Crash during transcode

Hi,
I regularly encounter crashes during the transcodes of OCF in proxy Op-Atom DNx36 25p.
I made a machine for our calibration room which currently serves as a transcoding machine.

Currently, the other 11 Resolve Studio workstations on a 2013 MACPro database do not have to worry about transcoding the project timelines. Unfortunately, the station Win10 Pro 20H2 which is the most powerful AMD 24 cores, RTX 3080 (Studio 460.89), 128GB Ram 16TB NVME crashes during transcoding and closes the Resolve application. The machines read the sources on an EditShare NL server in 10Gb and transcodes them into Proxy on a Nexis in 10Gb. The transcode also crashes when I transcode locally on NVMEs in RAID0.

The 2019 MAC Pro with 32Gb dual GPU transcodes at 150fps while the station I built allows for 290fps.

Can you help me solve this problem so that I do not have Resolve crashes after 60/80% of transcode.
But timelines are between 10 and 20 hours.

The machine does not crash every time. Some timelines are going very well. Other timelines crash but not in the same place. I see in the logs
Ignoring CUDA error 700: 'an illegal memory access was encountered', GPUManager\CudaBoardManager.cpp:317.

Here are the logs.
https://www.swisstransfer.com/d/676a1a4 ... 9da40c5fe9

Thanks
Boissard Michael

PC Win10Pro 20H2, Ryzen 3960X, 128Go DDR4, Gigabyte RTX 3080, system 512 Go SSD, data NVME 16To RAID 0

What's OCF?

Sorry,
OCF is Original Camera Files

Ahh. Thanks.

Could it be a thermal problem or getting close to the limits of the power supply?

Apparently, no thermal problem. The CPU is at 61°C, the GPU at 72°C and the 1600W power supply consumes 613W at max.

I have just performed several tests and when I limit the render speed to 100, I never have a problem with a crash. The same timeline without limitation sometimes crashes. I would like Resolve to add custom limitations or factors above (150 - 200 - 250).

I carry out many tests but I find it sad to limit the machine to 100 when it goes to 260/290 i/s (Faster than a MAC pro 2019 at 30,000 €).
I have a Z8 (xeon + 128GB) that I did not mount myself in which I installed the same RTX 3080 on a PCI-E 3 port and I have no crashes but I only go up to 130i/s.

Why is my super machine crashing?
Why such a big difference in speed between an HP Z8 at 14,000 € and my homemade station at 8,000 € ? Is it the PCI-E 4 and the high-end components that I have selected that allow this speed boost ?
Thanks