Welcome to Keen Software House Forums! Log in or Sign up to interact with the KSH community.
  1. Hello Guest!
    Welcome to the Bug Report forum, please make sure you search for your problem before posting here. If you post a duplicate (that you post the same issue while other people have already done that before) you will be given a warning point which can eventually lead into account limitations !

    Here you can find a guide on how to post a good bug report thread.
    Space Engineers version --- Medieval Engineers version
  2. You are currently browsing our forum as a guest. Create your own forum account to access all forum functionality.

[1.120] Abnormal GPU Usage

Discussion in 'Bug Reports' started by w4stedspace, Feb 5, 2016.

Thread Status:
This last post in this thread was made more than 31 days old.
  1. w4stedspace Trainee Engineer

    Messages:
    43
    Tested on the following cards at 1920x1080 resulting in the following temps:

    980ti Strix - 76c
    980ti Reference - 82c
    Titan Black - 85c (Thermal throttling)
    970 gtx - 80c (Thermal throttling)

    All cards immediately hit 100% VRAM usage on world load, then head swiftly to 100% GPU core usage also. The cards with poorer cooling then begin to thermal throttle. Take into account 3 of the above cards have 6gb of Vram and are about as top spec as you can find at the moment.. seeing 100% usage with 1080p res is highly abnormal. The issue can be contained somewhat by enabling v-sync assuming your card is running faster than 60fps, thus introducing an FPS cap to limit card usage.

    I'd be concerned that people unaware of checking GPU loads and temps will be causing some pretty extreme heat in their machines.
     
    • Informative Informative x 3
    • Agree Agree x 2
  2. Shabazza Junior Engineer

    Messages:
    689
    V-Sync should be standard annyway as no one profits from > 60 FPS besides of 3D monitor users with suiting footage.
    Without V-Sync on a top notch GPU you are basicly heating the room for nothing.
    But 100% VRAM usage is probably not normal. Are you sure it's showing you actually USED VRAM or is it just "reserved" VRAM?
    Windows for example also does not free unneeded RAM immediately.
    But it will free a part of this "reserved" RAM chunk whenever it's needed for an operation.
     
  3. w4stedspace Trainee Engineer

    Messages:
    43
    V-sync is know to introduce input lag, and having a higher fps than 60 also improves game responsiveness and input lag.. but that is a moot point and nothing to do with this.

    This is dramatically changed behaviour than even with v-sync enabled is capable of thermal throttling cards upto and including 970gtx's and Titan Blacks.

    I was running at 90fps previously using ~2gb Vram and about 30% core load, now it's 60fps with 6gb vram usage and close to 100% core load.
     
    • Informative Informative x 2
  4. LocInt Apprentice Engineer

    Messages:
    264
    Just buy better cooler lol.
     
    • Disagree Disagree x 9
  5. Tig Trainee Engineer

    Messages:
    51
  6. LeonserGT Apprentice Engineer

    Messages:
    459
    After recent update my gameplay last only 5-10 mins: after it I got image-freeze, but still hear game's working - shooting, interacting, moving. But video part is froze forever, only ctrl + alt + del and closing program will help. Does it have same origin - VRAM leak?
     
  7. Visentinel Junior Engineer

    Messages:
    843
    First of all, if you don't know what we are talking about due to lack of technical knowledge go away, useless comments like buy a better cooler are useless and just serve to embarrass yourself.

    I have noticed some strange GPU activity as well in the recent builds, in most cases the GPU was lucky to pass 2gb of vram usage but as of recent builds the game is filling the vram and maxing out the gpu core for what seems like no good reason, it is also introducing bottlenecks on the GPU causing annoying drops of fps.
    I have also noticed that weird wave moving across the voxel cells =\

    Vsync is a pretty important thing these days to avoid frame tearing, the input lag caused by vsync is due to frame pre-render what you need to do is set max pre-rendered frames to 1 :)
     
  8. Tig Trainee Engineer

    Messages:
    51
    My GPU usage in tbis game looks like a square wave of 100 and zero. Very strange. And I wonder if tbe frame limiting is showing a side effect of making the voxels ripple like that as the viewpoint shifts while standing there and the character is bobbing.

