Recently I upgraded my old machine (well, actually replaced it) with an AMD Ryzen 5 7600X. Initially with my old AMD RX 480. That already cut the XD processing time more or less in half compared to my old i5-6400 combined with the RX 480. Processing a single M50 file (DxO Standard + XD) went down from just less than a minute to approx 25s.
Even more recently I replaced the RX 480 with a AMD RX 6800XT. I processed the D850 files from the sheet with DxO Standard and XD: 25s. Unfortunately this card somehow did on me within a couple of days…
It was replaced with the NVIDIA RTX 4070. Processing the same D850 files now only takes 21s.
If you haven’t already made up your mind, this maybe helps. Please keep in mind: I’m using a desktop. Desktop hardware in most (if not all) cases is more powerful than the laptop variants.
The AMD RX 480 is a PCI-e 3.0 x16 card and the motherboard is initially was used on (MSI B150 Bazooka) and the processor (i5 6400) supported that too.
My new motherboard supports PCI-e 4.0 x16, but the RX 480 doesn’t so it won’t go any faster on the new motherboard. The PCI-e slot is simply clocked down to PCI-e 3.0 x16. Next to that, the GPU is only used for noise reduction, the CPU is doing the rest of the work.
So in this case the increased performance was due to the much faster processor.
Did you test or can you still test what those difference are between those 2 stations when :
1 - Only denoising correction (deepprime, deepprimeXD) is applied ?
2 - Other corrections but no denoising at all is applied ?
Just to understand which corrections are cpu or gpu intensive.
Ai Denoising is of course gpu intensive but the other ?
What would be interesting to know too, is how scalable are cpu intensive corrections on multicore cpu (do they fully use any number of cores) ?
Unfortunately I can’t anymore. The old desktop has been sold already. But the results of my old machine are in the sheet I mentioned earlier in this thread.
The Egypte image was meant to be processed using CPU only, with DxO Standard and DeepPrime (non XD). On my i5 6400 it took 185 seconds with CPU only and 36 with CPU + GPU.
Some results on my new machine (Ryzen 5 7600X + RTX 4070) CPU Only:
DxO Standard + DeepPrime XD: 263 seconds
Dxo Standard + DeepPrime: 65 seconds (120 seconds improvement over the i5 6400)
DxO Standard, no NR: 3 seconds
CPU + GPU
DxO Standard + DeepPrime XD: 7 seconds
Dxo Standard + DeepPrime: 4 seconds (32s improvement over the i5 6400/RX 480)
DxO Standard, no NR: 3 seconds
As can be seen in these results and also is mentioned in the preferences, the GPU is only used for DeepPrime acceleration.