St4rgun

@BIGNEWYIf you are already here, a word about the clouds: hopefully "some optimizations" made in 2.8 already to have smoother experience for the clouds in VR, which is true at least for me. But is it normal that while the close clouds are pretty nice even on Standard the others on the horizon and far away are very low resolution and have heavy aliasing despite of MSAA 2x and nvidia transparent AA 8x? Will be there some kind of development on eliminating ALL of those aliasing artifacts?

Let me make a small addition: it would be MANDATORY, because right now if I sit in an F86 and click "adjust controls" it will default to AH-64 Pilot controls, not even F86. Very annoying.

2.8 is broken.... Simple As I said also in the patch notes thread, for me the new frametime graph of the fps counter can be very spikey normally during a mission but it became super smooth if I start an instant mission with ANY airplane in cold and dark state. Which equals super smooth visuals in VR with the same details as when the simulation runs. It means that the GPU can render the extreme quality in VR without problem, but when the simulation of my own plane starts the CPU is so highly capped on the max. 2 cores that the rendering heavily suffers. Now I can visualize why the devs are working so hard on the multicore development. Actually I became extremely hyped on the hopefully incoming MC news as it seems that switching on the idle cores in my CPU will literally make WONDERS on the rendering smoothness also! Combine this with a future finalized Vulkan API I'm sure that VR will be much better. I really hope that performance wise the current 2.8 implementation is the bottom of the pit. @BIGNEWYCan you say a word about MC and/or Vulkan current status? As earlier said MC is in closed beta testing some time before. Will MC mean 2.9 and after that Vulkan 3.0? Is it possible that these two steps wont need another years?

Is it only news for me, or everybody noticed that the frametime graph can be very spikey normally but it became super smooth if I start an instant mission with ANY airplane in cold and dark state? Which equals super smooth visuals in VR with the same details as when the simulation runs. Now I can visualize why the devs are working so hard on the multicore development. Actually I became extremely hyped on the hopefully incoming MC news as it seems that switching on the idle cores in my CPU will literally make WONDERS!

I understand your argument, but calculating a x2 multiplier as a minimum factor for refresh rate is a bit steep IMHO. I would be really happy if my VR headset can provide a rock solid 90 fps at native resolution by having DCS stable 100 fps all the time, I would not require bigger headroom.

For me also the clouds are heavily aliased in VR on Standard, a bit better at Ultra (but still ugly aliased), but in Ultra the raindrops on the canopy tanks the performance to unaccetable levels. To be honest I can't see ANY improvement on the quality of the clouds in VR. Sometimes I have a strange feeling that devs write a changelog about the high quality clouds in VR then wait for the placebo effect of the players... Or someone made the build from an inappropriate source, I don't know. But the clouds are definitely not better in VR than it was before. I have to correct myself before some community manager does it: NO ONE STATED IN THE CHANGELOG, THAT THE CLOUDS WILL BE BETTER IN VR! Pardon for my ignorance. Exactly this was in the preliminary newsletter: "The Weather System includes optimization for VR, and we are confident that this will result in a more stable and smooth experience.[...]The rain effect has also been enhanced, and there have been fixes to rare volumetric clouds artifacts.". So sorry folks for waiting a globally nicer clouds in VR...

+1

DCS is much older than the current PC VR technology therefore the ancient graphics engine holds back the performance heavily on cutting edge hardware. As I understand right now (and someone please correct me if I'm wrong) the engine renders the stereo frame independently for each eye. so no optimization is in place to render the exact same scene from a slightly different eye position for L and R eyes. This combined with the high resolution of the modern HMDs is a performance disaster. We all hope, that in the near future multicore update will finish the internal beta testing and eventually crawl to OB and after that Vulkan implementation could change this resource hog we have right now. What can be done now is some tweakig with either the DCS settings, using OpenXR with OpenXR ToolKit (ie. fixed foveated rendering) and applying other tricks to have at least stable 45 fps (!!!) on a rig similar of mine at native HMD resolution. This is a shame as the G2 native refresh rate is 90 Hz, but it is at least usable. That's why I really like the new FPS counter and details window: if it will work correctly it can help finding the bottlenecks.

Move your DOT CURSOR over the graph, it will show up the labels. By the way you can use the dot cursor to minimize the windows as well.

For me the huge frame drops happens with clouds set to Ultra, but in Standard it's much better. Unfortunately the clouds are still heavy aliased in Standard.

Quick feedback on Persian Gulf, sunset, Overcast and rain 3 preset, M2000, SP. The new FPS counter is neat. The SIMULATION number is all over the place hence some performance drop. New lighting is great, very nice! Clouds are as heavily aliased as before (or worse) on Standard, while setting it to Ultra is a bit better but still pretty much aliased Rain droplet effect on canopy glass is beautiful in VR, but: Unusable on Ultra clouds because of heavy frame drop, kind of usable on Standard When turned off the droplets in VR the head tracking became broken in rain (moving the image glued to my head movement) Specs are in my signature. I'm using ingame MSAA 2x with nVidia Control Panel AA for transparency 8x supersampling.

