Ftdi plugin using 100% CPU.

[r0b0t0]

Hello everyone, I wanted to ask if anyone else has high CPU usage when using the FTDI plug-in in "OpenTX" mode on a 485 interace with FTDI chips? (i tested all the other output modes btw)

I tried on different windows 10 computers, intel i5 of different generations 4-7-10. Due to the nature of QLC, the percentage changes according to the number of cores, but the core used is the one that reaches 100%.
I tried changing the frequency one by one from 1Hz to 30Hz, changing the FTDI driver, using only 1 universe in QLC4 and beta 5. I tried with VCP uninstalled and installed. I tried an earlier version of qlc (4.10) and it's a bit smoother, the plug-in consumes 80-90% of the cpu and doesn't reach 100% so it's totally usable without crashing things. But in the last ones reaches 100% and becomes unstable.

I understand that in this type of cable, the processing of the dataflow is done by the host CPU, unlike more expensive interfaces where there are processors that take care of it, but using other software that sends DMX through this type of cable I don't have the excessive use of cpu.

Does anyone know what I may be doing wrong?

At the moment I am using the art-net output to a loop device and have installed art-net2dmx outputting the universe to the same ftdi chip to avoid getting stuck.

Greetings and thanks in advance

[GGGss]

I understand that in this type of cable, the processing of the dataflow is done by the host CPU, unlike more expensive interfaces where there are processors that take care of it, but using other software that sends DMX through this type of cable I don't have the excessive use of cpu.

This is true, but this shouldn't have such a high processor (core) load even on older computers.
Instead of altering the DMX frequency, you could opt to limit the number of channels transmitted. It all depends on how the FTDI communication is processed. In DMX world, the full potential of FTDI isn't needed at all, since DMX is one-way communication (dump and forget), whereas true FTDI is half duplex bi-directional communication.
So now the FTDI driver might be the culprit.

At the moment I am using the art-net output to a loop device and have installed art-net2dmx outputting the universe to the same ftdi chip to avoid getting stuck.

What you are saying is that art-net2dmx running in parallel to QLC+ on the same machine with the same interface does __not__ show such high processor (core) loads?
This surely would be something to invest some time in and could be a potential bug.

[r0b0t0]

What you are saying is that art-net2dmx running in parallel to QLC+ on the same machine with the same interface does __not__ show such high processor (core) loads?
This surely would be something to invest some time in and could be a potential bug.

Thanks for answering!

Yes, the same machine, same ftdi, all in local.
QLC+ -> ftdi dmx 100%cpu
QLC+ art-net out -> KM TEST Loopback Adapter -> art-net2dmx -> ftdi dmx ~2% CPU.

Same result, testing a fixture in channel 450-510.

It could be a potential bug, considering that I have QLC+ 4 and 5 installed on the same machine (I'm migrating things) and they share information in the system registry.

[Yestalgia]

I've cooked a computer with an infinite artnet loop before, maybe make sure it's setup correctly?

[GGGss]

I've cooked a computer with an infinite artnet loop before, maybe make sure it's setup correctly?

Lachie,
The OP is complaining about QLC+ cooking the processor. When he uses the Artnet work-around his processor is chill.

The only way to debug this will be, having the configuration sent to Massimo for him to verify what's wrong.
Unless he already has an idea.

[mcallegari]

I think this depends on the USB device which I probably don't have since the OP didn't mention which product he's talking about.
If it was a larger issue with any open DMX device, then all the ENTTEC clones wouldn't work too, which I don't think is the case

[r0b0t0]

Thanks for the answers
The adapter I have connected is an FT232R. On windows 10 64, using FTDI driver 2.12.36.4 05-07-2021.
From the serial number (A50285BI) I suspect that the adapter might have a fake chip. It strikes me that it only works badly in QLC. I mean, it works fine, but the CPU shoots up.

EDIT: I have continued doing tests and below 30Hz is where the CPU shoots up. With 40-50Hz the CPU drops to less than 1%. It is very rare behavior. But now it works great. I even tried using 2 adapters simultaneously connected, the same thing, under 30Hz the CPU shoots up, above 40Hz everything normalizes.