WINDOWS SYSTEM IS RUNNING MULTI-DISPLAY MODE!

1

Hi, 


I am running an experiment in which I run MATLAB code from a computer to be presented on a single attached display monitor. I get the following error when I run my script:


==============================================================================================================
PTB-WARNING: WINDOWS SYSTEM IS RUNNING IN MULTI-DISPLAY MODE! ALL FLIP STIMULUS ONSET TIMESTAMPS WILL BE LIKELY UNRELIABLE!
PTB-WARNING: STIMULUS ONSET TIMING WILL BE UNRELIABLE AS WELL, AND GRAPHICS PERFORMANCE MAY BE SEVERELY REDUCED!
PTB-WARNING: DO NOT USE MULTI-DISPLAY MODE FOR RUNNING REAL EXPERIMENT SESSIONS WITH ANY REQUIREMENTS FOR ACCURATE TIMING!
PTB-WARNING: ==============================================================================================================


I did a Perceptual synchronization test for synchronization which appears to be come out totally fine. Also, timing onsets with tic, toc in matlab shows virtually no error in onsets...


Should I be worried about going forward with running the script for experimental purposes? 

 


2
---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :

Hi, 


I am running an experiment in which I run MATLAB code from a computer to be presented on a single attached display monitor. I get the following error when I run my script:


==============================================================================================================
PTB-WARNING: WINDOWS SYSTEM IS RUNNING IN MULTI-DISPLAY MODE! ALL FLIP STIMULUS ONSET TIMESTAMPS WILL BE LIKELY UNRELIABLE!
PTB-WARNING: STIMULUS ONSET TIMING WILL BE UNRELIABLE AS WELL, AND GRAPHICS PERFORMANCE MAY BE SEVERELY REDUCED!
PTB-WARNING: DO NOT USE MULTI-DISPLAY MODE FOR RUNNING REAL EXPERIMENT SESSIONS WITH ANY REQUIREMENTS FOR ACCURATE TIMING!
PTB-WARNING: ==============================================================================================================


I did a Perceptual synchronization test for synchronization which appears to be come out totally fine. Also, timing onsets with tic, toc in matlab shows virtually no error in onsets...


Should I be worried about going forward with running the script for experimental purposes? 

 

-> If you are truly running a Windows PC with only one single monitor connected and active then you should not see that message. I'd reboot and see if the message goes away. Otherwise your systems graphics system or driver is broken and not trustworthy at all.


-> tic and toc is not a proper way to do any timing related to experiment data collection. Please read through the tutorials in the PsychDocumentation folder, look at our demos etc., otherwise you have a bigger problem than only a potentially broken experiment setup.


-> Windows Vista/7/8/8.1 systems in dual display mode exhibited timing problems on different tested occasions in the past and are in my experience fundamentally not trustworthy for precise visual timing. If you wanted to use them anyway you'd need to attach photo-diodes or other external measurement equipment to find out if they work for your specific purpose. I know of many ways of tricking yourself even with photo-diodes though, so i personally wouldn't even trust that if precise timing for every stimulus frame really matters. If you don't care about precise visual timing you can of course simply ignore the warning and carry on.


-mario


3
Thanks for the response. I think I was unclear.

I have two screens.

I time all screen flips as per the accurate timing tutorials. I simply used tic toc to get a 'second opinion'. According to my measurements the screens are being flipped with arbitrary temporal accuracy. Granted, maybe that sort of measurement is meaningless.

Anyway, I ran a sync test and that turned out fine. Is that meaningful?

If the presentation of screens were off by say 10 ms, it would not really be a problem for my experiment.

Cheers,

On Wed, Nov 30, 2016 at 8:01 PM mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:

---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :


Hi,


I am running an experiment in which I run MATLAB code from a computer to be presented on a single attached display monitor. I get the following error when I run my script:


==============================================================================================================
PTB-WARNING: WINDOWS SYSTEM IS RUNNING IN MULTI-DISPLAY MODE! ALL FLIP STIMULUS ONSET TIMESTAMPS WILL BE LIKELY UNRELIABLE!
PTB-WARNING: STIMULUS ONSET TIMING WILL BE UNRELIABLE AS WELL, AND GRAPHICS PERFORMANCE MAY BE SEVERELY REDUCED!
PTB-WARNING: DO NOT USE MULTI-DISPLAY MODE FOR RUNNING REAL EXPERIMENT SESSIONS WITH ANY REQUIREMENTS FOR ACCURATE TIMING!
PTB-WARNING: ==============================================================================================================


I did a Perceptual synchronization test for synchronization which appears to be come out totally fine. Also, timing onsets with tic, toc in matlab shows virtually no error in onsets...


Should I be worried about going forward with running the script for experimental purposes?

-> If you are truly running a Windows PC with only one single monitor connected and active then you should not see that message. I'd reboot and see if the message goes away. Otherwise your systems graphics system or driver is broken and not trustworthy at all.


