DCS2.5.4 Update a Mixed Bag of Tricks

[FragBum]

Okay just did the update and a quick test in NTTR.

 

Okay seems there is something new going on with CPU usage <now> hitting 28% (ramps up and down) this was around 16% in 2.5.3.

 

What is interesting is we now have a second core hitting 100% utilization here normally only core 11 is hitting high 80 + % but now there are 2 cores.

 

 

 

What happened to my peripheral vision in Rift only the tiniest center portion of the Rift image is in any sense clear and the peripheral portion of the display is really fuzzy, sorry guys this ain't good enough I rely on being able to roll my eyes to look at the extents when flying Gazelle at high speed through built up areas it now looks like someone smeared Vaseline around the major portion of the display.

 

This was very acceptable and usable in 2.5.3 but not anymore.:cry: <edit> no this doesn't work currently. This might have been a typo

 

My biggest question is with all the additional CPU utilization what exactly are we getting?

 

I will do some more testing but this reduction in image quality is really pointing to going back to 2.5.3 ATM

Edited December 21, 2018 by FragBum
<edit>

[Magic Zach]

Would like to hear some answers as well

[FragBum]

Still not sure what is causing this change in CPU usage currently.

 

The Rift image quality change, this seems most definitely a mixed bag.

 

The Rift now seems to have a cone or sweet spot with eyeballs looking straight ahead at the rift your face can be looking anywhere but now you have to look into this area and yes maybe that looks kinda okay maybe better. If you just move your eyes horizontally by only a few degrees left or right without moving your head the image now goes out of focus where as this was not the case in 2.5.3.

 

The vertical aperture seems wider and the effect is the same outside of this new sweet spot.

 

No Rift to face position is correct ;)

 

Actually it seems like there <is an artificial> Spherical Aberrations of the displayed Rift image.:noexpression:

Edited December 21, 2018 by FragBum
<edit>

[FragBum]

Actually Caucasus map feels great now :thumbup:

 

It feels a bit different in VR but looks and flies great. :thumbup:

[Schmidtfire]

Framerate is great at the cost of a tighter sweet spot. But jaggies are next to gone, FPS is a lot better and that was a huge problem for DCS VR.

To achieve this type of optimization, like the "shaders mod", peripheral vision loose a bit of clarity.

 

The overall immersion, quality and playability is way way up. Even during "heavy" lifting like rocket attacks, multiple bombs and intense dogfights it manages to provide a smooth experience now (at least for me)

Over time it will only get better, so I see this as a great leap forwards :) Great job Eagle Dynamics team!

Edited December 21, 2018 by Schmidtfire

[Wolf8312]

Framerate is great at the cost of a tighter sweet spot. But jaggies are next to gone, FPS is a lot better and that was a huge problem for DCS VR.

To achieve this type of optimization, like the "shaders mod", peripheral vision loose a bit of clarity.

 

The overall immersion, quality and playability is way way up. Even during "heavy" lifting like rocket attacks, multiple bombs and intense dogfights it manages to provide a smooth experience now (at least for me)

Over time it will only get better, so I see this as a great leap forwards :) Great job Eagle Dynamics team!

 

+1

Don’t even notice the sweet spot, all looks great for me! Huge step in the right direction!

[FragBum]

+1

Don’t even notice the sweet spot, all looks great for me! Huge step in the right direction!

 

Actually it's the oppersite for me I find it somewhat distracting the "sweetspot" is just such a small area. If anything it needs to be larger having said that I'm also wondering if there is an element of lens aberration in the Rift and now it's just more noticeable as you move your head and scan the DCS world image.

 

I am also thinking that the perceived changes in image just initially feels wrong and you need to relearn how it feels I spent a few hours yesterday and the change is way less grating than it was initially and even less so this morning.

 

The other perception change seems to be near field medium field rendering seems different but that was fairly easy to adapt to.

 

<Just to qualify the 2 comments above even to the point of feeling a bit queasy and I am a fairly seasoned VR sim pilot.>

 

Now the interesting stuff I've been watching task manager and to my surprise this morning DCS was using 27% CPU usage up from about 16~18% same for other maps but it varies depending but that's outstanding like about 80% more. I say surprise because I did check it yesterday and it seemed the same as it had been prior to 2.5.4 so I considered my initial observation that DCS was indeed more CPU intensive was wrong well not so.

