Fullscreen vs Windowed CPU/GPU

[DflippinK]

I think ive mentioned this before, but was hoping to get a little more insight on it...

 

I am curious what the difference to hardware is when running Fullscreen vs. Borderless Windowed mode.. It is apparent via monitoring hardware utilization, that when in Fullscreen mode, there is an approximate increase in CPU utilization of about 5-10% and GPU utilization is approximately 10-15% less vs Windowed mode.. L-alt Enter into Windowed mode, CPU utilization drops and GPU load increases.. In the past, i had noticed some stutters would happen in fullscreen mode where they didn't in Windowed mode... now they both run silky smooth, but the utilization variances between the two still remain.. Makes me wonder what is going on internally and weather or not it could be advantageous for some users to run the game one way or another.. Would love to hear some thoughts on this..

[imacken]

From my perspective, the difference is that with vsync off, I get screen tearing in windowed mode and not in full screen.

I almost always have to do alt-enter from the start when runnng in non VR to get into full screen.

[DflippinK]

From my perspective, the difference is that with vsync off, I get screen tearing in windowed mode and not in full screen.

I almost always have to do alt-enter from the start when runnng in non VR to get into full screen.

interesting... ive been playin around with it alot lately.. created a heavily saturated mission with a ton of units... in fullscreen mode, i was getting a range of 11-15% CPU total load.. and GPU at 54-68% load... but noticed the CPU primary core was flatlining at times creating an occasional lag... L-alt entered into windowed mode and CPU load drops 7-9%, primary core saturation drops well within its threshold, and the GPU increases load range to 67-81%... but now no flatlined cores and the mission runs silky smooth.. tested with both in game Vsync and Nvidia CP Vsync.. almost identical results between the two... its just strange.. as if some of the rendering is being swapped between the CPU and GPU dependent on which mode im in.. doubt thats the case.. but it sure acts like it.. :dunno:

[erniedaoage]

I am running dcs in window mode because i get a much smoother experience than in fullscreen mode. With the same settings i loose 30 to 50 frames in fullscreen and i get a lot of stutters, i am around 70 to 90 fps in fullscreen. When i change to window mode i get 100-140 and all the stutters are gone. But for me its the only game i have which runs better in window mode than in fullscreen.

[imacken]

interesting... ive been playin around with it alot lately.. created a heavily saturated mission with a ton of units... in fullscreen mode, i was getting a range of 11-15% CPU total load.. and GPU at 54-68% load... but noticed the CPU primary core was flatlining at times creating an occasional lag... L-alt entered into windowed mode and CPU load drops 7-9%, primary core saturation drops well within its threshold, and the GPU increases load range to 67-81%... but now no flatlined cores and the mission runs silky smooth.. tested with both in game Vsync and Nvidia CP Vsync.. almost identical results between the two... its just strange.. as if some of the rendering is being swapped between the CPU and GPU dependent on which mode im in.. doubt thats the case.. but it sure acts like it.. :dunno:

Interesting.

I have tried to do some tests, but am struggling to see how this is measured. I mean when vsync is off, then both GPU and CPU (single core) are at 100%, and when vsync is on, it is not a fair comparison as full screen is locked to 60Hz and windowed is not.

This then led me to ask which mode is which! How do we tell if we are in windowed or full? I have always assumed that the mode that gives me tearing - move head left/right quickly in cockpit and look at cockpit verticals - with 60 fps or greater, is windowed, but now I'm not sure!

[draconus]

From my perspective, the difference is that with vsync off, I get screen tearing in windowed mode and not in full screen.

I almost always have to do alt-enter from the start when runnng in non VR to get into full screen.

I do exactly the same. I cannot believe how people can play with vsynch off - it used to be mode for benchmarking only.

As for my low end rig I see much increased CPU load and much lowered GPU load in full screen. I would probably like the other way around but VS-off is not an option for me.

 

GPU has always been optimised to work in fullscreen so it's no-brainer but why CPU is higher - I don't know.

[DflippinK]

Interesting.