The small arrow is at the LEFT from the counter at the top row (the header of the counter window). Actually there're two different windows: one for the counter and graph and the other with the detailed info (if you clicked CTRL+Break two times), both can be minimized. Note however that that triangles can be clicked ONLY with DOT CURSOR in VR, and the blue cross will not switch to dot automatically if you hover it on the window. So the easiest way is to switch to the menu (to stop the game), then the dot cursor can be used to collapse the counter, then go back to game.

The VR optimization was mentioned in the newsletter on 14th October as following: "The Weather System includes optimization for VR, and we are confident that this will result in a more stable and smooth experience."

True, and the last technical word on Vulkan was the development of some kind of hybrid technology to be able to use Vulkan and DirectX together with a special graphics API to test parts of the new technologies. Since then no more details came out as I remember. I'm really curious what VR optimization will be included in 2.8. I hope this is not restricted only to the clouds... I'm sure that after introducing 2.8 ED will (or at least should) have to make a new development report on multicore and Vulkan status as detailed as the previous one was.

While we're waiting let's remember the forgotten celebration of the first birthday of the one-and-only Multicore Development Report, born on 15th October 2021! We have no idea what technologies 2.8 will include, but based on that report a year ago I had high hopes back then to have 3.0 including MC and Vulkan integration by now, not 2.8... Anyway, patience is golden, so waiting for the patch.

I'm using 10700k, 32 GB with a 1070 (!) and a G2 right now with OXR, single player only. More than usable without AA, framerate is 30ish on the ground at airports and 40ish in the air at altitude. But the fps variation is huge according to the modules, it's much worse with Apache, Mi-24 and much better with M-2000C, A-10C II, F-16, F-14 and especially F/A-18. The map also have huge impact, Nevada is the best, then Caucasus. Syria is doable, Marianas is totally unusable.

+1, it's a pain in VR. Serious bug, should be corrected ASAP.

All right, a small supplement for your kind words: https://www.psychologyofgames.com/2019/10/loading-when-were-willing-to-wait/ The essence is this:

Guys, calm down, we have to wait one more month to be able to celebrate the first birthday the one-and-only "Multicore - Development report" to date. In the last year I became patient. But I have high hopes that until this 15th October date we'll have at least one similar (or more datailed) status report on these topics. I'm not even aiming for the deployment, just a report. We'll see.

OK, I tried to reproduce the bug, but in all my attempts for making a landing mission failed, because the AI landed well, without half-sunken front tire. But after then I tried again to fly a quick circle with the same instant action mission as before and the same bug happened: half-sunk front tire, rolling off the runway and crash without braking. This track is smaller, so attached here. Beyond this I have no idea how to help more. If this is not enough I think you are on your own... M2000C_Caucasus_AI landing failure_v2.trk

All right, I'm managed to further test the situation. Unforunatelly upon making a mission for the landing with the AI wingman the AI managed to land correctly all the time without the described bug. BUT the front tire was NOT sunken in the concrete in these cases, so maybe that symptom is the key for the bug. As the TRK file exceeds the 5MB upload limit I'll post it in a two parted ZIP. M2000C_Caucasus_AI landing failure.zip Next part of TRK file... M2000C_Caucasus_AI landing failure.z01

Thanks, Flappie. Is it allowed here to upload a TRK file as 2 parted ZIPs?

M-2000C, singleplayer, Caucasus, Instant Action / Take off mission, One AI wingman. After taking off with the AI wingman I just flew a circle to immediately land to Sukhumi airbase. Before landing I instructed the AI to be in finger four formation, and we landed in formation. After touchdown I rolled down the runway (AI wingmen also does this normally) and at the end turned left for taxi. The AI however rolls down the runway in a straight line, then goes through the fields and hits the inner marker building without a trace of steering or braking. The collission is fatal. Further strange behaviours on AI flying and landing with the M-2000C, which are not fatal: abrupt slowdowns without opening the speedbrake (can be seen at slowing down before final in the turn, I had to use airbrakes and the AI can hold the formation without using them) not using high enough AoA at final (missing the vapor trails at the canard wings for AI airplane) always using parachute for quick slowdown instead of proper airbraking while rolling down the runway after touchdown the front wheels of the AI airplane seem somewhat half sinked into the concrete, maybe that causes the inability of steering at the end The track file is above 5 MB in size (5,8 MB), how shall I upload this? I experienced somewhat similar behaviour with A-10C II on the Nevada map at Nellis when I'm assigned to land on 21L, so I ordered the wingman to anchor at car park while I land, then when he lands he goes for 21R and after touchdown he goes right through the dirt in a straight line to the 21L and taxi place, very weird.

These AI animations are superb, but the actual AI behaviour will be the real thing. I'm afraid that's somehow connected to the multicore development: the AI pathfinding could be perfectly assigned to a separate CPU core, and I think none of them will be implemented without the other. As I know there's not a trace of any ETA for the AI pathfinding, so as for the multicore development. I hope we have the opportunity to see them in this year (if only partially), it would be sad to wait another years for these fine features to deploy.

For me right now is a hard stop after the AH-64D. I promised that no Multicore / Vulkan will mean no more buy for me. So with 20 modules and a TON of money already spent on VR headset and HOTAS in this GPU crysis I have to spare for a new GPU. It's much more important than a new module. If only we could get some info on DCS core progress then maybe I could revise my standpoint...