 

Tell me it's not placebo effect DCS just feels more responsive!

 

Based on the improvements and the concept that it can only get better in the long term 2.5.4 is a keeper. :thumbup:

Edited December 23, 2018 by FragBum
<edit>

[FragBum]

I have solved the image quality in Rift!

 

basically what I did.

 

Anisotropic filtering was causing the issue as the near field image looked good the mid to far field looked rubbish, full of aliasing artifacts. So I began tweaking and removed all MSAA processing but the issue was still persistent. I then removed my modest 1.1 PD setting still no change. Dropped the AF setting from 8 to none, bingo. Okay no where near as sharp over all so I ramped MSAA back up to *4 better still not great then I ramped PD up interestingly I can now run PD at 1.3 and avoid FPS drops in NTTR where as prior to 2.4.5 update I could not with or without AF and it looks amazing!

 

Not only looking awesome now and I thought I had it looking and preforming great in 2.5.3 I am seeing increases in CPU usage corresponding with improved image quality in (Rift) VR with improved responsiveness in the SIM.

 

Stay tuned and input welcomed.

[Zoomer]

I have solved the image quality in Rift!

 

basically what I did.

 

Anisotropic filtering was causing the issue as the near field image looked good the mid to far field looked rubbish, full of aliasing artifacts. So I began tweaking and removed all MSAA processing but the issue was still persistent. I then removed my modest 1.1 PD setting still no change. Dropped the AF setting from 8 to none, bingo. Okay no where near as sharp over all so I ramped MSAA back up to *4 better still not great then I ramped PD up interestingly I can now run PD at 1.3 and avoid FPS drops in NTTR where as prior to 2.4.5 update I could not with or without AF and it looks amazing!

 

Not only looking awesome now and I thought I had it looking and preforming great in 2.5.3 I am seeing increases in CPU usage corresponding with improved image quality in (Rift) VR with improved responsiveness in the SIM.

 

Stay tuned and input welcomed.

I think your 1.1 PD was too low, you are not going to get clarity with that. I run a PD of 1.4 with the shader mod, MSAAx4 and AFx16. I have found definitely a more smoother experience. For us running RTX 2080`s etc. DLSS would be a big help. We could drop MSAA and get a bump in res quality. The future of VR is in smart tech such as DLSS & VRS to render and fill the peripheral areas of our vision.

[Sandman1330]

I’ve noticed no difference at all this update. PD 1.5 and MSAA 4x, no shader mod. I haven’t seen any difference in quality, even looking specifically for the new sweet spot everyone talks about, I can’t see it.

 

FPS is also still pegged at 45 with ASW off, no change there. But then I’ve always been CPU limited (waiting for Ryzen 3000 series!!!), and I think the optimizations are aimed at reducing GPU usage.

[FragBum]

I’ve noticed no difference at all this update. PD 1.5 and MSAA 4x, no shader mod. I haven’t seen any difference in quality, even looking specifically for the new sweet spot everyone talks about, I can’t see it.

 

FPS is also still pegged at 45 with ASW off, no change there. But then I’ve always been CPU limited (waiting for Ryzen 3000 series!!!), and I think the optimizations are aimed at reducing GPU usage.

 

Yes I'm seeing FPS still 45FPS I do prefer ASW on. I simply can't run PD at 1.5 I hit CPU and GPU limits with the accompanying drop frames.

 

As for CPU usage you should find there is more CPU activity under 2.5.4 prior to 2.5.4 I would only hit about 16% CPU usage I am now seeing up to 30% CPU usage NTTR map.

 

Ah that sweet spot from what I can see it's seems to be straight ahead at center of view and effects different ranges of view depending on how much processing you dial also the width of it varies. The width can be very narrow maybe 1 degree at a few hundred mtrs to several degrees at 20mtrs. The area that is well processed seems to vary with the level of processing as far as I can tell. There is also a portion of lens aberration within the Rift as well.

 

Try this as in the video.

 

Park your self in front of a mesh fence move your head left to right but keep your eyes focused on the the portion of the fence straight ahead of you even though you head is looking left or right. You should see a difference or change in detail level of the mesh.

 

TBH I don't think the peripheral image is as clear as it was 2.5.3 was, indeed no where near it but I guess that will be fixed one day.