    I posted in the other thread my FPS drop was an issue with my power plan. With my motherboard and the i7-990x, I have to use Turbo to overclock. I have 2 power plans in Windows Power Options. One has processor management 10-100% (power saver) and the other is 100-100% ( high performance). After a gaming session, I switch to power saver and let it cool and it sleeps. Well I had left that on and it seems the game is not enough for it to Turbo boost the multiplier up to 30x and 4GHz. It was sitting way down in power mode and like 1.6GHz. 12x and not moving.

    I can switch power plans and watch the framerate drastically change...but the GPU is still going doing it's 0-100-0-100 usage squarewave. As soon as the radio button for the power plan is chosen, the framerate changes. You can toggle back and forth and change it each time.

    The waving/rippling voxel appearence is still there with it on the higher plan.

    Other thread post
    http://forum.keenswh.com/threads/update-01-120-highlights-bug-fixes.7379406/page-8#post-1286935376
     
    Last edited: Feb 6, 2016
  9. Visentinel Junior Engineer

    Messages:
    843
    No that doesnt seem normal, i can way iv seen glitches but your gpu square wave issue and turbo issues would have to be an issue with your pc specifically
     
  10. Kurazarrh Apprentice Engineer

    Messages:
    210
    W4sted, are you still seeing this issue after the hotfix? I'm running a GTX 970 Strix, and I'm seeing GPU usage around 70% most of the time, and temp is holding steady at 62C. The VRAM usage being at or around 100% I don't find unusual. It may not even be using it, just reserving it.
     
    • Informative Informative x 1
  11. Bibber Trainee Engineer

    Messages:
    75
    I have a 960M with 4 GB. Memory usage is ok but on planets I always have a constant GPU load of 100% ... WTF. That can't be normal.
     
  12. w4stedspace Trainee Engineer

    Messages:
    43
    This is what it looks like as of this morning, maximuum column most relevent as I have to tab out to screenshot giving it chance to back off a little.

    [​IMG]
     
    • Informative Informative x 1
  13. Kurazarrh Apprentice Engineer

    Messages:
    210
    You know what, I didn't think about this, but Bibber's post made me consider... I ran GPUTemp only while I was in space, and hardware performance/load changes drastically between space and planetary play. I may not get to it today, but I'll take some more measurements when I can, including from on-planet exploration.

    Also, W4sted, I notice your GPU temp max is now 66C rather than the higher numbers in your OP. Are you still seeing the GPU Core Load pegged at 100% all the time, or is it fluctuating more normally now? After my post last night, I kept playing, and my Core Load rose to around 82% but still averaged around 70%. Never saw it higher than 91%, and then only when I was alt-tabbing in or out of the game (I kept the GPUTemp window visible on a second monitor).

    Also, it may be helpful to know what graphics settings, both internal and external, we're each running. Maybe if we can find differences in our settings, we might be able to flush out exactly where your GPU load bug is coming from. If you think it'd be a useful metric, I'll post my settings when I have a chance.
     
    • Informative Informative x 1
  14. Tig Trainee Engineer

    Messages:
    51
    Are you running MSI Afterburner as a display/logging tool? You can add items to an OSD and also graph over time. Just set up the graphs the way you want them and ignore/minimize the overclocking part.

    On screen in the top left I have

    GPU load, GPU Temp, fan speed %
    Framerate
    GPU Memory Used
    CPU %, CPU Temp

    I leave the main Afterburner graphing app open on the second monitor along with the Radeon overclocking screen with the gauges. And a few gadgets for GPU, CPU usage.
     
  15. w4stedspace Trainee Engineer

    Messages:
    43
    The lower temps are because I've adjusted my fan curves to keep the temps more under control at the cost of it being damn noisy now :) Still seeing GPU load pegged though despite achieving 60fps (vsync on) without much issue.

