Your GPU Memory is full.

Hello Mathieu. You can choose your render speed here.

@Charles Bennett

Oh, thanks. I don't understand I had never noticed that one.

Unfortunately, I updated my drivers (and windows) and Resolve keep crashing now... I can't even reach the delivery page...

What people suggest that minimum is 8 GB of VRAM, is crazy. Sorry, but when you have a $300 Panasonic camera shooting 4k, and it's required of you to get a $800-$1500 graphics card to edit that 4k video (not counting the cost of the rest of the PC), obviously something is not right. Other software solutions can do just fine with 2 GB of VRAM for 4k, and I hope Resolve gets optimized. I personally just upgraded to a 4 GB nvidia card and I have no intention of spending more money on a new one. Particularly since Resolve 15 worked fine with it. I need Resolve 16 to be able to do that too, because they were released only 1 year apart. I expect the same type of system requirements -- not doubled.

ASRock Phantom Gaming D Radeon RX 570 DirectX 12 RX570 8G OC 8GB 256-Bit GDDR5 PCI Express 3.0 x16 HDCP Ready Video Card $140

https://www.newegg.com/Product/Product.aspx?Item=N82E16814930008&ignorebbr=1

Also window 8 is not recommended, w10 is.

Hi guys,

I just updated to DR 16 (Beta 1). And now I can't export timelines I was exporting in DR 15 without getting random "The GPU failed to perform image processing because of an error. Error Code: 0". The render doesn't stop but usually one frame will have glitch, which is so annoying I wish it would re-render the faulty clip, instead of having to backtrace which one has failed.

Here is my setup:

- Windows 10 (latest version)
- 2 x NVIDIA Titan X 12GB (Latest NVIDIA drivers as of today)
- 64GB DDR4 RAM

The timeline I'm rendering is part r3d 6k/8k and 5K CinemaDNG. I'm round-tripping to Premiere so I'm exporting at source resolution, which means 8k export for some files.

Even when I drop the render speed to 2 frames per second it will fail 2-3 clips... When I put the render at Maximum speed it will be close to 14 fps with 5-10 failed clips on a 200 clips timeline. I was able to get ~10 fps with DR 15 and no failed clips.

Any help is welcome! I could not find any info regarding "Error Code: 0". I also tried both CUDA and OpenCL mode with same results, OpenCL just being slower.

Thank you.

Hi guys,

I just updated to DR 16 (Beta 1). And now I can't export timelines I was exporting in DR 15 without getting random "The GPU failed to perform image processing because of an error. Error Code: 0". The render doesn't stop but usually one frame will have glitch, which is so annoying I wish it would re-render the faulty clip, instead of having to backtrace which one has failed.

Here is my setup:

- Windows 10 (latest version)
- 2 x NVIDIA Titan X 12GB (Latest NVIDIA drivers as of today)
- 64GB DDR4 RAM

The timeline I'm rendering is part r3d 6k/8k and 5K CinemaDNG. I'm round-tripping to Premiere so I'm exporting at source resolution, which means 8k export for some files.

Even when I drop the render speed to 2 frames per second it will fail 2-3 clips... When I put the render at Maximum speed it will be close to 14 fps with 5-10 failed clips on a 200 clips timeline. I was able to get ~10 fps with DR 15 and no failed clips.

Any help is welcome! I could not find any info regarding "Error Code: 0". I also tried both CUDA and OpenCL mode with same results, OpenCL just being slower.

Thank you.

I hope you have a backup of your database or at least project.
Downgrade to 15.3.1 and continue working with it.

16 Beta chews up on Vram for whatever reason - on both Linux and Windows. Doesn't seem to be an issue on MacOS installs.

I hope you have a backup of your database or at least project.
Downgrade to 15.3.1 and continue working with it.

16 Beta chews up on Vram for whatever reason - on both Linux and Windows. Doesn't seem to be an issue on MacOS installs.

Oh got it, hope they fix it asap! I already have few days of work within DR 16... I should have tried to export earlier... Right now I'm re-exporting the glitchy clips post-render. It should have at least an option to re-export failed ones or a list of the failed clips. Right now the only way is to watch each single exported clips to detect glitches.

But I wonder why you choose a nVidea Graphics Card with only 4 GB of vRam, when you can get a cheep AMD Graphics card with 8 GB of vRam?

Because nvidia is more standardized with cuda and other extensions. It's a more sure bet when it comes to API support than AMD is.

But I wonder why you choose a nVidea Graphics Card with only 4 GB of vRam, when you can get a cheep AMD Graphics card with 8 GB of vRam?

Because nvidia is more standardized with cuda and other extensions. It's a more sure bet when it comes to API support than AMD is.

