See this topic:
@TorsteinH I don’t normally sit and watch but that particular new element took long enough that it caught my eye!
The image is here, if I can find it again, @stuck beat me to it.
It is unlikely that you will have any problems because most haven’t, except me!
I was concerned that the DOPs I was using had been corrupted by PL9.x. so I took some I had been using on the 3060 system over to the 5060Ti and replaced all the test directories there and as soon as I started testing XD3 I got an error!?
It could be the build (OS) on that machine.
One way to test that would be to swap the GPUs, a somewhat tedious exercise but I might have to give that a try!?
So the weather is a bit colder this afternoon so I decided to open the computer cases to get a bit of extra warmth!!
At the same time I swapped the GPUs and reran the tests with the following results, both tests worked i.e. the 5900X with the 5600G and the 3060 with the 5900X.
So work that one out!?
The 5900X is about 2 times the CPU power of the 5600G and the 5060TI(16GB) is about 1.7 times the power of the 3060(12GB)!?
The runs times were 3 minutes 57s (5900X-3060) versus 3min 17s (5600G-5060Ti), longer runs are needed for a better evaluation.
I believe you overestimate the extra computing power of 5900x vs 5600G, according to this site it’s a bit over 10 %: AMD Ryzen 9 5900X vs AMD Ryzen 5 5600G - Benchmark, comparison and differences
And by the way, the St.Pancras photo worked just perfect on my PC with a 5700G and a rtx5060TI 16GB
@TorsteinH That is for single core performance not multi-core performance.
From that site you quoted
and other benchmarks put the multi-core performance at almost exactly 2 times.
Your hardware is very close to mine in specification but mine has consistently failed until the tests I published and that was with the 5060Ti working with the 5600G rather than the 5900X.
HOWEVER, I did further tests with the 5900X-3060 combination and things went downhill very rapidly.
I have a test directory where each edit is applied directly to a separate copy of the image, rather than to a VC of the image and ran that test on the 5900X-3060 twice, and got a new failure, the export worker was terminated after the first failed test and before running the second test which also failed.
I then created a new test directory and that failed in a number of different ways. I will continue those tests tomorrow.
Update:-
The reason why the 5900X has such a limited impact on the overall time is shown in the graphs I submitted, i.e. the graph shows limited peaks of processor activity indicating that there is limited use of the processor.
So having a processor that is capable of being twice as fast if all cores are in use is fine but the CPU usage is limited and sporadic so much of that CPU capability isn’t being used for much of the time, i.e. when the GPU is doing much of the work.
I apologise for the “weird” colours in the snapshot, I have been struggling with a colour issue for some time on the 5900X with the 5060TI but the snapshots taken with that combination appeared fine when viewed on my other machine(s).
However, swapping the 3060 back into the 5900X, where it was fitted until the 5060Ti displaced it in September. has made the problem even worse and the snapshots are now “contaminated” as well!