Hi,
it would be nice to have Multi GPU Support.
Speed is always critical, and so is quality, of course.
It would be great if each GPU could work on its own image, then you would have a +100% speedup.
Capture One and all apps from Topazlabs have had this for years and especially with Deep Prime it would be very useful.
People could then buy two cheap gpus and get the speed of one big one.
MrJynx, your feature request for multiple GPU optimisation is an interesting one. In general, when DeepPrime does have a properly optimised GPU it’s plenty fast. A bigger issue is not every GPU is supported and those that are supported are often not well-optimised. What this means is that hugely powerful high memory GPU’s like the top-end Nvidia family do not process any faster than an old GTX960 or GTX1050.
Many times a GPU is not supported at all or GPU acceleration is not supported at all on Mac (after years of bitter recrimination, the latest Photolab v4 seems to have solved the no Mac-GPU optimisation issue).
Better optimisation for single GPU is where most of us DxO Photolab “old timers” have pinned our hopes.
This does not make your feature request any less valid. This is just to share with you the context into which it fits. I’m not voting for multi-GPU optimisation now as single-GPU optimisation seems far more important at this point. I have a Radeon VII in one Mac Pro 5,1 with 16 GB of VRM and a Radeon RX 580 with 8 GB of VRAM in the second Mac Pro. I can’t imagine adding a second GPU to either. Either GPU is plenty powerful to render DeepPrime extremely quickly with proper code optimisation and full exploitation of the capabilities of the GPU.
This reminds me of the situation with Adobe Photoshop and multi-CPU optimisation. It turned out Photoshop could load all the processors but that after 4 processors each processor added actually slowed rendering!
I wouldn’t give Topaz as an example of efficient image processing. All of Topaz’s software runs very slowly and has always been a bit of a workflow nightmare. Happily Photolab 4 is much more fluid in hand. One trick I like is to only turn on Noise Reduction right before final render (of course I’ll test an image or two for noise reduction to have an idea of how much help it will be).
With Denoise AI my two Quadro RTX 5000 GPUs do process a 1DX III image in 2 to 4 seconds.
Capture One does export an image of a 1DX I/II/III in 0.7 seconds with these two GPUs.
CPU is a 24 Core Threadripper 3960X.
I use PL4 only as a Denoiser.
I convert (Denoise) images at the same time as i do develope them in Capture One.
The Nvidia cards from the Turing generation onwards can run several programs simultaneously without the graphics part stuttering.
Sounds like Topaz has made some progress with performance, thank heavens, albeit the hardware requirements as ever are absurd. You are unlikely to achieve performance anything like this with DxO Photolab. Prime Noise reduction with 5DSR files was something like 30 seconds per file. DeepPrime brought processing D850 files down to about 8 to 15 seconds depending on the crop. Of course your file sizes are much smaller so performance would be better.
I’m surprised you require a workflow where background processing of finished files requires real time. I’m also surprised you’d use Photolab 4 just for denoising when you already have a real time workflow with Topaz Labs DeNoise. I do prefer Photolab Prime and DeepPrime over anything Topaz Labs has to offer, as both look much less artificial than DeNoise to my eye but others don’t seem to notice such a difference.