Jump to content

METEOP

Members
 View Profile  See their activity

  • Posts

    222

  • Joined

  • Last visited

 Content Type 

Posts posted by METEOP

    2.5 TRACKIR STUTTERING

    in Game Performance Bugs

    Posted

    i5 6600k is 4 cores 4 threads, there is no hyperthreading in it.

    According to this thread : https://forums.eagle.ru/showthread.php?t=157374 , and to personnal tests, you loose performances in DCS (20% mesured in quoted thread with DCS 1.5.x) if you set affinity to 2 cores instead of all the 4 physical cores.

    I'd find it strange if you really have better experience with 2 cores affinity.

     

    I stand corrected! I also exchanged a pm with Goa about this I had a misconception about my processor.

     

    I guess what I am trying to point out is with my particular situation, the stuttering seemed purely a tracking issue when fps != vsync, and that monkeying around with affinity and priority did produce drastic improvements on my end!

     

    Now! Armed with the infinite wisdom that I possess indeed four physical cores it gives me a few ideas on how I can make it even smoother.

     

    I have always felt that departing from 2.2 and going to 2.5b, there is something that my system doesn't like very much.

    2.5 TRACKIR STUTTERING

    in Game Performance Bugs

    Posted

    Goa, I have had a similar problem to you since 2.5b

     

    I will take the time to write a longer post in the near future, but here is the short version.

     

    After many tests, and many disappointments, I started to pin down the correct symptoms:

     

    - Stuttering when framerate goes below 60hz (vsync'ed)

    - Framerate goes below 60hz when using sensors (A10C TGP and MAV) __with lots of speedtrees__

    - Framerate goes below 60Hz when lots of enemies are aquiring (me) and firing

     

    No sync is not an option, did not appear perfectly smooth like gsync under any circumstances.

     

    My problem is aggravated by the fact i'm not running in fullscreen, I need to extend to a second monitor for a glass cockpit (Ikarus). It is also aggravated by the two exports on that second monitor for both MFCD. Another factor that aggravates the problem is that I am running in nvidia resolution factors (2x factor of 1080p but no antialiasing).

     

    I start flying, very smooth. Get to IP, still smooth, fence in, turn on TGP and MAV, mild stuttering, enemies aquiring, medium to severe stuttering (about 50fps at that moment). At this point it was pretty much unplayable.

     

    After month of tinkering with everything I could, it started to dawn on me that at 50fps, I should still be able to get a pretty playable experience (remember the 15-20 fps days :glare:) but the stuttering was making it unplayable.

     

    One time, I was about to unleash mav hell to a sorry tank company when suddenly, everything went buttery smooth! I looked at the corner with the framerate, it was about 48-52fps. Launched mavs, went to circle pattern to look at the kill and then I realized my headtracker battery was dead cause it was not panning anymore.

     

    So I came to the realization that the stuttering was not 'in the game', it seemed to be caused by some weird timing problem with the TIR STUTTERING when going under refresh rate. With the correct symptom at hand, I started looking for alternative answers. I found it in Process Lasso.

     

    The biggest positive impact was to move the processes around by core affinity and elevating process priority for the Opentrack processes. Here are my Process Lasso configs by process :

     

    I have 16Gb RAM and an i5, so two physical cores (0 and 2) and two hyperthread cores (1 and 3)

     

    DCS

    - affinity 2,3

    - priority class: above normal

    - High performance process option ON

     

    Ikarus

    - affinity 0,1

    - priority class: below normal (would sometimes hang DCS if higher)

     

    Opentrack.exe and TrackIR.exe

    - affinity 1,0

    - priority class: High

    - High performance process option ON

     

    Realtime priority with Opentrack and Tir processes was even better but if the process hangs the whole computer is hung by that process and will not come back. So I dialed it back down to High.

     

    Goa, I would be curious to know what is the result from trying this on your end.

    Export.lua massive performance cost

    in Game Performance Bugs

    Posted

    Thread back from the dead! I think I have the same issue, i'm using Ikarus for the A10C and reverting back to Helios for Su27-33 since I just bought FC3.

     

    When I switched from 2.1 OA to 2.5B I noticed a big performance hit. Thought it was something else until I messed up my export.lua path and things suddenly were so smooth it gave me a vertigo!