 

I think your 1.1 PD was too low, you are not going to get clarity with that. I run a PD of 1.4 with the shader mod, MSAAx4 and AFx16. I have found definitely a more smoother experience. For us running RTX 2080`s etc. DLSS would be a big help. We could drop MSAA and get a bump in res quality. The future of VR is in smart tech such as DLSS & VRS to render and fill the peripheral areas of our vision.

 

 

I had the system tweaked well under 2.5.3 Rift image was near as good as screen display with the limitations of Rift of course and smooth constant 45FPS. however with 2.5.4 I have had to turn off Anisotropic Filtering it just looked horrible at anything past maybe 100mtrs.

 

I tweaked PD to 1.1 with MSAA at 4 to keep CPU and GPU from hitting 100% running any more PD and Rift became a stop frame video effect.

 

Oh as a side note we don't want any VRS (Vortex Ring State) that's bad for helicopters. :P :D

[BoneDust]

I’ve noticed no difference at all this update. PD 1.5 and MSAA 4x, no shader mod. I haven’t seen any difference in quality, even looking specifically for the new sweet spot everyone talks about, I can’t see it.

 

FPS is also still pegged at 45 with ASW off, no change there. But then I’ve always been CPU limited (waiting for Ryzen 3000 series!!!), and I think the optimizations are aimed at reducing GPU usage.

 

Same for me. PD running 1.6 MSAA 4X AS 16X No shader mod. FSP pegged at 45 but rarely drop below that.

[FragBum]

Well more on performance and CPU usage. (Now that I have image quality back to about what is was like prior to 2.5.4)

 

After tweaking last night I checked Caucasus map and there is a small but possibly significant difference here in the NTTR screen shot one core seems to be pegged at 100% but in the Caucasus map this doesn't happen. as per

 

 

Average usage for test mission Caucasus map.

 

This is from my first post but looks representative for NTTR test mission, notice middle row 4th column core seems pegged at 100%

 

 

Interestingly NTTR was the least CPU/GPU intensive map prior to OB 2.5.4 whilst that in itself isn't an issue as the map still preforms well I just wonder if that core should be pegged at 100%?

 

Anyone else seeing this increase in CPU usage it appears quite significant?

[chumpnut1323]

Can't confirm fps improvement or not, but there's definite improvement with stuttering. I actually haven't noticed any during my single player test's so far. ED deserves my money now :)

[FragBum]

Well it's official Wags says no improvements in VR with 2.5.4

 

Fair enough however for me 2.5.4 broke Anisotropic Filtering big time I've tweaked settings to try and get it as good as 2.5.3 VR performance but it's not as good.

 

TBH DCS seems more responsive with 2.5.4 and it appears to be using more CPU resources which is making my GPU the system bottle neck* now with current settings. So my question is what is DCS doing with a nearly 80% increase in CPU usage (NTTR) for ultimately poorer VR performance?

 

Is my DCS install broken? I've run the repair utility no change!

 

And yes I noticed that straight up after the update

 

Is my DCS install broken? I've run the repair utility no change!

 

Is the extra CPU usage part of some multicore framework for the next versions to use, well who would know there is no mention of it. I'd be okay with 2.5.4 if officially it represents a pathway to further improvements but after a lot of tweaking 2.5.3 is looking pretty inviting as far as VR image quality goes.

 

* nearly 12 months ago my old 4Ghz CPU and 980Ti rocked out 90FPS even in Normandy now a 4.7GHz CPU and a 1080Ti are lucky to rock out 45FPS with drop frames even on NTTR with 2.5.4. :cry:

[DflippinK]

Well it's official Wags says no improvements in VR with 2.5.4

 

Fair enough however for me 2.5.4 broke Anisotropic Filtering big time I've tweaked settings to try and get it as good as 2.5.3 VR performance but it's not as good.

 

TBH DCS seems more responsive with 2.5.4 and it appears to be using more CPU resources which is making my GPU the system bottle neck* now with current settings. So my question is what is DCS doing with a nearly 80% increase in CPU usage (NTTR) for ultimately poorer VR performance?

 

Is my DCS install broken? I've run the repair utility no change!

 

And yes I noticed that straight up after the update

 

Is my DCS install broken? I've run the repair utility no change!

