles, and terrain settings seem to influence CPU the most, and the settings in the audio config can make a difference with a slower CPU in larger battles, but any powerful GPU will need a solid CPU to support the render threads.

All you really need to see is per-core CPU load and the GPU load at the same time. Task manager should be able to do this well enough, though MSI afterburner or Sysinternal's Process Explorer may do it better.

As for testing similar loads, I'd recommend the tutorial missions.

This. Be careful using monitoring programs, though. Oculus has named MSI Afterburner as well as quite a few other monitoring programs as culprits when troubleshooting stuttering and framerate drops in VR. I'm not sure if they're just covering their butts, but it's worth considering. Task manager should do the trick.

 

[dmulligan]

Shadows, particles, and terrain settings seem to influence CPU the most, and the settings in the audio config can make a difference with a slower CPU in larger battles, but any powerful GPU will need a solid CPU to support the render threads.

All you really need to see is per-core CPU load and the GPU load at the same time. Task manager should be able to do this well enough, though MSI afterburner or Sysinternal's Process Explorer may do it better.

As for testing similar loads, I'd recommend the tutorial missions.

There are several terrain settings; quality, work, LOD, material, and sampler. I'll start with all of them turned down and work from there.
Which settings translate to particles?

Task manager is too transient, the graphs don't stick around for long enough which is why I asked for others. I'll give the two you suggest a try.

Thanks!
David

 

[Pville_Piper]

Last night I was checking things out in VR, near or in stations with medium VR settings my I6700k was running 30-40% utilized, the GTX 970 was 80-90% utilized and my Vram was 100% utilized. Kinda weird that my Vram is the bottle neck.

I in ASW whenever I'm near stations or near the surface so I'm maxed out.

 

[dmulligan]

This. Be careful using monitoring programs, though. Oculus has named MSI Afterburner as well as quite a few other monitoring programs as culprits when troubleshooting stuttering and framerate drops in VR. I'm not sure if they're just covering their butts, but it's worth considering. Task manager should do the trick.

Interesting and good information. Google has indeed pointed me to recent positive tests showing Afterburner causing otherwise unexplained frame drops.

 

[dmulligan]

I am using the Oculus Debug Tool with the Performance HUD. Unfortunately the overlay is in the center of the screen. I've chosen to drive a course around Dav's Hope looking for Polymer Capacitors while I play with settings. I started with VR Low and I am turning up settings one at a time watching my frame rate and "App Motion-to-Photon Latency." I am trying to keep my frame rate at ~90Hz with only occasional drops . The latency appears to be different than the 11ms latency quoted elsewhere because I can maintain 90Hz with ~25ms latency. Some settings push that up to 35+ which brings the frame rate down to 45 and drops from there.
% overhead has been an eye opener. With VR Low I am still getting mostly below 0% overhead which is quite bad.

 

[NW3]

Interesting and good information. Google has indeed pointed me to recent positive tests showing Afterburner causing otherwise unexplained frame drops.

Thanks for the tip. I use Afterburner and I always thought my (mostly minor) frame drops were due to ED code. I'm going to do some testing without Afterburner running.

 

[Alec Turner]

I've chosen to drive a course around Dav's Hope looking for Polymer Capacitors while I play with settings.

While you're there why not take a crack at the latest Buckyball Race!

https://forums.frontier.co.uk/showt...ub-presents-Land-of-Hope-and-Glory-(Dec-5-11)

 

[xondk]

Thanks,
David

Best way to monitor, can be logitech ARX, you don't need logitech keyboards or such to get the arx component, that allows you among other things to see cpu and gpu load, in a scene where you are losing fps or feeling laggy, check or to see what is maxed out, if gpu is maxed then gpu is the problem, if cpu is maxed its likely cpu, if neither, something else is up, likely related to software.

 

[dmulligan]

After some more testing I've found that using VR Low settings and no AA or SS, I can get 90 fps pretty reliably with 0-15% overhead and 19-20ms latency. I think I needed to turn off the Windows 10 creators update recording stuff to get that improvement.

I've noticed in some low to mid spec GPU threads people are running with SS and not running AA. Is AA mostly superfluous when using SS? The white papers on both technologies make it seem like they are.

Until recently once a CPU was fast enough the game wouldn't perform better. I understand that an 3rd gen top of the line i7 is the minimum CPU recommended for ED. Modern CPUs are much faster, so how well would an 8th gen i5 with 6 cores/6 threads or Ryzen 5 with 6 cores/12 threads perform on a 1080, Rift and ED?

 

[dmulligan]

I just found a 3770k used which is both the minimum recommended CPU and the best CPU my motherboard can take. I'm glad I searched here though. It sounds like it's barely enough. It's just not with spending any more on the CPU/mb/RAM setup.

D

 

[dmulligan]

After disabling Windows' Xbox DVR i am getting much better results with approximately medium VR settings. It's at the point where I doubt upgrading to an 8th gen i7 would make me $800CAD happier if you get my meaning. I'll post my settings next time I'm in front of my desktop.

