1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

[Request] Queue depth monitoring

Discussion in 'Rivatuner Statistics Server (RTSS) Forum' started by CaptaPraelium, Oct 6, 2018.

  1. CaptaPraelium

    CaptaPraelium Master Guru

    Messages:
    229
    Likes Received:
    62
    GPU:
    1070
    This would be very helpful for monitoring the behaviour of games under the hood; for predicting input lag, correctly configuring 'pre-rendered frames' settings or framerate limiters accordingly, seeing the impact of recording software or overlays or chat apps or other background applications, identifying bugs with game patches or windows updates, etc.

    I know this is possible to monitor, using non-realtime tools which log to files for post-analysis, such as GPUView, but to have it in realtime would be priceless. I've spent a lot of time starting the game, logging a few seconds, exiting the game, parsing the logs, make a change, repeat... when all I really needed was a few line graphs to tell me the depth of these queues.

    Suggested monitoring items would be for example:

    Context CPU Queue Depth (for currently active application ie our game)
    Context CPU Queue Present calls (so we can see how many complete frames are queued for the GPU)
    Context CPU Queue Present call Latency (each time a present call is emptied, updates with the time it spent in the queue)
    GPU Hardware Queue Depth (for correlation with CPU depth to identify causes of idle time)

    Basically, a very highly simplified approach to what GPUView can tell us. Once, I had to use FRAPS and analyse logs afterwards, to find 1% lows, but not any more... This would just be as good as that :) I'm not sure how possible it is, to implement this in realtime, though....

    For those not familiar with the purpose of monitoring such things, You may be interested in this documentation to start with. https://graphics.stanford.edu/~mdfisher/GPUView.html
     

Share This Page