Blackmagic Forum

Hi,
I just started trying Resolve 17 and have found 1 issue,
Every time I try and use magic mask my workstation shuts down.
I am running latest W10Pro and latest Nvidia Studio drivers.
My workstation is Dell T7910 with 1300W power supply and specs as in signature.
I have never experienced this before with Resolve or any other program.
I have noticed that both GPUs are working very hard when magic mask is running, I usually get about 20secs of rendering then shutdown.

you might need a more powerful GPU, or more GPU memory
Try different GPU drivers

Hi Peter,
thanks for your reply, I do have 2x GTX1080Ti with 11G ram each, surely enough or do I have to look at new 3 series cards??

That's crazy. i didn't even thing that was possible until a few weeks ago when Adobe released the AI assisted face refinement tools. I have a decent workstation but the GPU pulls so much current when these filters are running that it shuts the computer down.

Have found a half fix, I have to turn off one of my graphics cards off in GPU configuration.
Just trying to work out how much it slows down other areas.

I have a RTX2080ti that's overclocked and I can confirm this happes to me as well. But it doesn't shut down the PC, it just shuts down the graphics card. I have to restart my monitor and it kicks in again.

Not sure if it's because it draws more power or if it overheats, but ugh... performance bugs that risk damaging hardware should not be in public betas. These kinds of bugs can cost people a lot and even if we agree to using beta software, Blackmagic has a responsibility to secure that it's at least safe to use the software.

I can use Magic mask with a humble Radeon 580. It's not fast, but neither does it shut down or run too hot.

Christoffer Glans wrote:Not sure if it's because it draws more power or if it overheats, but ugh... performance bugs that risk damaging hardware should not be in public betas. These kinds of bugs can cost people a lot and even if we agree to using beta software, Blackmagic has a responsibility to secure that it's at least safe to use the software.

If your system is shutting down or the graphics card is crapping out, the problem isn't BMD's, it's either Microsoft's or Nvidia's or a problem with the hardware. The worst we can blame BMD for is Resolve crashing. If your system hangs, reboots, shuts down, or bursts into flames, the blame for that is elsewhere.

It's a bit like running some high load process which causes a high draw from your PSU which in turn causes a circuit breaker to trip. Do we blame the software for the power outage?

It crashed on me when I had an image editor open, but after I closed that and relaunched Resolve, I was able to use the Magic Mask without any problems.

Jason Tackaberry wrote:

Christoffer Glans wrote:Not sure if it's because it draws more power or if it overheats, but ugh... performance bugs that risk damaging hardware should not be in public betas. These kinds of bugs can cost people a lot and even if we agree to using beta software, Blackmagic has a responsibility to secure that it's at least safe to use the software.

If your system is shutting down or the graphics card is crapping out, the problem isn't BMD's, it's either Microsoft's or Nvidia's or a problem with the hardware. The worst we can blame BMD for is Resolve crashing. If your system hangs, reboots, shuts down, or bursts into flames, the blame for that is elsewhere.

It's a bit like running some high load process which causes a high draw from your PSU which in turn causes a circuit breaker to trip. Do we blame the software for the power outage?

You do know that there are software viruses that make software cause hardware failure. If the software has bad code it can definitely damage hardware. Hardware and software aren't separate, they work in tandem, and it's the hardware manufacturers job to secure handling software and vice versa. RTX2080ti is not a card that is unknown to Blackmagic, and because Resolve uses GPU heavily, that symbiosis isn't unknown or untestable.

Any time when you can use the current hardware setup, clock and power draw in other software and games at the GPUs maximum without a problem and a new software comes around and shuts the card down, that is definitely on the software side. Why else would overclocking be stress tested? What do you think the stress tests do to the GPU card? Having a software shut down the card is not a sign that the GPU is overclocked and can't handle stress, it's a sign of bad code that shouldn't be in the software.

Version 17 is working faster and more stable on my alienware 15, nvidia gtx 980, 16gb ram, 2 ssd + 1 hdd.

Keeping my fingers crossed

Christoffer Glans wrote:Any time when you can use the current hardware setup, clock and power draw in other software and games at the GPUs maximum without a problem and a new software comes around and shuts the card down, that is definitely on the software side. Why else would overclocking be stress tested? What do you think the stress tests do to the GPU card? Having a software shut down the card is not a sign that the GPU is overclocked and can't handle stress, it's a sign of bad code that shouldn't be in the software.

You're confusing trigger with root cause. Any userspace application that can hork a system may be doing something extremely stressful, but it's the job of the OS and hardware to be robust against userland abuses. If I run an app that BSODs or panics the kernel, the bug is with the kernel (or drivers) or harware. The app may have triggered it, but the blame for the instability lies with one of the layers underneath. If the app "fixes" the issue, at best it merely works around triggering the underlying problem.