I have tried to do some tests, but am struggling to see how this is measured. I mean when vsync is off, then both GPU and CPU (single core) are at 100%, and when vsync is on, it is not a fair comparison as full screen is locked to 60Hz and windowed is not.

This then led me to ask which mode is which! How do we tell if we are in windowed or full? I have always assumed that the mode that gives me tearing - move head left/right quickly in cockpit and look at cockpit verticals - with 60 fps or greater, is windowed, but now I'm not sure!

,

 

oh yes.. i see... you are correct.... if you want to measure the load differences, you can only do so with the in game Vsync.. Nvidia Vsync only works in full screen... but it does provide the same load result as fullscreen with in game vsync.. sorry about that... but if you use in game Vsync, it will work in both Fullscreen and Windowed... and you can just L-alt Enter between the two both locked at 60hz..

[draconus]

but if you use in game Vsync, it will work in both Fullscreen and Windowed

It doesn't work like that for me. I have VS-on option in DCS but I have to LAltEnter to fullscreen (I believe) to make it work.

Edited August 8, 2018 by draconus

[DflippinK]

Iit doesn't work like that for me. I have VS-on option in DCS but I have to LAltEnter to fullscreen (I believe) to make it work.

... really??... what happens if you dont Lalt Enter??.. do you have Fullscreen checked or unchecked in game settings?

[draconus]

... really??... what happens if you dont Lalt Enter??.. do you have Fullscreen checked or unchecked in game settings?

I have Fullscreen option checked all the time in DCS too. Unless I LAltEnter I am probably in WindowedBorderless and VS doesn't work then.

[DflippinK]

I have Fullscreen option checked all the time in DCS too. Unless I LAltEnter I am probably in WindowedBorderless and VS doesn't work then.

 

 

Well... i most often have fullscreen unchecked in DCS... but the Vsync works for me both ways... maybe try unchecking fullscreen and restarting DCS in windowed mode first before LAlt entering...

[imacken]

,

 

oh yes.. i see... you are correct.... if you want to measure the load differences, you can only do so with the in game Vsync.. Nvidia Vsync only works in full screen... but it does provide the same load result as fullscreen with in game vsync.. sorry about that... but if you use in game Vsync, it will work in both Fullscreen and Windowed... and you can just L-alt Enter between the two both locked at 60hz..

 

I'm not using nVidia vsync, just in-game one. It only works in full screen. On first loading, DCS is in windowed mode as my FPS is huge. It's not until I do an alt-enter, that the FPS goes to 60.

[DflippinK]

I'm not using nVidia vsync, just in-game one. It only works in full screen. On first loading, DCS is in windowed mode as my FPS is huge. It's not until I do an alt-enter, that the FPS goes to 60.

 

 

well... i wonder what the issue is... Vsync will lock my system to 60 no matter which mode i am in... again though,.. i usually start DCS in windowed mode with the Fullscreen checkbox unchecked so i can easily work between my main display and my reference monitor.. and then LAlt Enter between the two... but i have also tested it with Fullscreen checked and it also works.. the only thing that changes Vsync for me is weather or not i have it enabled or not... hmm, wonder if maybe you and Draconus might have a display setting outside of DCS that might be affecting it?.. for me, all system display settings are locked at 3840x2160 with 60Hz refresh rate.. and all Nvidia Vsync options are set to (application controlled) default setting.. ideally leaving DCS as the only dictating factor of weather or not Vsync gets used or not...

 

 

edit: just checked out your monitor... nice!!.. any chance you might have a GSync setting that could be affecting it??

Edited August 8, 2018 by DflippinK
mistated: no* Vsync (thanks imacken)

[imacken]

well... i wonder what the issue is... no Vsync will lock my system to 60 no matter which mode i am in... again though,.. i usually start DCS in windowed mode with the Fullscreen checkbox unchecked so i can easily work between my main display and my reference monitor.. and then LAlt Enter between the two... but i have also tested it with Fullscreen checked and it also works.. the only thing that changes Vsync for me is weather or not i have it enabled or not... hmm, wonder if maybe you and Draconus might have a display setting outside of DCS that might be affecting it?.. for me, all system display settings are locked at 3840x2160 with 60Hz refresh rate.. and all Nvidia Vsync options are set to (application controlled) default setting.. ideally leaving DCS as the only dictating factor of weather or not Vsync gets used or not...

