Modartt user forum - cpu overload protection curiosity

Re: cpu overload protection curiosity

2019-12-02T09:35:25Z

Luc, thanks... in this case though, I don't think the DPC latencies are the main problem. I have spent many many hours with LatencyMon in order to get it to where it is. That (in conjunction with Dell tech support) is how I ended up disabling the battery and disabling C-states.

At this point, the oddity to me is that I seem to be hitting "CPU overload protection" cutoffs in pianoteq, even though it *seems* like I'm not actually overloading the CPU. Or maybe I am and I just don't hear/see the symptoms... not sure... that's why I was wondering just what exactly it is that triggers the overload protection, and what exactly the "performance index" numbers mean.

Re: cpu overload protection curiosity

2019-12-02T09:14:58Z

Only one answer comes to mind: use this freeware to check your system, it could be a very simple background task creating havocs
https://www.resplendence.com/latencymon

cpu overload protection curiosity

2019-12-02T08:15:46Z

Testing the demo version on a laptop with an i9-9980HK (dell precision 7740, 8-core, 2.4ghz, turbo to 5ghz). I'm plugged in, and have had to disable the battery in device manager as well as disable C-states in order to get glitch-free performance with minimal buffers using an RME interface.

When I run Pianoteq, I get occasional "CPU overload" cutoffs.

Interestingly, if I turn off overload detection, then it seems to work fine and I hear no audio glitches. Even when playing with my arms across lots of keys, with pedal down, and voice limit set to 96. :-) So it seems to be cutting off voices way before I'm hitting an audio issue.

This is happening before I hit the currently set 48 voice limit.

What is up with that? What values actually go into calculating whether or not there is a "CPU overload"?

Also, my Performance Index is reported to be at around 136. Again, when I turn off the detection and bang on lots of keys, I can get the Index up to around 139 or maybe 140.

I don't understand why it is either cutting off unnecessarily, or conversely why there are overloads but I don't hear any evidence of them when Overload Detection is off.

Thanks for any insight...