Questions:
If you could only use anti aliasing or super sampling which would you choose for ED?
Does ReShade help on an Oculus Rift?

 

[vic]

Reshade does not work with VR!

 

[dmulligan]

Ok. I just entered a Combat Zone for the first time since getting my Rift. The ghosting isn't as bad as normal stuttering since the stars move properly but it's still very annoying. If lowering my settings doesn't help I'm going to have to bite the bullet and upgrade my CPU, mb and RAM.
Is an i7-8700k going to be better enough than an i5-8600k to be worth the extra money? I have a 1080.

 

[zarrazee]

Ok. I just entered a Combat Zone for the first time since getting my Rift. The ghosting isn't as bad as normal stuttering since the stars move properly but it's still very annoying. If lowering my settings doesn't help I'm going to have to bite the bullet and upgrade my CPU, mb and RAM.
Is an i7-8700k going to be better enough than an i5-8600k to be worth the extra money? I have a 1080.

I went for an i7-8700k o/c at 4.7Ghz. I love it. VR is so much better than my old rig. The 8th generation is so expensive atm, i thought another two hundred pounds is neither here nor there.

 

[dmulligan]

Although I have no doubt that one of the 8th gen i5-8600(k) CPUs would have been a significant upgrade and would have worked well, but I couldn't shake the feeling that I would have regretted not getting the i7. Then I found this post by IronmanG saying he wishes he had gone with a 8700k instead of the 8600k.
https://forums.frontier.co.uk/showt...ooking-good!?p=6142867&viewfull=1#post6142867
I got a i7-8700k today! I will build it tonight. Which cooler did you use zarrazee? I'm hoping to get 5.0GHz out of mine using an Arctic Cooling Liquid Freezer 240.

 

[zarrazee]

Which cooler did you use zarrazee? I'm hoping to get 5.0GHz out of mine using an Arctic Cooling Liquid Freezer 240.

I went for Cryorig R1 Ultimate Dual Tower CPU Heatsink with 2x 140mm Fan. It was a b*t*h to fit, took me several attempts, but now it's on I aint taking it off again! I'm idling at 26 degrees celsius on the CPU and 20 on the motherboard whilst in windows. Not bothered checking whilst in-game. I should do really, lazy sod!

I actually got a pre-o/c tested chip from Overclockers. As it comes with a 3 year warranty I'm not going to bother trying to push it anymore.

Good luck with your new chip, I'm sure you'll be very happy with it!

 

[TorTorden_VR]

I went for Cryorig R1 Ultimate Dual Tower CPU Heatsink with 2x 140mm Fan. It was a b*t*h to fit, took me several attempts, but now it's on I aint taking it off again! I'm idling at 26 degrees celsius on the CPU and 20 on the motherboard whilst in windows. Not bothered checking whilst in-game. I should do really, lazy sod!

I actually got a pre-o/c tested chip from Overclockers. As it comes with a 3 year warranty I'm not going to bother trying to push it anymore.

Good luck with your new chip, I'm sure you'll be very happy with it!

I'm running a measly corsais 115i, also a mother to fit in the case, not enough room to wedge a piece of paper between the fans and the RAM modules, and I had to dismantle the IO port hoodie to get in at all.

Anyhoo, using corsair link, I can tell my reservoir usually stays just over room temp by a few degrees, and the cpu itself not much warmer than that.

Still not bothered overclocking though.

 

[dmulligan]

Time to finalize this thread.

Here is what I learned about running a sub-spec CPU (i5-3450 3.1GHz):
1) Using the lowest possible VR quality settings it is possible to use a fast modern GPU, such as the GTX1080, and CPU that isn't quite as fast as recommended at least when using Oculus' ASW. Note: ASW requires Windows 8.1+.
2) Using almost medium VR quality settings ASW was annoying in Conflict Zones. The effect was ships ghosting/bouncing back and forth as they quickly cross my field of view when ASW is filling in frames that would otherwise be dropped. Lower settings may have alleviated this.
3) SRV driving is bad enough normally. ASW is okay occasionally here but with almost medium settings I still found myself wondering if my nausea was my driving or ASW kicking in and/or frame rate dropping. Too bad the performance debug hud is always in your face when turned on. If it's me I might be switching to 2d for long SRV excursions.
4) Since there is no deterministic way to test changes it is difficult to determine the best settings. Some changes have obvious effects, and others are too close to effectively notice the differences in minimum and average frame rates, dropped frames, and ASW frame counts. There is no apparent way to reset the debug hud counters.

My lower spec hardware kept up most of the time I spent playing the game. It is frustrating that only occasionally during close up combat in very busy CV or Res, often flying near or inside stations and always planetary surfaces cause frame drops. I think this frustration is what finally led me to the decision to upgrade sooner than later. I had the choice. If I didn't have the choice I was tending towards using the lowest settings only due to exceptional load.