| Reference | Downloads | Github

Lagging and slowdown when running experiments

Hi all,

A group of us have recently experienced major lagging issues while running our experiments. Given that the same issue affects multiple users across multiple studies, issues with the experimental coding itself can’t be the explanation. Also, older experiments which worked fine before are also affected (in addition to newly-written experiments using the most recent Psychopy release).

The issue itself largely occurs during responding methods (key presses, mouse clicks and rating scales). Participants experience long delays (sometimes above ten seconds) before they can progress forward in the experiment.

Our standard lab computers use integrated graphics cards. Testing on PC’s with discrete graphics cards outside of the lab removes the issue. Also, we have tested some older versions of Psychopy on our lab PC’s and the issue disappears with 1.83.04, although we haven’t tested every release.

The unusual aspect of this is how suddenly users were affected. We have speculated that a Windows 10 update might have had a role to play - some kind of issue between Windows 10, the more recent Psychopy releases and our specific hardware set up? (it’s quite hard to disentangle). As most of us are Linux users outside of the lab, we are also going to test whether an operating system change might help.

Has anyone else experienced a similar problem to this? Any thoughts or advice would be greatly appreciated. Please feel free to request more details (I have tried to make this as brief as possible).


1 Like

I have also experienced this issue! presenting times for trials needing responses has been set as 2 seconds but it will always present for much longer (15-20 seconds) and im not sure what to do about it so i’d also welcome any advice!

It sounds like you’ve found the issue right there. We don’t recommend using integrated graphics cards on Windows, although sometimes the issue can be resolved by installing the latest driver direct from the manufacturer (don’t be content with one provided by Windows).

Thanks Michael - using discrete graphics cards going forward seems like the most sensible option, although we’ll have a look at driver suggestion too.

The only odd part of this is that rolling back to an earlier version of Psychopy has also provided a quick fix to the problem - would this simply be that newer versions require more hardware resource?