 

Is the extra CPU usage part of some multicore framework for the next versions to use, well who would know there is no mention of it. I'd be okay with 2.5.4 if officially it represents a pathway to further improvements but after a lot of tweaking 2.5.3 is looking pretty inviting as far as VR image quality goes.

 

* nearly 12 months ago my old 4Ghz CPU and 980Ti rocked out 90FPS even in Normandy now a 4.7GHz CPU and a 1080Ti are lucky to rock out 45FPS with drop frames even on NTTR with 2.5.4. :cry:

 

 

i dont like that you are seeing additional CPU utilization.. mostly because i am not... but we do have different sets of variables... so who knows... what i can say is, on my 4k 2d setup, my CPU utilization is the same as it was last update... so no change for me in that regard.. i do run with vsync at 60fps... so that will give me quite a bit of consistency comparing CPU utilization between updates, but it may also suggest that the variable you are seeking may be limited to a VR setting or maybe an updated/changed process of drawing frames in a variable framerate environment... I notice your task manager shots are displaying dual graph charts... are you using DSR in nvcp?... also, i noticed that you have 157 processes running.. Have you ruled out any inhibiting processes that may be affecting performance?... i am generally running 95-100 processes, but always keep an eye out and generally disable or set to manual, anything that might affect priority tasks.

[Gryzor]

The stuttering when you destroy a airplane never occours in 2.5.4. That is a big improvement for me, specially in dogfight combats!

[FragBum]

i dont like that you are seeing additional CPU utilization.. mostly because i am not... but we do have different sets of variables... so who knows... what i can say is, on my 4k 2d setup, my CPU utilization is the same as it was last update... so no change for me in that regard.. i do run with vsync at 60fps... so that will give me quite a bit of consistency comparing CPU utilization between updates, but it may also suggest that the variable you are seeking may be limited to a VR setting or maybe an updated/changed process of drawing frames in a variable framerate environment... I notice your task manager shots are displaying dual graph charts... are you using DSR in nvcp?... also, i noticed that you have 157 processes running.. Have you ruled out any inhibiting processes that may be affecting performance?... i am generally running 95-100 processes, but always keep an eye out and generally disable or set to manual, anything that might affect priority tasks.

 

On my way out soon so just quickly if I look at the processes DCS with .y test mission I'm seeing DCS at about 25% prior in 2.5.3 and pretty much 2.5.x DCS used around 16% usage.

 

Usage goes back to normal when I close DCS and is consistent when I start DCS now with 2.5.4.

 

Cheers.

[FragBum]

_cont_

 

Okay DCS performance is about the same maybe better but that could be placebo :doh:

 

are you using DSR in nvcp?

No off, I couldn't see that improve VR in DCS.

 

 

The only thing I have changed which appears to have made no difference to DCS and it's new found appetite was upgrade the nvidia driver which fixed some minor display glitch type events.

 

TM points at DCS prior DCS was lucky to hit 16% maybe 17% but that could just be TM or 100/6.

I also had to run itunes to get total CPU usage to maybe up to 18~19~% CPU usage previously.

 

 

One thing I have noticed is that it is easier to push the GPU to hit the wall. Damn it,..$ :cry:

 

As for the VR image quality I have no idea it's just different, cause or effect,.. :huh:

[DflippinK]

_cont_

 

Okay DCS performance is about the same maybe better but that could be placebo :doh:

 

No off, I couldn't see that improve VR in DCS.

 

 

The only thing I have changed which appears to have made no difference to DCS and it's new found appetite was upgrade the nvidia driver which fixed some minor display glitch type events.

 

TM points at DCS prior DCS was lucky to hit 16% maybe 17% but that could just be TM or 100/6.

I also had to run itunes to get total CPU usage to maybe up to 18~19~% CPU usage previously.

 

 

One thing I have noticed is that it is easier to push the GPU to hit the wall. Damn it,..$ :cry:

 

As for the VR image quality I have no idea it's just different, cause or effect,.. :huh:

 

 

k... gotchya... the reason i was asking about DSR was because it is the only thing that i have ever noticed that sometimes makes the graphs in TM display oddly as they appear in your screengrabs.. I did a little more testing to confirm my CPU utilization is the same as before, and in fact it may be even less.. Again though, you are likely chasing 90fps where i am locked at 60.. regardless, if nothing else has changed on your setup or config, i am a bit weary of what might be demanding an additional 10% of your CPU as compared to 2.5.3xxx... With regards to changes in VR, i suspect, while not particularly mentioned, that we are seeing some of the results of Kegety's shader mod being implemented by ED... this however would only have an effect on GPU utilization and not have any effect on CPU.. might explain some of the image quality differences in VR though.