 

 

edit: just checked out your monitor... nice!!.. any chance you might have a GSync setting that could be affecting it??

 

I’ve lost track of what the issue is!

If you search around, you’ll see that the Full Screen option in-game does nothing. Everyone has to do alt-enter to get into full screen after game starts no matter what that option is set to.

If you have vsync off, then you shouldn’t be locked to 60 anywhere!

Yes, GSync is on, but that should only kick in when vsync is off and frame rates are above 60 to remove the tearing effect I described earlier.

[DflippinK]

I’ve lost track of what the issue is!

If you search around, you’ll see that the Full Screen option in-game does nothing. Everyone has to do alt-enter to get into full screen after game starts no matter what that option is set to.

If you have vsync off, then you shouldn’t be locked to 60 anywhere!

Yes, GSync is on, but that should only kick in when vsync is off and frame rates are above 60 to remove the tearing effect I described earlier.

 

 

i understand your position... what i am suggesting to you is that in Game Vsync works in both Fullscreen and Borderless Windowed mode... in order to achieve this result, you may have to eliminate any variables that are preventing you from doing so... Vsync in game checkbox should be the only determining factor of weather or not Vsync is in use..

 

ex...

 

Vsync OFF= unconstrained fps variable framerate...

Vsync On= 60 fps no matter Fullscreen or Windowed..

 

 

...no matter what anyone with any performance issues is talking about..

 

 

in achieving this result, you will then be able to measure the differences in CPU/GPU utilization between Fullscreen and Windowed mode by switching between the two with LAlt-Enter as many times as you like while flying.

[Art-J]

In-game V-Sync has been limiting fps to 60 for me, but at the same time never got rid of tearing (thus indicating that borderless windowed mode is still on and V-Sync kind-of-works-but-not-quite). Whether I want to use in-game V-sync or NVidia one, Alt-Enter has always been the only way to switch to actual fullscreen + no tearing = V-Sync really operational.

 

I use only NV one simply for ease of switching between locked and unlocked frames for performance evaluation.

[DflippinK]

In-game V-Sync has been limiting fps to 60 for me, but at the same time never got rid of tearing (thus indicating that borderless windowed mode is still on and V-Sync kind-of-works-but-not-quite). Whether I want to use in-game V-sync or NVidia one, Alt-Enter has always been the only way to switch to actual fullscreen + no tearing = V-Sync really operational.

 

I use only NV one simply for ease of switching between locked and unlocked frames for performance evaluation.

 

 

are you monitoring fps??.. if you are seeing any tearing, it is because you are out of sync.. which would suggest Vsync is not working correctly.. or the GPU is dropping below the threshold of the refresh rate... which can be caused by any number of variables..

[imacken]

i understand your position... what i am suggesting to you is that in Game Vsync works in both Fullscreen and Borderless Windowed mode... in order to achieve this result, you may have to eliminate any variables that are preventing you from doing so... Vsync in game checkbox should be the only determining factor of weather or not Vsync is in use..

 

ex...

 

Vsync OFF= unconstrained fps variable framerate...

Vsync On= 60 fps no matter Fullscreen or Windowed..

 

 

...no matter what anyone with any performance issues is talking about..

 

 

in achieving this result, you will then be able to measure the differences in CPU/GPU utilization between Fullscreen and Windowed mode by switching between the two with LAlt-Enter as many times as you like while flying.

 

Ok, I’ve spent about 2 hours testing all this out. I tested with GSync on full and windowed, full only and off. For each of these I test with vsync on and off with both full screen ticked in game and unticked.

I monitored the GPU/CPU load in-game with switching between full and windowed monitoring the load at all times, and there is no difference whatsoever on my system.

What I didn’t realise before is that I could set GSync to work in windowed and full screens, as well as full only, so now I don’t see any tearing at all. It is now impossible for me to tell which mode I am in!