     

    Not sure if there is anything that can be done, but I'll try confirming nonetheles...

    How much time do you have to play

    in Chit-Chat

    Posted

    With two young children and a consulting business that wont quit, I'm actually training myself to sleep less so I can get some wing time.

     

    Still I manage to get about 2-4 hours a week in the pit.

     

    And ocasionally with my little guy on my lap we do a quick 30 mins instant action bit bombing session.

     

    Yeah.. I'll be proficient in about a decade or two at this pace.

    2.5 runs smooth until i start bombing stuff

    in Game Performance Bugs

    Posted

    Hi METEOP. I had a severe fps problem with my GTX980 and I fixed it pretty easily. Have you checked that your GFX card is using the full x16 speed of your PCIe bus already? If you haven't, get GPU-Z and run it once. It will only take you a minute.

     

    My rig is less powerful than yours, by the way (i5 3750K, 16GB RAM, GTX980).

     

    Excellent suggestion, however it does not seem a problem on my system.

     

    According to CPU-Z I'm using the full 16x lanes @ PCIe 3.3

     

    I would also thing that I would have issues with other games if it was a problem and that is not the case...

     

    At this point I think I will just wait for an update and see what happens. Thanks all for your help!

    2.5 runs smooth until i start bombing stuff

    in Game Performance Bugs

    Posted

    Give this a try: set your "resolution of cockpit displays" to 512 or 1024 every frame

     

    https://forums.eagle.ru/showthread.php?t=202541

     

    https://forums.eagle.ru/showthread.php?t=199880

     

    Maybe an issue with frame timing, wich can cause heavy stuttering, that did the trick for some of my friends having the same issue...

     

    Thanks for your suggestion, I did try all these options and could not notice any difference.

     

    So, reading from all your suggestion, it seems Nobody else has this problem?

     

    Maybe I am crazy after all :megalol:

    DCS World Stutter during gameplay

    in Game Performance Bugs

    Posted

    Just for testing to see if there was any software/driver conflicts, I temporarily installed a clean Windows 10. +nvidia drivers + ms visual c++ to get the game to work.

     

    Still stuttering. This time with fps drops to half the usual rate after 5 minutes or so of flying AND with micro-stutters along with the usual larger stuttering.

     

    Although our stuttering issues seem ta have different causes, check my other thread at:

     

    https://forums.eagle.ru/showthread.php?t=201162

     

    I have narrowed down my issue to TGP and NAV MFCD usage. There is a switch in graphics.lua which you might want to play with, it's the maxfps which by default is 180.

     

    Maybe briging it closer to 60 fps would balance your stuttering issue?

    2.5 runs smooth until i start bombing stuff

    in Game Performance Bugs

    Posted

    Sorry about the delay and for bringing this back up again.

     

    After a good while spent testing here is some more info on the stuttering issue, I'm still on the fence as to reporting it as a bug or waiting it out.

     

    While it appears stuttering in 2.5OB can be caused by a number of issues for different pilots, I've narrowed it down (on my system at least) to electro-optical sensor usage, specifically the MAV targeting screen, and somewhat lesser of a problem is the TGP screen in the MFCD also.

     

    TO RECREATE THE PROBLEM:

     

    - Any map and mission will do, I'm running the A-10C, not sure about other airframes. Smoke compound the problem so you might take an instant action mission with smoke markers.

    - Line up for attack on ground target, get low about 3K-5K AGL

    - Turn on TGP on one MFCD, MAV on the other

    - At this point the stuttering should be pretty severe

    - Attack target, launch a MAV, gun down the target to a burning crisp

    - While overflying through smoke and explosions, the stuttering should be very bad

     

    ANALYSIS:

     

    I tried many benchmarking applications and finally I found the bottleneck appears in the GPU. Using nvidiainspector, I was able to point out the stuttering occurs when the GPU is saturated (100%).

     

    When flying without using the EO sensors, under any condition the GPU is running smoothly at around 60-70% usage so there is lots of overhead available.

     

    However, turning on the EO sensors, especially the MAV, will eat up up to MORE THAN 30% OF A 1070GTX GPU under any combination of graphic setting.

