True. But somewhere in most cases of this sort of thing, there may be something or other looping, going round and round and leaking away memory and performance. It has to be something.

I ran about 6 similar long timed tests last 3 days and every single time fps, gpu utilization, Watts, temperature went way way down slowly.

It did not go down even a tiny bit after over 2 hours this time. Utilization is still 100% (the test is still flying now). I saw it go down to 13% yesterday after a long time (flying straight ahead on auto pilot while I'm away doing something else in another room). (All the tests were run like that, straight flights on autopilot with no input from me).

FPS would start from 30, 35, or 40 fps and sink all they way to single digits every other time. All the way down to 4 or 5 fps in two hours.

2 hours ago, Fielder said:

And I just tested for 35 minutes, fps when UP from 24 to 27, and gpu utilization stayed exactly at 100% the whole time, temp went up 80 to 81, and watts down 174 to 172, fan speed increased slightly. Stable, no more problem.

I think this was what is called a memory leak from bad programming (my bad programing in MSFS Controls).

I think you are using the term memory leak incorrectly.  You were seeing performance degradation, not a memory leak.  A memory leak would indeed consume your available memory until that process can be forced to be terminated and free the memory back up.  It would inevitably end up in the application crashing or worse taking down the OS and requiring a reboot..  

This is one of my big pet peeves because a memory leak in coding terms is big.  If Asobo had a true memory leak in their code chances are more than a dozen or so people would be impacted...   Performance degradation does not = memory leak and "bad controller programming" isn't a memory leak.  Sounds more like a USB issue.  I have 2 throttles, a Warthog, an iris yoke, MFG rudder pedals, a steering tiller, 3 opencockpit modules a G13 and a button box, so in addition to the same controller as you I have 10 other controllers connected with no issues. 

No memory leak I can see in the code.  What is your memory consumption?

Edited August 20, 20214 yr by psolk

" What is your memory consumption? "

I don't know because Alt R does not tell me that in its window.

Semantics is of course important in order to understand what is being said. But semantics does not change the observed facts of degraded performance observed over time caused by something hogging resources.

I would now say, based on your message, that 'memory leak' is the wrong term to describe what was happening in my case.

10 minutes ago, Fielder said:

" What is your memory consumption? "

I don't know because Alt R does not tell me that in its window.

Semantics is of course important in order to understand what is being said. But semantics does not change the observed facts of degraded performance observed over time caused by something hogging resources.

I would now say, based on your message, that 'memory leak' is the wrong term to describe what was happening in my case.

This isn't semantics.  The entire thread was started based on the premise of a memory leak without ever looking at memory consumption.  From what I can see this also wasn't degraded performance by something hogging resources, what resources were being hogged?  You actually said your HW consumption was going down not up with degrading performance as your GPU was used less.  

So you've got degrading performance, GPU usage going down, haven't looked at memory consumption but started a thread that there is a memory leak in SU5 (only impacting certain users)  which is a big issue "if true."  I'm just saying, throwing out the term memory leak is actually a big big deal in coding and it's hurling a massive accusation against the code quality.  It's also usually monitored by memory consumption so to even speak about a memory leak without monitoring memory is impossible.   

Perhaps more accurate is "degrading performance over time, unsure as to cause"  

When I see memory leak I think fatal coding flaw.