    Have a bunch of logging tools, I personally likke HWinfo because it tracks everything in one place and often shows info other programs lacks, but I do have some other options for tracking graphs if needed.

    Settings for me right now are maxed, 1920x1200, with the exception of grass density being on 2.0 and using FXAA instead of MSAA becuase in SE it just seems to look better.
     
  16. Commander Rotal Master Engineer

    Messages:
    4,979
    Yo @w4stedspace, it may be a nice touch to include a log via pastebin. Not sure if that'd help too much but it's worth giving it a try. Maybe they're overtaxing some card somewhere, i don't know. Coding is all voodoo to me.
     
  17. Tig Trainee Engineer

    Messages:
    51
    Here's some graphed history. This is general play.


    At the end I fly around with jetpack for longer periods of time. Look at the core clock drop while flying
    (flat framerate is me tabbing out)

    [​IMG]


    Mining with ship and 6 drills. Look at low Core clock entire time and framerate. Was around 11-13 fps
    I did notice they have stopped the other cores and threads from going crazy while mining. Maybe too much now

    [​IMG]
     
  18. tyriael_soban Apprentice Engineer

    Messages:
    113
    I play in 1600x900 with most of my settings smack-dab in the middle, save for ansitropic filtering which i have turned off (despite the nVidia recommendation that i turn it to 4x) on an 870m, i'm also seeing this 99% load with roughly 81C temp.
    I should note that this isn't something SE hasn't already done to this laptop's gpu, a few weeks ago before they really started hammering home on the optimization of planets, i was barely making 20fps and sometimes my sim would drop to .4, with all my cpu cores almost loaded out and the gpu at maximum.

    [​IMG]
     
  19. w4stedspace Trainee Engineer

    Messages:
    43
    Here's the results from a few minutes running on my 2nd machine. Specs = 2600k, 770gtx, 32gb ram. Settings identical to my main machine.

    Again, maximum line is the important one.. same behaviour, 100% core load and very high vmem usage. To be clear, this machine has 4 NF-F12's for the case and 2x 240mm rad's (also wiith nf-f12's) cooling the gpu and cpu, highly overkill cooling setup and this is a machine that lives in another room so can be very noisy - most people would be seeing 70-80c easily at this load.

    [​IMG]
     
    Last edited: Feb 11, 2016
  20. Tig Trainee Engineer

    Messages:
    51
    I think you should be more interested in the averages. A single spike/event can set the maximum percentage.

    On loading the game, it should load up almost the entire amount of vmem right from the get go and also spike the load. My memory ramps up quickly to near max just before it finally loads, as it should, and stays there.

    Get in game, then reset the numbers and see what it does. Screen cap it before exiting.
    Temps look good over time and load is only 41% average over time.

    Watch it on a graphing history.
     
  21. Tig Trainee Engineer

    Messages:
    51
    Notebooks are tough to keep cool. Especially running them hard.


    Check your gaming power plan and see that it has the following settings. Use the High performance one.

    Active cooling ( cools when it sees temp change before thermal throttling )

    Processor power management:
    CPU min percent 100%
    CPU Max percent 100%

    If you haven't tried that 100/100 before, you should see a FPS increase.


    Can you set the GPU fan manually using a utility? Have you tried that? I didn't see fan RPM or percentage in your screencap.
     
  22. Kurazarrh Apprentice Engineer

    Messages:
    210
    I dunno, W4sted, that secondary machine looks like it would probably run at healthy temps even without the fan tweaked to run at over 1900RPM (!!!). With the temp at 54C, I suspect you'd see temperatures more similar to Tig's and mine, in the low to mid 60s.

    Your super-high temperatures are really strange, though. I'd have called it a fluke except for your OP where you mentioned several people are experiencing similar temperatures. Could it possibly be an issue with ambient temperature? You don't keep your house at 35C in winter or something, do you? :p

    I should have some time tonight to get some better readings. I'll see what I can do to try to force the game to run in true fullscreen. I've been having trouble with my dual monitors--it only wants to run in fullscreen window. If I can get it to run in true fullscreen, I'll use HWInfo to get a better idea of temperatures over time.

