edmuss
-
Posts
1740 -
Joined
-
Last visited
Content Type
Profiles
Forums
Events
Posts posted by edmuss
-
-
Turn off the debug mode in the vrperfkit settings file by setting the flag to false
-
I believe each headset is different, I think it's to do with DCS not reading in the factory calibration settings for the headset it something along those lines.
The values should be the offsets from centre to move the rendered display I think.
-
Yup same thing I have, I've not tried alt-tabbing as it only takes a few seconds to restart DCS. The actual GPU frametime is visible behind the purple, still down below 16ms but the dropped purple frames are then pulling it back above 30-40ms.
@BIGNEWY has this already been reported? Seemed to start with the last OB patch.-
1
-
-
Does your frametime chart fill up with purple? I have the same, originally I suspected it was lack of VRAM but I think that there is a memory hole somewhere, you can fly once but then have to restart DCS if you want to fly again (even if it's the same empty mission on the same map). I think I've read that it happens if you switch planes as well in MP aswell? Not sure myself as I'm SP only.
edit: I have an existing thread about the problem, not sure if it's been seen by the devs or whether it's been reported earlier.
-
1
-
-
Graphics settings attached, in open sky on Syria I'll be sitting around 12-15ms (85-65 ish fps), down low over Beirut it'll jump from 18-22ms (55-45 fps). A10C II easy instant action it sits between 13 and 18ms unless a CBU crater has gobbled up all the frames!
I run the G2 at 60Hz so the frametimes can go all the way up to 30ms before motion smoothing starts to drop out, I can't reliably hold below the 20ms threshold to run 90Hz and keep the clarity and visual detail that I like - why spend all the money on the headset and GPU and run it looking like a calculator
Note that I also turn MFAA on in the nvcp and it makes a significant improvement to the shimmer for no cost to frametimes, as you run MSAA, turn it on and you can probably drop your PD down to grab some more GPU headroom. Im still toying with terrain shadows, I like them on but it costs another 1-2ms.
-
2
-
-
PD 1.6 on the aero would be 3648 X 4352 per eye which in total is a little under 32 million pixels, the G2 pushes just over 9 million pixels. It's fairly unsurprising that you're not getting high frame rates
This is assuming that steamvr is set to 100%, if it's set higher then you're pushing even more!
Wind the PD back to 1.0 and make sure steamvr is set to 100% as a base, check performance and then wind the resolution up and down as you need.
Best it looks amazing at that resolution though
-
Brand new RTX FE card from nvidia is the cheapest GPU, there are monthly stock drops here in the UK and you'll pay 469 for a 3070 and 649 for a 3080 etc. etc.
I think if you can build it around a 3080FE then you might just be able to squeeze it into budget but then you might be making compromises elsewhere. Have you got all the necessary peripherals (monitor/mouse/keyboard/hotas/rudder etc.) an as that could add a significant cost on as well. Also possibly look at a prebuilt system with a lesser card in (gtx1660ti) and make do till you can grab an RTX FE at msrp and sell the old card on ebay.
-
2 hours ago, darkman222 said:
@edmuss Sorry its miselading what I posted. The screenshot from the task manager is while I am running a stress test on the CPU. Prime95.exe This was only to see how much turbo boost my Ryzen will give.
DCS is on a separate m2 HDD (drive E) and drive D youre seeing is my data drive indexing files at the moment.
Aha, that makes sense!
The CPU will achieve maximum boost when not stress testing all cores as there will be more thermal overhead. My 3600 would only boost to 4ghz normally and temperatures into the high 60s, overclocked it to 4.4 across all cores, reduced voltages and now it's faster and cooler. I am running a thermalright truespirit 140bw with no case so no airflow restrictions.
As mentioned above a cinebench single core benchmark will give you a better indication of what it will boost to. Also I think you'll see no difference for large expenditure of cash to make the hardware upgrade.
My experience is (with G2@60Hz) that CPU times can be higher than GPU and it will kick in motion smoothing but still be smooth as long as the frametimes don't get above 30ms (roughly half refresh rate fps), this is regardless of the GPU pulling 12 or 20ms. I run most settings high so you should be able to do so, try turning the settings up and see if it massively affects the GPU frametimes; I aim to tune to 22ms max GPU time at worst case in SP (empty mission flying low over cities in syria for example) and then when up high frametimes are down to 10-12ms. When mission AI gets involved I have headroom up to 30ms to play with before it gets janky and stutters.
-
On 1/19/2022 at 10:46 AM, darkman222 said:
To the i9-12900KF owners. This is what the task manager shows. Although I know my 5900x is capable to turbo boost up to 5 Mhz it is unlikely to see it.
Is someone around here with a i9-12900KF and can tell me what his task manager shows? Not really sure if that is a feasible way to compare. But you might realize it is really hard for me to decide if I want to invest the money or stick with my 5900 for the next year. Maybe until the new Ryzen 6000 series comes up.
Am I the only person to see that your CPU and HDD are both pegged to 100%? What's on the HDD? Where is windows and DCS installed?
This isn't a DCS issue, if this was whilst playing DCS there's something seriously wrong with your machine. If you're playing DCS I would anticipate no more than 15-20% CPU utilisation in general - 12 cores, DCS uses 2. Look at what it taking up the CPU cycles because anything taking them up will then pull your GPU back down.Also CPU does seem very hot, not too sure on the 5900x but my 3600 overclocked to 4.4ghz never tops 60°C. I would look into that, heat causes the CPU to throttle and will not hit the boost speeds that it needs.