-> tic and toc is not a proper way to do any timing related to experiment data collection. Please read through the tutorials in the PsychDocumentation folder, look at our demos etc., otherwise you have a bigger problem than only a potentially broken experiment setup.


-> Windows Vista/7/8/8.1 systems in dual display mode exhibited timing problems on different tested occasions in the past and are in my experience fundamentally not trustworthy for precise visual timing. If you wanted to use them anyway you'd need to attach photo-diodes or other external measurement equipment to find out if they work for your specific purpose. I know of many ways of tricking yourself even with photo-diodes though, so i personally wouldn't even trust that if precise timing for every stimulus frame really matters. If you don't care about precise visual timing you can of course simply ignore the warning and carry on.


-mario


4



XXX---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :

Thanks for the response. I think I was unclear.

I have two screens.

-> That's bad on Windows 7/8/8.1, even if you only use one screen for the Desktop/Matlab GUI

I time all screen flips as per the accurate timing tutorials. I simply used tic toc to get a 'second opinion'. According to my measurements the screens are being flipped with arbitrary temporal accuracy. Granted, maybe that sort of measurement is meaningless.

Anyway, I ran a sync test and that turned out fine. Is that meaningful?

-> It's better than not, but no guarantee in itself. The failure case on Windows Vista/Windows-7 etc. is clicking with the mouse on any GUI Window on your non-stimulation display. E.g., if you'd click into the Matlab window to input something, or press some buttons etc. That will give keyboard input focus to that window and take away window focus from PTB's onscreen window. Loss of keyboard focus from the PTB window means that that window loses various graphics system optimizations which are crucial to get accurate/reliable timing and tearfree stimulus display. PTB will often not detect this problem and you'll just misstimulate your subjects. PTB is also in no way able to recover from such a situation or prevent it.

-> As long as you only look on your Matlab GUI but don't touch anything, you may be fine. Everything else is asking for trouble.

-> Windows-10 seems to be more robust in such a dual-display scenario as long as one display is for the GUI, the other for single-display visual stimulation. But i could only test with two Windows-10 machine, so ymmv. PTB on Windows 8 and later is incapable of detecting some of the hazards it could still detect on Windows-7, and entirely incapable of doing anything against them, so i don't know which system Win7 or Win10 is worse. Needless to say, Linux has no known problems with dual-display setups.

If the presentation of screens were off by say 10 ms, it would not really be a problem for my experiment.

-> If an error of up to 2 video refresh cycles on Windows-7 (potentially larger on Windows-8 and later due to DWM interference), e.g., 33 msecs on a 60 Hz display, and potential tearing artifacts under certain loads and/or with certain stimuli are still not a problem for your experiment then you would be ok. The severity of effects is load and hardware/software dependent, so just because PerceptualVBLSyncTest doesn't show obvious trouble doesn't mean that different stimulation scripts can't show trouble.

-> Running Windows-7 truly single display is usually safe timing-wise if PTB doesn't complain and our various tests don't show signs of trouble.

-mario

Cheers,

On Wed, Nov 30, 2016 at 8:01 PM mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:
 

---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :


Hi, 


I am running an experiment in which I run MATLAB code from a computer to be presented on a single attached display monitor. I get the following error when I run my script:


==============================================================================================================
PTB-WARNING: WINDOWS SYSTEM IS RUNNING IN MULTI-DISPLAY MODE! ALL FLIP STIMULUS ONSET TIMESTAMPS WILL BE LIKELY UNRELIABLE!
PTB-WARNING: STIMULUS ONSET TIMING WILL BE UNRELIABLE AS WELL, AND GRAPHICS PERFORMANCE MAY BE SEVERELY REDUCED!
PTB-WARNING: DO NOT USE MULTI-DISPLAY MODE FOR RUNNING REAL EXPERIMENT SESSIONS WITH ANY REQUIREMENTS FOR ACCURATE TIMING!
PTB-WARNING: ==============================================================================================================


I did a Perceptual synchronization test for synchronization which appears to be come out totally fine. Also, timing onsets with tic, toc in matlab shows virtually no error in onsets...


Should I be worried about going forward with running the script for experimental purposes? 

 

-> If you are truly running a Windows PC with only one single monitor connected and active then you should not see that message. I'd reboot and see if the message goes away. Otherwise your systems graphics system or driver is broken and not trustworthy at all.


-> tic and toc is not a proper way to do any timing related to experiment data collection. Please read through the tutorials in the PsychDocumentation folder, look at our demos etc., otherwise you have a bigger problem than only a potentially broken experiment setup.


-> Windows Vista/7/8/8.1 systems in dual display mode exhibited timing problems on different tested occasions in the past and are in my experience fundamentally not trustworthy for precise visual timing. If you wanted to use them anyway you'd need to attach photo-diodes or other external measurement equipment to find out if they work for your specific purpose. I know of many ways of tricking yourself even with photo-diodes though, so i personally wouldn't even trust that if precise timing for every stimulus frame really matters. If you don't care about precise visual timing you can of course simply ignore the warning and carry on.