    Tig, does MSI Afterburner work with non-MSI products? If it does, those graphs would be mighty sexy to monitor, but I'm using a Gigabyte mobo and ASUS Strix GPU...

    [Edit] Healthy temperatures, not speeds.
     
    Last edited: Feb 11, 2016
  23. Tig Trainee Engineer

    Messages:
    51
    Yes. I'm using it to monitor my Gigabyte R9 380x card. It loads a main screen for overclocking, which also works, but the sensors are configurable under settings and displayed in a second window. It also has an OSD option for any of the items.

    Open the second window and then minimize the main window. It also remembers over reboots that you just wanted the graphing and reloads it where you had it with the main screen minimized. I have it running on the second monitor.

    Edit: My ambient temp currently is 72/22 and normally 78/25. ( I'm in Florida, my AC is set on 77 lol and heat on 72 )
     
    Last edited: Feb 11, 2016
  24. w4stedspace Trainee Engineer

    Messages:
    43
    Hehe, maybe yeah... I just cranked things because I saw how hiigh it was getting originally. You can tell the ambient temp by the minimum value on the 2nd machine, as that was only turned on for that test... aka, cold enough, about 20-21c.

    The averages are kinda false also, the program was open for a total of 7mins during which only about 3mins were spent actually in SE, was just to grab some quiick values.

    I'm doing some building atm, i'll get both machines on and graph it for you so it's without doubt :)
     
    Last edited: Feb 11, 2016
  25. w4stedspace Trainee Engineer

    Messages:
    43
    Both machines in same world, with a bit of logging from prior to world load.

    Main (980ti, 6gb VRam) also note mem usage.
    [​IMG]

    2nd machine (770gtx, 2gb VRam):
    VRam was also maxed
    [​IMG]


    Interesting to note, the peaky section you can see on my main rig was when I was flying around with jetpack, just like the poster above. However, flying on my 2nd machine doesn't make a diifference, maybe due to the dramatically lower vram?
     
    Last edited: Feb 11, 2016
  26. w4stedspace Trainee Engineer

    Messages:
    43
    Not sure what to include a log of tbh, just seems to happen whenever planets are invovled
     
  27. Tig Trainee Engineer

    Messages:
    51
    I think the memory usage is as it should be. Game loads as much as it can into memory on the card. Basically why the memory is there, to avoid the slower dynamic loading of textures/resources.

    It is suprising though that the Nvidia cards are actually loading up higher (better?) GPU wise than either of the AMD cards i've tried.

    I first started with dual HD6850s 1g cards in non crossfire and got the same usage pattern that I have now. 0-100-0-100
    You would think that original card would have been sweating bullets at 100%. Then I throw in a i7-990x and a R9 380x and get similar activity. Yet your Nvidia cards are going full out.

    Edited higher/better GPU wise

    Also it may be the way my driver reports GPU load activity. Still, seeing the core at half speed sucks.
     
    Last edited: Feb 11, 2016
  28. Tig Trainee Engineer

    Messages:
    51
    You're welcome for the new graph watching addiction! :pbjt:
     
  29. Commander Rotal Master Engineer

    Messages:
    4,979
    I meant the regular SE log file (SpaceEngineers.log) :/
     
  30. tyriael_soban Apprentice Engineer

    Messages:
    113
    Its an Asus ROG G720js, i can actually put my hands anywhere on the machine (without flipping it over during operation, obviously) apart from the heat vent for the gpu, i cannot say ive been able to do that with a laptop before ...
    I scaled my legs with a dell D510 while i was in college playing Empire At War on it :woot:
    The tools i have do not allow me to change the settings beyond specifying if the machine is allowed to use active cooling or not, since i never do any serious gaming on batteries, the fans are always on in that respect, the cpu is set to full when on mains power as well.

    The framerates and sim speeds i reported in my earlier post where from v1.113ish when nearly everyone and their dog was having sim issues when they got anywhere near planets.
     
Thread Status:
This last post in this thread was made more than 31 days old.