Freedom of choice, not required by Davinci Resolve.
Cuda is a closed system from NVidia, the open standard is OpenCL supported by all GPU makers.

But I wonder why you choose a nVidea Graphics Card with only 4 GB of vRam, when you can get a cheep AMD Graphics card with 8 GB of vRam?

Because nvidia is more standardized with cuda and other extensions. It's a more sure bet when it comes to API support than AMD is.

Freedom of choice, not required by Davinci Resolve.
Cuda is a closed system from NVidia, the open standard is OpenCL supported by all GPU makers.

Nvidia supports OpenCL.

AMD does not support CUDA.

So yes, OpenCL is an open standard... but Nvidia's OpenCL performance isn't exactly trash... and going AMD removes the ability to use a superior API on superior GPU hardware.

Going Nvidia actually gives you more choice, regardless of what you think of CUDA and it's "closed nature." OpenCL is still availability.

OpenCL is also deprecated on macOS.

Because nvidia is more standardized with cuda and other extensions. It's a more sure bet when it comes to API support than AMD is.

Freedom of choice, not required by Davinci Resolve.
Cuda is a closed system from NVidia, the open standard is OpenCL supported by all GPU makers.

Nvidia supports OpenCL.

AMD does not support CUDA.

So yes, OpenCL is an open standard... but Nvidia's OpenCL performance isn't exactly trash... and going AMD removes the ability to use a superior API on superior GPU hardware.

Going Nvidia actually gives you more choice, regardless of what you think of CUDA and it's "closed nature." OpenCL is still availability.

OpenCL is also deprecated on macOS.

How is NVidia's Metal support these days? AMD's is 100%.
Metal is from the minor player Apple...

[quote="Eugenia Loli"]

Because nvidia is more standardized with cuda and other extensions. It's a more sure bet when it comes to API support than AMD is.

Freedom of choice, not required by Davinci Resolve.
Cuda is a closed system from NVidia, the open standard is OpenCL supported by all GPU makers.

Nvidia supports OpenCL.

AMD does not support CUDA.

So yes, OpenCL is an open standard... but Nvidia's OpenCL performance isn't exactly trash... and going AMD removes the ability to use a superior API on superior GPU hardware.

Going Nvidia actually gives you more choice, regardless of what you think of CUDA and it's "closed nature." OpenCL is still availability.

OpenCL is also deprecated on macOS.

How is NVidia's Metal support these days? AMD's is 100%.
Metal is from the minor player Apple...[/quote]Apple ships Intel and AMD GPUs. For most people, it doesn't matter.

Sent from my SM-N960U using Tapatalk

Quite a few of us have had this problem with version 16. The cure was to update the GPU driver with the latest version. You will find the latest version on the Nvidia website. If given a choice go for the Creator version of the driver.

I already had the latest GPU creator version driver before updating to DR 16 but still get the memory full errors (I only have 4GB gpu memory). I'm not sure there is any "one thing" we can do at the moment to fix this on our end. One other thing that did help to reduce the number of errors I was getting was to set my preferences to use cuda cores vs auto. I still get errors but not nearly as often.

Quite a few of us have had this problem with version 16. The cure was to update the GPU driver with the latest version. You will find the latest version on the Nvidia website. If given a choice go for the Creator version of the driver.

I already had the latest GPU creator version driver before updating to DR 16 but still get the memory full errors (I only have 4GB gpu memory). I'm not sure there is any "one thing" we can do at the moment to fix this on our end. One other thing that did help to reduce the number of errors I was getting was to set my preferences to use cuda cores vs auto. I still get errors but not nearly as often.

NVidia is the best and when it doesn't work the way you want, just buy more of it.

And when it still doesn't work the way you want, don't blame NVidia.

I have 10 seconds timeline, only Deflicker and Stabilizer enabled.
The same warning on Windows computer with 2x1080 GTX 8GB each (and 64 RAM)
and the same situation on iMac with Radeon 4GB (and 40GB RAM)

So...

Same issue with DVR 16 and DVR15
on Windows computer with Nvidia GTX 2x1080 8GB Each, 64GB RAM
on iMac with Radeon 4GB with 40GB RAM

10 sec timeline, Deflicker and Stabilizer enabled, no color corrections

Just had this on a single clip (the others were fine) even with grading disabled. Turns out that clip was set to Super Scale 4x, which was clearly flooding the GPU memory (could see it on Task Manager). Dropping to 2x (or disabling it) fixed it.

Strange thing is, it must have worked when I first enabled Super Scale on that clip (on the same system) - but not when I reloaded the projected and hit it.