-mario


5
Mario,

I appreciate the comprehensive answer. See in-text.

Ghislain d'Entremont
BSc, Neuroscience, Honors
Dalhousie University
tel: 902-802-5671 (text)

On Fri, Dec 2, 2016 at 7:19 PM, mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:




XXX---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :

Thanks for the response. I think I was unclear.

I have two screens.

-> That's bad on Windows 7/8/8.1, even if you only use one screen for the Desktop/Matlab GUI

Good to know.


I time all screen flips as per the accurate timing tutorials. I simply used tic toc to get a 'second opinion'. According to my measurements the screens are being flipped with arbitrary temporal accuracy. Granted, maybe that sort of measurement is meaningless.

Anyway, I ran a sync test and that turned out fine. Is that meaningful?

-> It's better than not, but no guarantee in itself. The failure case on Windows Vista/Windows-7 etc. is clicking with the mouse on any GUI Window on your non-stimulation display. E.g., if you'd click into the Matlab window to input something, or press some buttons etc. That will give keyboard input focus to that window and take away window focus from PTB's onscreen window. Loss of keyboard focus from the PTB window means that that window loses various graphics system optimizations which are crucial to get accurate/reliable timing and tearfree stimulus display. PTB will often not detect this problem and you'll just misstimulate your subjects. PTB is also in no way able to recover from such a situation or prevent it.

Understood.

-> As long as you only look on your Matlab GUI but don't touch anything, you may be fine. Everything else is asking for trouble.

Fortunately, I will not be clicking with the mouse on any GUI Window.

-> Windows-10 seems to be more robust in such a dual-display scenario as long as one display is for the GUI, the other for single-display visual stimulation. But i could only test with two Windows-10 machine, so ymmv. PTB on Windows 8 and later is incapable of detecting some of the hazards it could still detect on Windows-7, and entirely incapable of doing anything against them, so i don't know which system Win7 or Win10 is worse. Needless to say, Linux has no known problems with dual-display setups.

Got it.


If the presentation of screens were off by say 10 ms, it would not really be a problem for my experiment.

-> If an error of up to 2 video refresh cycles on Windows-7 (potentially larger on Windows-8 and later due to DWM interference), e.g., 33 msecs on a 60 Hz display, and potential tearing artifacts under certain loads and/or with certain stimuli are still not a problem for your experiment then you would be ok. The severity of effects is load and hardware/software dependent, so just because PerceptualVBLSyncTest doesn't show obvious trouble doesn't mean that different stimulation scripts can't show trouble.

I see. Ok, it seems still that for my purposes I want have any troubles.

-> Running Windows-7 truly single display is usually safe timing-wise if PTB doesn't complain and our various tests don't show signs of trouble.

-mario

Cheers,

On Wed, Nov 30, 2016 at 8:01 PM mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:

---In PSYCHTOOLBOX@yahoogroups.com, <ghislaindentremont@...> wrote :


Hi,


I am running an experiment in which I run MATLAB code from a computer to be presented on a single attached display monitor. I get the following error when I run my script:


============================== ============================== ============================== ====================
PTB-WARNING: WINDOWS SYSTEM IS RUNNING IN MULTI-DISPLAY MODE! ALL FLIP STIMULUS ONSET TIMESTAMPS WILL BE LIKELY UNRELIABLE!
PTB-WARNING: STIMULUS ONSET TIMING WILL BE UNRELIABLE AS WELL, AND GRAPHICS PERFORMANCE MAY BE SEVERELY REDUCED!
PTB-WARNING: DO NOT USE MULTI-DISPLAY MODE FOR RUNNING REAL EXPERIMENT SESSIONS WITH ANY REQUIREMENTS FOR ACCURATE TIMING!
PTB-WARNING: ============================== ============================== ============================== ====================


I did a Perceptual synchronization test for synchronization which appears to be come out totally fine. Also, timing onsets with tic, toc in matlab shows virtually no error in onsets...


Should I be worried about going forward with running the script for experimental purposes?

-> If you are truly running a Windows PC with only one single monitor connected and active then you should not see that message. I'd reboot and see if the message goes away. Otherwise your systems graphics system or driver is broken and not trustworthy at all.


-> tic and toc is not a proper way to do any timing related to experiment data collection. Please read through the tutorials in the PsychDocumentation folder, look at our demos etc., otherwise you have a bigger problem than only a potentially broken experiment setup.


-> Windows Vista/7/8/8.1 systems in dual display mode exhibited timing problems on different tested occasions in the past and are in my experience fundamentally not trustworthy for precise visual timing. If you wanted to use them anyway you'd need to attach photo-diodes or other external measurement equipment to find out if they work for your specific purpose. I know of many ways of tricking yourself even with photo-diodes though, so i personally wouldn't even trust that if precise timing for every stimulus frame really matters. If you don't care about precise visual timing you can of course simply ignore the warning and carry on.


-mario