     

    This was not an issue in 2.2A all other things being the same.

     

    I've further narrowed down the conditions that causes the issue : THE TREES.

     

    If the sensor is not looking at trees, then the GPU usage for the sensor is about 5%, more than reasonable! It's when the sensor is looking at trees it will just eat the GPU for breakfast.

     

    So.... how to make this _playable_ ? Here's what I did, please chime in if you have some more ideas.

     

    1. Graphic settings

     

    Here are my current graphic settings, anything else will either not affect the performance of will degrade it severely. Note that I am runing in 1080p with a second 1080p monitor for my glass cockpit.

     

     

     

    ["graphics"] = {

    ["DOF"] = 0,

    ["HDR"] = 0,

    ["LensEffects"] = 0,

    ["MSAA"] = 2,

    ["SSAO"] = 0,

    ["anisotropy"] = 4,

    ["aspect"] = 1.7777777777778,

    ["chimneySmokeDensity"] = 0,

    ["civTraffic"] = "high",

    ["clouds"] = 1,

    ["clutterMaxDistance"] = 1200,

    ["cockpitGI"] = 1,

    ["disableAero"] = true,

    ["effects"] = 3,

    ["flatTerrainShadows"] = 0,

    ["forestDistanceFactor"] = 0.8,

    ["fullScreen"] = false,

    ["heatBlr"] = 0,

    ["height"] = 2160,

    ["lights"] = 2,

    ["motionBlur"] = 0,

    ["multiMonitorSetup"] = "ikarus-a-10c",

    ["outputGamma"] = 2.2,

    ["preloadRadius"] = 120100,

    ["scaleGui"] = false,

    ["shadowTree"] = false,

    ["shadows"] = 1,

    ["sync"] = false,

    ["terrainTextures"] = "max",

    ["textures"] = 2,

    ["treesVisibility"] = 17250,

    ["useDeferredShading"] = 1,

    ["visibRange"] = "High",

    ["water"] = 2,

    ["width"] = 1920,

    },

     

     

     

    Yes I use MFCD and RWR exports to an Ikarus touchscreen cockpit, but the problem appears even without them. So I did try with no exports, only one fullscreen etc... same problem.

     

    2. graphics.lua

     

    .../DCS/Config/graphics.lua

     

    maxfps = 64

    sync = true

     

    I've found the maxfps trick in an old 1.x thread. If I put it under 60 the tracking is not so smooth, so I'm really wondering if some frames are not being dropped that are not detected by FPS meters... 64 fps is a good tradeoff for me between being smooth without sensors and playable with sensors.

     

    The thinking behind this is, it seemed to me that even though the mainscreen would drop frames, the exported MFCD would run very smooth. This appeared odd to me, since i'd rather get 10 fps in the sensor than hurting the main view... So I started looking for ways to limit the max fps of the MFCDs.

     

    I've found by default the maxfps is 180, by reducing the maxfps I am putting a hard limit on what the sensors can hog out of my GPU in worst case. Too bad I have not found a way to limit only the MFCD fps...

     

    Oddly enough, if I put some VSYNC either in DCS or the Nvidia control panel after a few minutes of flying the fps degrades severely and it's unplayable, but the sync=true switch in the config file seemed to make the image mnore stable without causing any problems.

     

    3. MFCD mip filtering

     

    .../DCS/Mods/aircraft/A-10C/Cockpit/Scripts/MFCD/indicator/MFCD_useful_definitions.lua

     

    mip_filter_used = false

     

     

    Another optimisation I've discovered is disabling the mipmapping filter for the MFCDs. Turning it off seems to make thing better.

     

     

    RECOMMENDATIONS:

     

    I see a few options availlable to the developpers.

     

    The easiest I would think would be to limit the MFCD max FPS, or at least make them less of a priority than the main views.

     

    Another would be the speedtree algorithm, it is quite possible that the sensors for the A-10C need to be optimized to take this new algorithm into account.

     

    I sincerely hope this will become a priority in the near future, since in it current state and without the aforementionned optimisations, it is alas UNPLAYABLE for now. Even with the optimisations it is barely OK.

     

    Furthermore, I would really like to use my 30% GPU overhead that I keep free for sensor usage. The base engine is incredibly fast and beautifull so cranking up the graphics would be even better!