And that assumes Resolve is doing something "abusive." It's entirely possible it's using a less used and undertested API or instruction set. For example, AVX-512 instructions generated no end of instability early on, often highlighting insufficient cooling solutions, even though every other workload was stable on the same system. We don't blame the applications using AVX-512 for crashing systems.

It's likely the power supply that is insufficient for the huge power draw of the GPU:s, that's hardly a fault of the operating system or the application.

Similar issue with Magic Mask on PC. After making a selection and starting tracking, Resolve locks up. There are no other apps running other than all the Windows background stuff.
i9 10 core, Dual quadro p4000, 64GB Memory
Magic Mask does work on our trashcan MacPro, just quite slowly.

roger.magnusson wrote:It's likely the power supply that is insufficient for the huge power draw of the GPU:s, that's hardly a fault of the operating system or the application.

likewise just because it says 1300W doesn't mean its capable of sustaining that load indefinitely. You'd like to think it can.Good PS's brand new can. questionable quality, age, expect 70% load rating.

when was the last time you vacuumed / blew dust out of your system ? its yearly maintenance for me. if you have never done that, might be time. may also be worth looking into the PS and replacing it, especially if its over 5 years old.

roger.magnusson wrote:It's likely the power supply that is insufficient for the huge power draw of the GPU:s, that's hardly a fault of the operating system or the application.

Not if the system is stress-tested in different ways and timeframes. Overclocking should always stress-test to find the most stable setup and when that is done, no load should be worse than what that stress-test achieves as the stress test is pushing the absolute maximum performance load onto your system.

how long takes rendering a 20sec clip with a mask applied on your workstations... my counter will not stop counting the rendering time upwards.
Never had such problem with any other project even with resolve 16 fusion with 300+ masks and keyersapplied a 30 min film with h264 only took 2-3 hours so please don't tell me that the Magic Mask has MAGIC requirements.

MacBook Pro 16 2019:
2,6 GHz 6-Core Intel Core i7
64 GB 2667 MHz DDR4
AMD Radeon Pro 5600M 8 GB !!!

Similar Problem. Windows 10.

Magic Mask hangs the system, the fan on the computer ramps up an down a lot but nothing is happening.
Just a grey screen. Use task manager to kill it eventually.

In addition rendering from the deliver page has become so slow it is painful.
Now I loaded a project archive backup from resolve 16 (about 20 minutes long 1920 x 1080). 16 rendered in about 8 to 10 minutes, the same project in resolve 17 takes about 45 minutes to an hour.
My settings/paths are the same etc.

I am running an Asus Rog Stryx with an Nvidia 1080 gpu about 2 years old. Never had a problem with resolve 16 rendering or system crashes at that resolution.

Any help?

I found a fix that works on my machine. The GPU Configuration was set to OpenCL. We reinstalled CUDA drivers. Now everything seems to be working.

Wow. Thanks Michael.
I could not figure out why everything was so slow. I even uninstalled 17 and went back to 16 to check.
In 16 everything was working and renders were minutes instead of hours under 17.
Downloaded the cuda drivers selected GPU configuration (cuda was an option now) and everything is OK.
Thanks again.

I have similar workstation as the original post. Same thing happens after about 10-20seconds, immediate shutdown. Maybe there needs to be some kind of throttle option in the GPU or CPU preference for these kind of calculations?

Shuts my system down about 5 seconds after drawing the tool on a face. Have to pull the power plug to reset. This is not a great feature too say the least.

Bob

[quote="Christoffer Glans"]I have a RTX2080ti that's overclocked and I can confirm this happes to me as well. But it doesn't shut down the PC, it just shuts down the graphics card. I have to restart my monitor and it kicks in again."

Like others, mine shuts the computer down. Fans stop. The power light on the front of the MacPro is dark. It is, for all intents and purposes.... bricked. Until I pull the power cord. Then I can reboot.

Of course, I can't update the drivers for the Rx580. They are pretty much controlled by Apple.

I think by the responses that represent the randomness of this problem point to code that has some bugs. I can't imagine how the coders are going to fix this, making the magic mask a viable tool... but they are pretty sharp pencils. It will get fixed...hopefully soon. I can sure use this tool NOW.

Bob

I'm using an RX 580 here without any problem for magic masks.
Could it be a thermal problem or the power supply?

Hi,

Adding to this post as it describes experience I've been having using Magic Mask i.e.: after making stroke and processing, it will playback for 5 - 10 seconds, then the Mac Pro fans suddenly run at full and it is no longer possible to work in Resolve and computer shuts down. On reboot there is a kernel panic msg. I have attached the report from that.

Note that I usually work in 4K timelines, and even with some heavy processing on clips in Resolve it runs well and I never have fans at full/freeze up like this. This issue is happening on a simple HD timeline and 1080p clip and only if Magic Mask is applied.

