Dual GPU, how much memory will be?

Hi, Resolve sometimes fill whole 6GB GPU memory. I dont want optimize effects and resolution, because it runs realtime, so it is solution buy second (same model) GPU?
In SLI it use memory only of one card, but Resolve manage cards independently. So do I will have with two cards available 12GB or compute power of two, but only 6GB memory? Thanks

Hi, Resolve sometimes fill whole 6GB GPU memory. I dont want optimize effects and resolution, because it runs realtime, so it is solution buy second (same model) GPU?
In SLI it use memory only of one card, but Resolve manage cards independently. So do I will have with two cards available 12GB or compute power of two, but only 6GB memory? Thanks

6GB for 4k is not enough.
2x 6GB won't help you with the memory issue, Resolve sees 6GB

Try to get a Radeon VII for around $700, it has 16 GB of blazing fast memory and is currently the fastest GPU in Resolve. https://www.pugetsystems.com/labs/articles/DaVinci-Resolve-15-AMD-Radeon-VII-16GB-Performance-1382/

When there are more than 1 GPU, DaVinci Resolve uses the memory of only 1 GPU, the one with lower amount of memory. In order to have more GPU memory the GPU card(s) needs be replaced.

Rather, I expect it will use memory in both cards, but they won't necessarily process the same payload (a video frame for instance). That's why you can't think of it as doubling the memory and why the smallest one will still be the bottleneck.

Ok, thanks all for reply. So I will upgrade to new GPU next year.

Does Davinci 17 have better memory handling or is it still the same?

Also, If you have two graphic cards and put one for gui. Say you have 24GB for gpu, and 11GB memory on the GUI card. WIll Davinci still only use 11GB "all in all". Or how does it work?

Would love to not spend the money on the rtx3090 and go with two 12GB cards instead.

What has been said above is still valid for 17.

Uli, it's still not clear to me exactly how it works.
I make some example of memory's. How will Davinci use it?
Example 1
One gui card 11GB + gpu card 24GB.

Example 2
One gui card 11GB+ one gpu 16GB + second gpu 24GB

In both cases, Resolve will be limited to 11 GB of VRAM.

With recent versions of Resolve and any modern GPU a separate GUI card makes no sense.
That's a concept from 5 years ago or older.

In both cases, Resolve will be limited to 11 GB of VRAM.

GUI card is not involved in processing, do you have any source from BMD which explicitly states that GUI card limits processing mem?

I don't know if that was ever made clear, to be honest. I've only seen them say that Resolve is limited to smallest amount of available VRAM.

Yes, in the context of processing. What gui card does though is it renders just the GUI (obviously) and thus only needs the final view buffer from processing, not all sources.

So, I'd suggest doing some research by changing cards and configs and find out.

It's a valid theory, Hendrik. I don't know if it's correct.

This topic may be useful: viewtopic.php?f=21&t=89574&hilit=multiple+gpu

[Peter Chamberlain]

Some incorrect infö above.

Resolve uses the GPU memory on every card it has in the system that is selected for Resolve use.
Generally it doesn't share memory to other cards.. lets not get into exceptions.
If you have a GPU in the system but its not selected in preferences, its just a waste of space.
If you have mis matched GPUs, the slowest GPU will define how fast the system can process.
If you have GPUs with difference memory capacity, the lowest capacity will define how big an image and the type of processing before its runs out of memory. The cards with extra memory waste that memory.
If you have a separate UI GPU, that's something we generally don't favor now, whereas we did 10 years ago.
If you have one more powerful and more memory GPU compared to a second, use the better one for the UI but not as UI only. It might make a little difference if the gpu performance is close.
One very powerful GPU with lots of RAM is better than two average GPUs, for many reasons....
Anyone who says multiple GPUs are a waste of time are picking a use case that doesn't need them, after all the large facilities that put 4 x 24GB GPUS in their system are spending that money for fun.
Most users will be happy with one good GPU.
All image processing is in GPU.
Some new GPUs, with the Resolve Studio version. also support compression and decompression in a hardware stage in the GPU.
GPU driver selection makes a big difference in reliability.

Thank you for the clarification, Peter. My first RTX 3090 with 24GB of VRAM should arrive soon.
I am thinking about putting a second one into the system

Could you give an example where 2 or more GPUs are beneficial?

Some incorrect infö above.

Resolve uses the GPU memory on every card it has in the system that is selected for Resolve use.
Generally it doesn't share memory to other cards.. lets not get into exceptions.
If you have a GPU in the system but its not selected in preferences, its just a waste of space.
If you have mis matched GPUs, the slowest GPU will define how fast the system can process.
If you have GPUs with difference memory capacity, the lowest capacity will define how big an image and the type of processing before its runs out of memory. The cards with extra memory waste that memory.
If you have a separate UI GPU, that's something we generally don't favor now, whereas we did 10 years ago.
If you have one more powerful and more memory GPU compared to a second, use the better one for the UI but not as UI only. It might make a little difference if the gpu performance is close.
One very powerful GPU with lots of RAM is better than two average GPUs, for many reasons....
Anyone who says multiple GPUs are a waste of time are picking a use case that doesn't need them, after all the large facilities that put 4 x 24GB GPUS in their system are spending that money for fun.
Most users will be happy with one good GPU.
All image processing is in GPU.
Some new GPUs, with the Resolve Studio version. also support compression and decompression in a hardware stage in the GPU.
GPU driver selection makes a big difference in reliability.

So, if there is a gpu selected for gui only, does it limit the memory use of processing cards or not? It is a technical question, not what users should prefer to use or which configuration is faster.

Could you give an example where 2 or more GPUs are beneficial?

I was wondering that, too.

if there is a gpu selected for gui only, does it limit the memory use of processing cards or not?

"[Resolve] doesn't share memory to other cards.. lets not get into exceptions."

"[Resolve] doesn't share memory to other cards.. lets not get into exceptions."

What has this got to do with sharing memory? Two equal gpus for processing don't share memory to each other either. Everything above single gpu is an exception.

I take this as if you want to use multiple gpus. Make sure they have the same amount of memory. Likely is best you get two identical cards. Then resolve will use the both cards full memories.

Two cards might make sense if each frame takes a long time, like with heavy temporal noise reduction. But if I understand correctly, two cards might slow down faster operations because of all the copying back and forth.

I have just been offered a used RX 580 8G to accompany my existing one (new ones are $400-500). ) I'd primarily need it for 3 frame TNR or Super Resolution to upscale 1080p to 2160p. I'm currently getting 13 fps with the latter. Would the second RX 580 help push the playback framerate to 24-25 fps or the scaling wouldn't be that linear? The 2 PCI-E slots will work at x8 in that case.

I'll also need a more powerful PSU, of course.

You can expect about 80% gain for TNR.

You can expect about 80% gain for TNR.

Thanks, Uli. If Super scale scales the same, that would mean I won't be able to hit real time with it.

However, I 'm being stupid, Super scale is really only relevant at render stage.