Render crash with RAW clips

Computer: i9, 3090 GPU, 64GB RAM
Resolve: 18.1.2 Build 6
Camera: URSA Mini 4K, RAW (Cinema DNG)

Problem: Computer motherboard beeps when rendering the RAW clips, followed by BSOD.

I've experimented with multiple timelines and projects, trying different render settings and codecs. If my source material in the timeline is anything but the RAW clips from the camera, the render is fine. CPU gets a little activity, but the GPU takes the brunt and no crashes or beeps. If my source material is the RAW files, CPU instantly hits 100% and the GPU hardly does anything.

I updated to the latest GPU drivers and same result. I also "forced" the GPU in my Resolve preferences. Everything plays just fine in the timeline, it's only on export that it crashes, and only with RAW clips. H.264, ProRes, etc... all play and render just fine.

I thought about trying the latest beta for 18.5, but I have a bunch of projects in the works and would prefer the stability of an official release.

I even opened an older project that was just h.264 and did a render - everything was fine. Imported my RAW clips, added them, and it crashed. Tried different and also brand new RAW clips. Same problem.

Final note: these are just vanilla clips, no color or effects applied.

Update:
Been at this all day. I even broke down and installed the beta for 18.5.

Sometimes, after a hard restart on the computer, Resolve will render the RAW footage just fine, using the GPU. But if I try for a second render, it's the CPU that kicks in and maxes out and the computer shuts down.

Any ideas?

What's your PSU situation? 3090 is a very power hungry GPU with huge transient spikes. Once you hit one of those if your PSU can't handle it it will trip the protection and shut down or restart.

If your PSU is fine then how are you supplying power to the GPU cable wise? You should always use separate cables from the PSU. Daisy chaining leads to power issues.

Could also be that the CPU is overheating. Is there any way you can monitor this?

It's a 1000w PSU. GPU is hooked up directly to it, no daisy chain, no adapters. I've opened the monitoring app the came with the motherboard. CPU is fine with rendering regular footage, maybe 60 degrees C, but once I toss the CDNG stuff at it, the load increases to 100% and the temp skyrockets and it starts beeping. If I can stop the render right away, I can prevent the BSOD and shutdown. It's just strange that the GPU doesn't render the RAW clips like it does everything else.

The CPU is also cooled by a Noctua NH-D15, which I thought would be more than enough. It barely fits the case.

I have no cDNG stuff around it this time, but what if you disable decoding of it by the GPU? Might be slow, but stable.

It's a 1000w PSU. GPU is hooked up directly to it, no daisy chain, no adapters. I've opened the monitoring app the came with the motherboard. CPU is fine with rendering regular footage, maybe 60 degrees C, but once I toss the CDNG stuff at it, the load increases to 100% and the temp skyrockets and it starts beeping. If I can stop the render right away, I can prevent the BSOD and shutdown. It's just strange that the GPU doesn't render the RAW clips like it does everything else.

The CPU is also cooled by a Noctua NH-D15, which I thought would be more than enough. It barely fits the case.

If the CPU skyrockets in temperature then that's your culprit. How old is your CPU, have you ever repasted it? The paste might have dried out.

One last, highly unlikely thing, 1000W might not be enough if you have like a 12th gen i9 that consumes 400W under heavy load (cDNG decompression is a heavy load) and then the 3090 transients to over 600.

What you could do if you don't wanna check the CPU paste and stuff is to limit the render speed in the delivery tab to under 100% and see if that's stable.

Still, hitting 100% should throttle the system not shut it down or restart.

I have no cDNG stuff around it this time, but what if you disable decoding of it by the GPU? Might be slow, but stable.

It's just strange that the GPU doesn't render the RAW clips like it does everything else.

You can't. There is no decoding with cDNG like there is with H264 or RED. That part is done by the CPU always, afaik, and the debayer and processing done by the GPU once its converted to FP32.

Well, it seems to be a hardware issue anyway.

Well, it seems to be a hardware issue anyway.

CPU overheating would be my guess. Or PSU tripping the over current protection. Especially if there's no BSOD or anything and it just shuts down.

And dried out thermal paste is a good guess too.

Thanks for the input everyone.

I didn't know that the GPU wasn't used for processing cDNG, so that's good to know. The computer is a new build, less than a month old, though the CPU is a 10th gen. I guess I'll take it apart and reapply a liberal amount of thermal paste. That's really the only thing that I can think of too.

So here's the strange part. My computer was off all day yesterday. I just turned it on and launched Resolve. I wanted to try to transcode 3 minutes worth of cDNG to MXF for Avid. The entire render, my CPU was between 18%-28%. That's it. Temps were between 55 and 61. All five shots were transcoded, no overheating, no issues.

Immediately after the render, CPU went down to 0% and temp back down to 37. Right away, I added the same job again, didn't change anything, and within seconds the CPU was at 100% and it started to beep. I cancelled the render before it could shut down the computer.