-
You might be able to build a box for 1500 but you'll be limited on the GPU front.
Essentials to look for: -
CPU with fastest core speeds, currently DCS is mostly single threaded so the faster the better. This will hopefully change soon as ED are working on the multicore engine upgrade. Make sure whatever you get is overclockable.
32gb ram, faster should be better if built into a balanced system.
All thing's Windows and DCS installed on SSDs.
For DCS VR you want the biggest GPU you can afford, it well never be big enough so you will have to make a compromise.
With a lot of tweaking I can get good stable frametimes of 12-24ms with my 3070 running a g2 @100% - 60hz.
In the current market I would look to get a 3070fe and then you should be able to build a reasonable machine for the budget.
-
I have started using the low bit depth A10C normal maps that @Taz1004 did and I suspect that has recovered a bunch of vram usage, but like clockwork if I start up a subsequent flight without restarting DCS I get a wall of purple in the frame graph. Switching between high and medium textures doesn't seem to make a huge difference to the actual frametimes.
-
I think I've come to realise that this is the vram memory leak bug. If I load a mission I can run textures on high all round and it's fine, if I reload the same mission it then drops all of the frames. Restart DCS and it loads up just fine.
Anything I can provide to the devs to try to help support this?
-
I had a quick and dirty test with the simple canopy glass and I agree, it costs about 0.5-1ms but obviously with the baked in reflections obscuring everything in certain lighting conditions.
-
I use steamvr SS 100% with PD 1.0, fsr @0.7, MSAA x2 + MFAA with DCS settings mostly high. Using the Clearwater version of kegetys mod and taz's experimental low bit depth normal maps for the A10C.
To be honest I could turn MSAA off and be fine with it but it's a lot nicer flying over rooftops with it on.
-
I find that at 100% on the G2 it helps a lot with shimmer over MSAA 2x, makes it worthwhile running MSAA at all for me.
I do run openvrfsr which I think adds extra shimmer but this helps quell it down.
-
MFAA effectively boosts the quality of the anti aliasing, almost as good as the next rank up so 2x MSAA + MFAA is almost as good as 4x MSAA for little to no performance hit. As long as you had MSAA enabled in DCS for your testing, I would say that performance impact is indeed nill.
The difference is a visual improvement for the same settings meaning that you can drop a rank of MSAA, retain visual quality and gain FPS back
-
I guess it should 'only' be a case of mapping a 3D coordinate to each control node point out similar. Probably completely wrong with that guess though
At least the toolkit is there, with the recent improvements to leap motion within DCS they will hopefully be more uptake and increased demand to get the 3rd parties moving on it. I only own ED modules (which I don't have enough time to learn properly!), so chances of me using others in the near future is remote!
-
Good stuff, I like the terrain shadows so I leave them in and take the frametime hit.
If I can run over Beirut rooftops at 22ms peak then I'm happy, running at 60hz I still have bags of GPU overhead (up to 30ms before motion smoothing drops out). If I go skyward then frametimes are down to 13-15ms on the same settings.
-
6 minutes ago, tomeye said:
Where do you enable MFAA, in nvidia control panel maybe?
Yes, MFAA only works if MSAA is applied to the output. Turning MFAA on if you don't run MSAA in dcs does nothing.
-
Just now, Fiztex said:
I don't think MFAA works with DCS, but will give it a try.
It does, and very well indeed
Beirut rooftops with mostly high settings 2x MSAA+MFAA is as good as 4x MSAA but frametimes are 16-18 rather than 22-26.
-
48 minutes ago, Fiztex said:
Did some testing and simple canopy = 0 doesn't seem like a good solution, it eats around 10% of GPU although indeed solves reflection problem - if I use it I can't do PD 1.1 and I'd very much prefer to as it looks great with msaa 4x. Also didn't notice difference between #define ENABLE_DIFFUSE_SS 4 and 2, 2 is good enough for full res Quest 2 though. Will try to find a way to kill responsible texture and see how it goes.
I'll have a play with the simple canopy on/off and see if I can notice a difference, it may be that there is performance gain but I just never noticed it with all the other changes I've made. I'm certainly not as thorough as @speed-of-heat at testing and recording!
If you're using Nvidia, enable MFAA and knock MSAA down to 2x, big performance jump there
-
15 minutes ago, speed-of-heat said:
so
#define ENABLE_DIFFUSE_SS 4?
Yup that's the one, seems to clean up the cockpit text a little.
-
1
-
-
Simple canopy = 0
and
albedo supersampling = 4 (edit: whatever the maximum value is)
are my immediate go to changes on the various versions of kegetys mod. I simply cannot get 3Dmigoto to work along side it* and I personally see better quality and performance with Clearwater.
*The albedo supersampling gives better text clarity for me than the HUD/MFD sharpening.
-
1
-
-
32 minutes ago, speed-of-heat said:
so
#define ENABLE_SIMPLE_CANOPY 0?
17 minutes ago, Fiztex said:Cockpit glass and HUD I solved by liveries and some textures from mods, but how to kill reflections from instruments on a10c and Ka-50? Will backup my fxo and metashaders and give it a try later today.
Yup
That gets rid of the static baked in reflections on the A10 and KA50 for me, both from HUD, canopy and guages/MFDs
-
1
-
Poor VR Performance unless game is restarted.
in Virtual Reality
Posted
I've been testing the extended alt tab till the frametime graph settles back down and it seems to be a quicker fix than restarting; doesn't always work though.
I think I have all smoothing options set to auto/enabled.
Still certain that's it's a big within DCS.