[DflippinK]

Ok, I’ve spent about 2 hours testing all this out. I tested with GSync on full and windowed, full only and off. For each of these I test with vsync on and off with both full screen ticked in game and unticked.

I monitored the GPU/CPU load in-game with switching between full and windowed monitoring the load at all times, and there is no difference whatsoever on my system.

What I didn’t realise before is that I could set GSync to work in windowed and full screens, as well as full only, so now I don’t see any tearing at all. It is now impossible for me to tell which mode I am in!

 

 

sweet!! glad you got it working!!... do you have any monitoring overlays that you can observe while in game?... I use a reference monitor but an overlay should also be able to display GPU/CPU loads.. If you are on a single monitor and want to see if you are in Fullscreen or Windowed, you can do something like leave open a file folder, monitoring app,.. whatever ect.. before you start DCS... then while flying, if you LAlt Tab, it will put you in windowed mode every time and you will see an options to select the folder, app, whatever... either way, LAlt Tab should put you into windowed mode... once you have established that you are in windowed mode, you can just LAlt Enter between that and Fullscreen...

[imacken]

sweet!! glad you got it working!!... do you have any monitoring overlays that you can observe while in game?... I use a reference monitor but an overlay should also be able to display GPU/CPU loads.. If you are on a single monitor and want to see if you are in Fullscreen or Windowed, you can do something like leave open a file folder, monitoring app,.. whatever ect.. before you start DCS... then while flying, if you LAlt Tab, it will put you in windowed mode every time and you will see an options to select the folder, app, whatever... either way, LAlt Tab should put you into windowed mode... once you have established that you are in windowed mode, you can just LAlt Enter between that and Fullscreen...

 

OK, I have spent more time on this today, and have some interesting results.

Using Afterburner OSD overlay, I get the exact same results as you stated at the top of this thread, i.e. reduced GPU/increased CPU loads in full screen compared to windowed mode.

However, interestingly, Task Manager reports the opposite effect in GPU terms. It shows the same value as Afterburner in windowed mode, but jumps to 100% in full screen when it should be showing a drop. (I look at TM history after being in full screen DCS after a minute or two.)

[DflippinK]

OK, I have spent more time on this today, and have some interesting results.

Using Afterburner OSD overlay, I get the exact same results as you stated at the top of this thread, i.e. reduced GPU/increased CPU loads in full screen compared to windowed mode.

However, interestingly, Task Manager reports the opposite effect in GPU terms. It shows the same value as Afterburner in windowed mode, but jumps to 100% in full screen when it should be showing a drop. (I look at TM history after being in full screen DCS after a minute or two.)

hmmm.. interesting indeed.. i'll check it out hopefully later tonight or this weekend..

[DflippinK]

just checked.... but on my system, im not seeing any variances between monitoring softwares... task manager shows the same result as CAM monitor and Precision X.. CPU high/GPU low in fullscreen, and CPU low/GPU high in windowed..

[imacken]

just checked.... but on my system, im not seeing any variances between monitoring softwares... task manager shows the same result as CAM monitor and Precision X.. CPU high/GPU low in fullscreen, and CPU low/GPU high in windowed..

See attached screenshot. Afterburner and TM give very different values for CPU. DCS is on CPU11 - obviously! Weird.

Also, how do you get Precision XOC to show CPU load? I can only see CPU temps in my version.

[DflippinK]

See attached screenshot. Afterburner and TM give very different values for CPU. DCS is on CPU11 - obviously! Weird.

Also, how do you get Precision XOC to show CPU load? I can only see CPU temps in my version.

 

 

it does look like a difference in your OSD and TM values... could just be different sensors being averaged between the two... and yes, sorry.. Precision X does only report GPU usage... but also reports the GPU load the same as TM and CAM..

[imacken]

it does look like a difference in your OSD and TM values... could just be different sensors being averaged between the two... and yes, sorry.. Precision X does only report GPU usage... but also reports the GPU load the same as TM and CAM..

Yep, better off using MSI Afterburner, it has more options for the OSD.