     

    I would like some people to test these optimisations on their end and let me know what you think about filing an official bug report?

    running dcs on virtual box

    in Multiplayer Server Administration

    Posted

    Xen server is also open source and runs somethin like 75% of vm on the internet

     

    I've used virtualbox, vmware and xen, hyperv (microsoft) never tried it.

     

    just my opinion... Attempting a gpu passthrough would not be easy and I would not waste my time on it.

     

    But hey! If you manage to pull it off it would be quite a feat! I use virtualisation a LOT for my work and at home, mostly with xen now.

    DCS World Stutter during gameplay

    in Game Performance Bugs

    Posted

    I'm having stuttering no matter the settings.

    High, med, or low. VR or without VR. msaa on or off. High or balanced power performance. It all stutters. This while getting 150 - 180 fps.

     

    I'm testing it on only 1 monitor with a gtx 1080ti. Installed on SSD. Even tested it on my other computer with a gtx1070 still stutters. Tried nvidia drivers 388.71 as well as 390.77.

     

    Did everything suggested on this forum. Deleted shader cache, turned off C1e halt state... I give up.

     

    Don't give up!!!

     

    I'm not exactly sure precisely what did it but it's solved now for me.

     

    Is your stutter problem when you use electro optical sensors i.e. TAD or MAVERICKS?

    We need to have a chat about memory...

    in Game Performance Bugs

    Posted

    I replied to your other post also.

     

    We did not see these issues during our testing, this is why larger testing with a open beta is important.

     

    Hey BIGNEWY thanks for chiming in, I'm certain that ED people are looking at the threads so I'm trying to be as constructive as I can.

     

    I've seen in another thread that upping your pagefile to 16Gb solves the MP issue for ram usage. Sorry I cannot say for myself as I don't MP, yet.

     

    2.5 is drop dead gorgeous for sure! I just wish I could bomb something without there being major stutters.

    We need to have a chat about memory...

    in Game Performance Bugs

    Posted

    I tried to test the memory leak in DCS (singleplayer).

    Idk if it can help finding the problem :thumbup:

    Here is what I found:

     

    RAM USAGE REPORTED FROM THE DCS BUILT-IN FPS counter ETC... (Ctrl+pause)

    It should contain both the real RAM usage and the paging files usage

     

    -Main page (default): 1.6 Gb

    -Inside the mission editor (no units at all, first join): more or less 6 Gb

     

    Here is the tricky part::music_whistling:

     

    I begin looking at the map with different zooms and moving trying to "save" every corner of the map;

     

    -Ram usage after that procedure: 8Gb

     

    Then I fly my empy mission, hopping from an airfield to another one

    -Ram usage: 10 to 12 Gb

     

    So I quit the mission, and memory usage stays the same in the main page. After a restart of the client it returns to 1.6Gb.

     

    RAM USAGE REPORTED BY WIN TASK MANAGER (only the DCS.exe process)

     

    -Main page (default): 0.8Gb

    -Inside the mission editor (no units at all, first join): 2.6 Gb

    -After the "save" of the map: 3.6 Gb

    -After moving the camera a bit in the empty mission: 5.4Gb

    -When leaving the mission editor and going back to main page: 5.1Gb

    -After client restart: 0.8Gb

     

    I have 16Gb of ram DDR4 and 22Gb of paging.

     

    Settings:

    BuW5NWd.jpg

    PC in sig

     

    Interesting, we almost have the same computer, do you have stuttering when turning on the tgp and making low passes over enemies?

    2.5 runs smooth until i start bombing stuff

    in Game Performance Bugs

    Posted

    I'm with you on this.

     

    I'm firmly in the camp that suspects that many of the familiar stutters that have plagued ED games (since back in the LOMAC days!) are related to the sound engine.

     

    Rockets hitting the ground, bombs exploding, an aircraft being hit, contacting ATC, ejecting, crashing....all these events, at least the first time you encounter them during a session, cause DCS to stutter. They have at least one common denominator - they trigger a sound file.

     

    Nope that is not the issue, it's turning on the TGP that does the most stuttering, way before pulling the trigger or anything sound related...

×
×
  • Create New...