I've updated to 17.4.3, I've tried several options as suggested here - turning off one or the other GPU, no change to issue. GPU Processing is set to auto (Metal).

It's over a year since these posts and there's no definitive answer so perhaps it's been solved? For now the best solution is simply not to use Magic Mask.

Thanks in advance if anyone has any other solutions to try.

BobTrim wrote:Shuts my system down about 5 seconds after drawing the tool on a face. Have to pull the power plug to reset. This is not a great feature too say the least.

Bob

if you dont have the pixlas mod, also a RX-580 can use to much watt on a peak!

then the PSU switched off for safe the backplane

For what it's worth I went ahead and upgraded my RTX 2060 Super with a RTX 3080 TI hoping to utilize the Magic Mask. Immediately shuts down the system without spending more than 5 seconds on creating the mask. Real disappointment.

I understand that some people with modest specs can slowly use the Magic Mask. Even if I could slowly use it would be better than the system just shut down. Hell, I just spent the money for an overpriced GPU. Maybe in the Blackmagic Preferences, there could be a throttle for Magic Mask use?

Specs in the signature.

I think that Dell has a 1300W power supply so I'd be surprised if it was shutting down the the load of a single RTX3080Ti.

Love to know what's causing these issues because I was *almost* about to install Resolve 18 to have a play but I can't afford to be jumping back and forth if it's too flakey

Two possible issues.

One 1300 sounds huge and is huge but you need to check if it's all available for the GPU or if it's spilt off. The label should show how the 12V is setup providing how many amps.

The other thing is many of these cards have very high spike draws. Older PSU see that being a danger and shut down. Newer PSUs have been designed to deal with the spikes.

Do you have any software installed that creates a virtual display? Another user had this issue while Oculus Virtual Desktop Streamer was installed.

Nick2021 wrote:Two possible issues.
One 1300 sounds huge and is huge but you need to check if it's all available for the GPU or if it's spilt off. The label should show how the 12V is setup providing how many amps.
The other thing is many of these cards have very high spike draws. Older PSU see that being a danger and shut down. Newer PSUs have been designed to deal with the spikes.

Hey there Nick2021, I'm not really sure what I am to look for regarding the amps. I went ahead and got an image of the PSU as well as the specs. If it isn't too much, could you look at this?

https://i.ebayimg.com/images/g/6ykAAOSw ... -l1600.jpg

https://www.dell.com/support/manuals/en ... cdf23ce58f

Is it possible that even though the machine seems to run otherwise, there could be a malfunction aside from inadequate specs? Otherwise, I suppose I could go with a Thunderbolt3 eGPU case, right?

To me that looks like your 12V is split into multiple 18A lines. That's 216watts. Your card at full power needs almost twice that.

Moving the card to a thunderbolt box might get it to run but it makes the port a potential new bottleneck.

Nick2021 wrote:To me that looks like your 12V is split into multiple 18A lines. That's 216watts. Your card at full power needs almost twice that.

Thank you for that! At least I have some direction to solve my GPU issue.

One more thought. You're using multiple 8 pins I think. Make sure they aren't all on one cable.

If Dell was nice they've labeled everything. Pick different PCI cables and see if that helps.

Nick2021 wrote:You're using multiple 8 pins I think. Make sure they aren't all on one cable.

Nick2021, you are exactly right! I had a Y cable which was part of the problem. And the rest was solved by digging around the back panel to reveal the power distribution card. There I noticed that there are three 8pin VGA power plugs, two for CPU 1 (and one of those is an 8pin connector cable being converted to two 6pin connectors which I removed the cable) and then a third 8pin VGA power connector and cable for CPU 2 which I moved to replace now empty secondary 8pin VGA power connector and to be closer to the GPU. That takes care of the proper cabling for my 3080 TI (mine requires only two 8pins).

The second step to ensure no crashes was to add the free MSI Afterburner software and 'undervolt' it (YouTube it if unfamiliar, they walk you thru it, and is very easy), which is recommended anyhow to optimize the power use of the system, especially for when it is closer to idle. But also helps prevent the spiking that can occur which can cause a shutdown.

So, the system probably isn't reaching its full potential, but in a pinch, this got me back up and running.

I have catched the same trouble. In my case PSu with good enough for my graphics card. And any stress-testing passed by successfully, but magic mask anyway crashing my system. I thought it's a driver issue but any actions in this way didn't solved the issue. As Stephen Buckley said the solution:
Step 1 - check powering of the GPU, if everything fine with that, then
Step 2 - Use an "undervolting of GPU" (google for that).
The situation is treacherous because it happens only with magic mask in Da-Vinci. Any renderings/stress-testings/games will works fine on the system, but not magic mask.