Overall, after comparing my 2.5.4 results with those of 2.5.3xxx, i would say utilization is approximately the same if not better on my system... was previously running at 8-9% overall CPU and currently running at 6-7%.. negligible really.. but still a bit better than before...

 

[FragBum]

Okay some more basic observational testing with some surprising twists.

 

I have a basic free flight mission for all the maps all tests used the Gazelle only because I'm slightly more confident at returning it to the ground in one piece. :thumbup:

 

All tests are using Oculus Rift and 1080Ti (damn it ordered 2080Ti :doh:)

 

All settings constant between tests in 2.5.4. Comparative notes do refer to 2.5.3 which used slightly different settings but constant for all maps. Due to something changing in 2.5.4 I have had to put AF to off from 8* before the update however I am now able to increase PD to 1.3 up from 1.1 in 2.5.3. Also CPU rarely cracked 17% for DCS.

 

Normandy

GPU always pushing it GPU well into 100% utilization with associated drop frames. this map has been rather hard to fly since 2.5.x changes.

 

GPU ~100% VRAM 7GB

 

CPU Tot 24% DCS 20% (up from 2.5.3)

 

MEM Tot 4.7 DCS 3GB

 

PG This was also a drop frame fest in 2.5.3 but more playable than Normandy, just. Now surprisingly very playable and smooth. :thumbup:

 

GPU 70~80% VRAM 7.9GB ( GPU Usage down from 2.5.3)

 

CPU tot27% DCS 23% (up from 2.5.3)

 

MEM Tot 13.8 DCS 5.4GB

 

Caucasus Was very playable with the odd drop frame however now worse than PG map was??

 

GPU ~75%+ 7GB VRAM (but regularly hits 100% and drop associated frames)

 

CPU Tot 21% DCS 18% (up slightly from 2.5.3)

 

MEM Tot 11GB DCS 3.9GB

 

NTTR This was the best performing map under 2.5.3 using the least resources GPU/CPU. It was rare for this map to actually drop framesrate below 45FPS however now it's maybe an infrequent event certainly more noticeable in 2.5.4.

 

CPU Tot 28% DCS 24% (up from 2.5.3)

 

GPU 70% VRAM 6.86GB

 

RAM Tot 11GB DCS 4GB

 

I'll upload old and new settings soon for comparison as I said this is purely observational it's curious that TM is showing more usage under 2.5.4 again my install might well be broken but seemingly in a good way. I am still vexed at the drop in overall VR image quality however near field seems about the same overall image quality might be missing AF which now looks real fuzzy in VR now. :huh:

[FragBum]

WIP

 

I'll keep this here as a place holder After the DRM process was sorted I installed the RTX2080Ti and well tried it out of course. :thumbup:

 

I'll go over the tests and report back but I can say it's a significant improvement no not 90FPS but with settings the same there is a significant reduction in GPU utilisation across all maps for better performance and a lot more head room.

 

cheers.

 

Normandy

Very flyable now

 

GPU ~65 to 85% VRAM 6.1GB

 

CPU Tot 29% DCS 25% (up from 1080Ti)

 

MEM Tot 9.9 DCS 3GB odd don't think it's because of the GPU DCS same usage??

 

PG

Very playable and smooth, still an odd drop in FPS .

 

GPU 70~80% VRAM 7.3GB ( no real difference here )

 

CPU tot 31% DCS 26% (up a smidge from 1080Ti)

 

MEM Tot 13.7 DCS 5.6GB

 

Caucasus

Was very playable with the 2080Ti

 

GPU 45 to 60%6.3GB VRAM

 

CPU Tot 17% DCS 14% (down slightly from 1080Ti)

 

MEM Tot 10.5GB DCS 3.3GB

 

NTTR

This is again the best performing map under 2.5.4 flies well now.

 

GPU 45 to 60% VRAM 6.1GB (down fom 1080Ti

 

CPU Tot 27% DCS 24%

 

RAM Tot 10.9GB DCS 3.8GB

Edited January 5, 2019 by FragBum