DCS Benchmarks with Eye-Tracking 5K+ 2080ti

Finaly got eye-tracking working, so here are some rough benchmarks in DCS.
All benched 1st 60 seconds after the mission says ‘start benchmark’. Looking straight ahead (or trying to keep looking straight ahead)
High CPU mission.
Text is from FRAPS summary
stereo_mode_use_shared_parser = true in graphics.lua

Tests run with both Oculus (via opencomposite) and SteamVR rendering.

opencomposite 50% Pitool 1.25 FR Off
2020-08-31 20:00:32 - DCS
Frames: 2979 - Time: 60000ms - Avg: 49.650 - Min: 44 - Max: 52

opencomposite 50% Pitool 1.25 FR aggressive
2020-08-31 20:06:35 - DCS
Frames: 2660 - Time: 60000ms - Avg: 44.333 - Min: 34 - Max: 56

opencomposite 50% Pitool 1.25 FR Conservative
2020-08-31 20:10:32 - DCS
Frames: 2674 - Time: 60000ms - Avg: 44.567 - Min: 34 - Max: 56

Steam VR. 58% 57% Pitool 1.25 no FR
2020-08-31 20:23:04 - DCS
Frames: 3408 - Time: 60000ms - Avg: 56.800 - Min: 47 - Max: 62

Steam VR. 58% 57% Pitool 1.25 FR conservative
2020-08-31 20:37:23 - DCS
Frames: 2866 - Time: 60000ms - Avg: 47.767 - Min: 40 - Max: 54

steamvr 58% 57% Pitool 1.25 no FR
2020-08-31 21:06:37 - DCS
Frames: 3475 - Time: 60000ms - Avg: 57.917 - Min: 48 - Max: 61

opencomposite 33% Pitool 1.25 no FR
2020-08-31 21:12:07 - DCS
Frames: 3453 - Time: 60000ms - Avg: 57.550 - Min: 45 - Max: 61

opencomposite 33% Pitool 1.25 FR Conservative
2020-08-31 21:16:11 - DCS
Frames: 2834 - Time: 60000ms - Avg: 47.233 - Min: 34 - Max: 59

TLDR: I loose frames when running foveated rendering. :frowning_face:

5 Likes

Well that doesn’t sound right. Hopefully there’s something wrong with your test method settings.

3 Likes

We had said before DCS is one of the few titles that doesn’t benefit. Those tests are indeed about what our own tests showed. Titles that are CPU bound often don’t gain.

You will see very significant gains on titles where your graphics card is the limitation most of the time (but not 100% of the time).

11 Likes

lol…isn’t dcs cpu bottlenecked to begin with? not sure if ffr would help. I doubt a 3090Ti
would help either

(EDIT @PimaxUSA beat me to it)

6 Likes

a case in point of “the buyer is not listening”. Others read it and write on reddit “product is bad”

2 Likes

I realise that DCS is CPU bound, and very single core speed dependant. I have spent years tweaking settings to make it useable in VR, since the first week of the release of the Vive, so I’ve done the yards folks.

Here’s the quick bench of the Caucasus F-18 Air to air refulling, about the lowest CPU taxing mission I can find.

opencomposite 33% Pitool 1.25 FR OFF F18 Refuelling mission
2020-09-01 10:08:35 - DCS
Frames: 5383 - Time: 60000ms - Avg: 89.717 - Min: 80 - Max: 91

opencomposite 33% Pitool 1.25 FR Balanced F18 Refuelling mission
2020-09-01 10:12:00 - DCS
Frames: 4836 - Time: 60000ms - Avg: 80.600 - Min: 75 - Max: 83

As you can see, the Foveated rendering still takes nearly 10FPS.

This is VERY early tech, and the software will improve.

I’m still happy to have got the eye-tracking, and will re-visit it when appropriate, but right now, it’s going back in the box.

1 Like

That is literally your only software title?

2 Likes

As reported, these statistics are very far from conclusive.

As my own tests show with Fixed Foveated Rendering, I would still expect Dynamic Foveated Rendering to offer significant improvement (~3ms at 30Hz, 9%) in DCS World use cases that are not particularly CPU heavy, including FA18C air-air refueling in the Caucasus region. In fact, aside from Smart Smoothing artifacts, some extremely CPU intensive community-made maps, and some newer terrain, it is probably worth using most of the time. And if that doesn’t seem to be the case, you can probably tweak things until it is, by cranking up the resolution.

If you have it, you still may as well use it for most things DCS World, based on this info so far. All it doesn’t buy you is a comfortable margin of safety for future DCS World updates, or the ability to turn off Smart Smoothing outright (even combined with a ~35% boost from next gen GPUs).

  • Framerates are an unreliable and unimportant statistic. Frame latency is far more useful, and commentary on any spikes in the performance graphs is needed.
  • FPSVR is by far the only trustworthy VR performance margin reporting tool. FRAPS? Really? For VR?
  • Exact impact - in milliseconds - of CPU and GPU frame latencies is extremely helpful to know.
    * Which headset is used?
  • What PiTool Render Quality, SteamVR Video vertical resolution, and DCS World Pixel Density are used?