I can't figure out why it would work fine the first time, then seconds later acts up the second time.

Working at 100% shouldn't cause overheating. I wonder if you have a bad CPU fan, or a maybe a bad mobo if it's a PWM fan.

For a new machine, it doesn't sound like thermal paste, and "more" isn't better than less. And the fact that you report, in later posts, that it works sometimes.... FWIW, I've had a D15 installed for years, with one pea-size drop of paste.

You might want to eliminate v18.x as a possible source..... Uninstall and try 16 or 17. And have you tried a "stress" test? That would at least let you observe performance at the outer limits, outside Resolve.

And you aren't overclocking, are you? If so, get rid of it.

So here's the strange part. My computer was off all day yesterday. I just turned it on and launched Resolve. I wanted to try to transcode 3 minutes worth of cDNG to MXF for Avid. The entire render, my CPU was between 18%-28%. That's it. Temps were between 55 and 61. All five shots were transcoded, no overheating, no issues.

Immediately after the render, CPU went down to 0% and temp back down to 37. Right away, I added the same job again, didn't change anything, and within seconds the CPU was at 100% and it started to beep. I cancelled the render before it could shut down the computer.

I can't figure out why it would work fine the first time, then seconds later acts up the second time.

First time the clips were probably read from the disk which is slower and it gave the CPU time to breathe and second time they were cached from RAM which removes the bottleneck and uses the CPU 100%.

Download HWINFO64 and watch which of the component skyrockets in temperature and other anomalies that might occur the moment you add the load.

One of my colleagues just repasted his GPU a few days ago. The paste was pretty dry and there were parts of the core that were not covered but the temperatures were within normal range.

The anomaly was the clocks were jumping from 300mhz to 1300mhz every second which shouldn't happen. This eventually lead to black screens. Repasted and everything works normally, 1300mhz all the way.

Give HWINFO a chance and see which component acts weird. No point going blindly.

If the CPU skyrockets in temperature like you say above, then there's an issue with either the paste or the cooler. Being a new build matters not.

Really appreciate all the help guys.

I ran a test in Blender by opening a scene and forcing it to use the CPU. It hits 100%, and the temp hits 100, but nothing shuts down, so it must of been getting really hot in Resolve. Fortunately the GPU works great in Blender so my CPU is never running that high, but I was curious what would happen.

I have found success with changing the render speed to 50. Now the CPU only hits 28% and the temp stays under 80 while under stress. I tested this multiple times for a total of ten minutes and haven't had a single failure, so I think that solves this problem. I have another issue unrelated that I'll start a new thread for, but in the meantime, I'd say this one is solved.

Really appreciate all the help guys.

I ran a test in Blender by opening a scene and forcing it to use the CPU. It hits 100%, and the temp hits 100, but nothing shuts down, so it must of been getting really hot in Resolve. Fortunately the GPU works great in Blender so my CPU is never running that high, but I was curious what would happen.

I have found success with changing the render speed to 50. Now the CPU only hits 28% and the temp stays under 80 while under stress. I tested this multiple times for a total of ten minutes and haven't had a single failure, so I think that solves this problem. I have another issue unrelated that I'll start a new thread for, but in the meantime, I'd say this one is solved.

Great that it works for you.

Different loads (Blender and Resolve). However, 100degrees is quite a lot even for 100% utilization. Keep an eye on this, sooner or later you'll need to fiddle with it.

Edit: T Junction for 10th gen 10900K is 100celsius so you're maxed out. No wonder it shuts down.

"The T-junction of the Intel® Core™ i9-10900K processor is 100°C, which means that any temperature value below that degree is considered normal and expected. "

https://community.intel.com/t5/Processo ... -p/1283904

Yeah, I should've done some more research before choosing that processor. I was trying to save a few bucks on the build and figured that the GPU was going to be used more so I didn't mind using a slightly older CPU. Sometimes you gotta learn the hard way.

Look for a better cooler or some extra case fans. That should do the trick. There's nothing wrong with a 10th gen CPU. Still rocking a Threadripper Zen 1 on one system. Works like a champ.

What happens if you drop the render speed a lot lower, at least to 25fps or even 10fps?



Do that, and it might help reduce the stress level on the CPU and the GPU and help you finish rendering the job effectively.

What happens if you drop the render speed a lot lower, at least to 25fps or even 10fps?



Do that, and it might help reduce the stress level on the CPU and the GPU and help you finish rendering the job effectively.

I have found success with changing the render speed to 50. Now the CPU only hits 28% and the temp stays under 80 while under stress. I tested this multiple times for a total of ten minutes and haven't had a single failure, so I think that solves this problem. I have another issue unrelated that I'll start a new thread for, but in the meantime, I'd say this one is solved.