Help needed with understanding the readouts from the profiler correctly

I am trying to benchmark performance of different spatializers in FMOD. To this end I looked at the profiler and am not sure if I am reading the graphs correctly. Since the spatializers work differently internally, interpreting the numbers correctly is not straight forward.
Here a screenshot of the profiler data super imposed from both spatializers:

I am interested in the end how much processing power each spatializer ends up using. So should I only compare the numbers for system/mixer (total) and system/update (total)? Would this include the cost for the master bus? Or would the total processing power used be system + master bus per thead for total / update? Or is the cost for the master bus included already in the system readout?

Also, is performance data inaccurate if I get the warning: “monitoring:module:addData: not enough bandwidth to send monitoring data… due to backlog”?

Many thanks for helping me understand these numbers.

Best wishes,

It does seem the issue is that your game has too much going on at once for the profiler to keep up. The full error for monitoring:module:addData: not enough bandwidth to send monitoring data… due to backlog should have another message that says discarding %d commands (%d bytes worth) to tell you which commands were dropped.

Whilst the profiler should be able to handle 145 instances, I’m not sure if there is anything else going on during your tests. There is no real workaround for this other than to scale down your test. Hopefully this will give you some proper full results to work with.

Thanks Richard,

As you’ve suggested, I’ve reduced the performance test to 64 voices max in the end. That was sufficient in the end to get an overall idea. I also noticed that the error mentioned tend to appear when I’ve stopped the profiler which makes sense now that I think about it. So it’s better to stop Unity rather than the profiler to end recording data.

1 Like