Most important of all, is whether it is possible to see the awful green backlight texture when FA18C instrument lighting is turned on using DFR, as it is easily observable with FFR. A report on that would be useful indeed.

EDIT: Ah, I see ‘High CPU mission’. Since you do that stuff, you may not have as much use for DFR. Still though, you should be able to get some benefit with FA18C air-air refueling in the Caucasus region, especially in single-player. Get FPSVR, and do some tweaking.

4 Likes

@PimaxUSA Lol, no. its not my only software title. Last look at steam with VR as a filter shows 157 titles.

DCS is why I have VR though.
DCS and ED are why a LOT of people have VR, and I would say why a lot of those people have a Pimax.

But I’m sure that Pimax would know that, given that you guys have spent lots of time researching what people are playing, or trying to play with your headsets. Oh, wait, that right, I’ve never seen any sort of survey from Pimax looking at that sort of (I would have thought, critical) information.

Have a look at the numbers of simmers that will abandon Pimax in an instant when the Reverb starts shipping.

So, While the community bands together to weather the storm that is Pimax, we post what we feel is useful information in an attempt to help others.

@mirage335, Thanks for your extensive resources and time on tweaking DCS. I do have FPSVR, but find that for quick and dirty FPS comparisons, FRAPS works for me.

I’m happy to run the same benches with FPS VR.
Back in a tick.

6 Likes

Thank you. Please also comment on any spikes - the presence or total absence of these can reveal either more subtle problems with DCS World, or (rare, applicable to only specific DCS World versions) problems resulting in some latency not shown on FPSVR.

Also, FWIW, you might I posted a topic on these forums about the importance of DCS World to Pimax. Most of those who spoke up were more lukewarm about DCS World getting priority - perhaps more enthusiastic responses would have been seen on the DCS World forums.

3 Likes

Eye-tracking suddenly stopped working. :confused: trying to get going again.

1 Like

The way you write you make it look like it would be pimax’s fault that DCS runs like crap. It simply isn’t.
It’s a poorly optimised simulator. A great sim for sure but it simply is very CPU bound and since DFR only helps the GPU this is kinda obvious.
We’ll probably see the very same issue with MS Flight simulator once VR is added since it is hevy on the CPU as well.

2 Likes

never played it. Woah it’s free on steam? nice

1 Like

Sorry, but it would be more accurate to say the way Shrike wrote it was simply confusing.

Pimax, in my opinion, is selling more or less a consumer (even if enthusiast) product, and thus, has at least a good reason, if not a responsibility, to be motivated to ensure VR experiences are optimized for their headsets. DCS World in particular, since I still think it is one of the use cases at least some users will buy Pimax headsets for with usable visual quality as an expectation.

On that point, Shrike is correct.

Unfortunately, from there Shrike’s writeup prematurely makes broad generalizations about expected performance, without sufficient detail, possibly without the right measuring tools, and concluding with “it’s going back in the box”.

I get the impression Shrike actually has some of the right test cases, with extensive DCS World experience, and just did not write the conclusions precisely, so I don’t want to be too harsh. At the same time, part of quality science is precise writing - right now this is junk science.

@BNP Is correct to question the test method. Inadequate for the sweeping broad generalizations drawn.

@PimaxUSA Is of course correct that gains with DCS World from DFR/FFR will be limited. I would clarify though that casual DCS World use cases are a little more GPU bound, and can benefit significantly from DFR/FFR - just enough to push visual quality a bit higher especially on the lower-res 5k+ which needs more subsampling. FWIW, my spreadsheets have long documented that tradeoff, and this sort of thing is why I recommend multiple profiles depending on what simulator loads are expected.

@dstar Is exactly right about how these sort of premature conclusions get taken out of context and repasted elsewhere with even more broad generalizations along the lines of “product is bad”. Even worse, this is how addressing junk science can attract the wrong attention making the problem worse.

1 Like

Well, your post has caused a huge post on reddit about that, pimax is lying to people with the tracker. :wink:

Unfortunate. Already? Link?

Quick search on the reddit thread I can find has no mention of DCS.

@Shrike
When you get eye tracking working again with DCS, please, first thing, see if the green backlight texture with the FA18C instrument lighting turned on, is irritatingly visible at night, preferably a pitch black night. This is very important - it is likely to be the primary difference between FFR/DFR with DCS.

Since 3 days, about the bad tracking, inly usb 2.0 ports , blabla. Not to be overlooked.
op meant well (only slowly in his original post, which confirmed it was running), a second came along, then it was twisted on reddit.

But nothing to do with this thread, correct?

1 Like

This USB 2.0 Arie begann with DCS here in this forum. on reddit, only the usb part can you find.

1 Like