Gigapixel 4.4.0 CPU vs GPU Different Results

Topaz Products Product Technical Support
21

Andy, Using Sharpen AI 1.4.0 the auto system used my GPU. I have a Radeon RX 580 (4GB memory) GPU and a Ryzen 5 2600X (6 core, 12 thread) CPU. Using Sharpen the time was only 14 sec for GPU and memory high but that is not really a good test. I got the same time for Stabilize and Focus which are much slower.
Method --------Time (Focus algo) on 3456 x 4608 Jpg
GPU_low 1:47 (min & sec)
GPU _high 1:09
CPU VINO on 1:44
CPU VINO off 4:34

Even though the Intel VINO optimization was made for Intel CPUs and I have an AMD Ryzen, it was much faster with it on. I would recommend anyone that is using their CPU and if it is an AMD Ryzen, to turn this on. You didn’t state if your test was with the Sharpen setting but from the times I would guess it was.

I haven’t tested Gigapixel but the memory labels may be wrong.

22

Yes, I was doing a simple sharpening with the Auto values.

23

I reran these tests with Gigapixel AI 4.4.3 and Sharpen AI 1.4.2 and the timings remain essentially the same.

24

I reran these tests with Gigapixel AI 4.4.4 and the timings remain exactly the same.The banding issue when using Max Quality AI Models with the test picture above is also unchanged. It appears that these recent 3rd tier updates are just attempts to resolve the various user interface issues, and are not adding to the machine learning or optimizing/improving the processing code. I’ll test it again when/if we ever get to version 4.5.0, or 5.0.0.

25

@ weasel-126738: I noticed this too and using the CPU seems to be the best option for me to get the most sharpen image with as many details as possible.

26

I tested the Zion canyon image again with the new version 4.5.

I do see some speed improvements, but I’m running a different CPU/GPU combination now so it’s not Apples to Apples.

The banding issue when using Max Quality AI is gone!

Also, the anomalous finding that Medium ran faster than the High memory usage setting is gone as well.