Seems like Grax continues to be a work in progress as it’s always kind of been. It’s tough to lose nebula detail like you showed a little past half way in. I still like NXT while linear with a small amount like 30 or 40 and then some Topaz at the very end so I can eliminate any remaining color or lum noise. Doing it in two smaller phases seems to be more efficient at least for me.
Hey Sascha. QQ. How long does it take to denoise, let's say a 24-megapixel image on a Mac? In PC with NVidia and CUDA GPU acceleration it takes 20-30 seconds. Without GPU acceleration it might take up to a minute.
On my 14” M3 Max MacBook Pro with 128GB of RAM and the maximum number of CPU and GOU cores, I ran the PixInsight benchmarks, and it came out around 22,000. I think that shows that PixInsight is not well optimized for the Mac, when the top performing machines are in the 40k range for the benchmark. But, I don’t know of a way to benchmark GraXpert versus NoiseXterminator. That would be an interesting test.
I have old PC using AMD Vega 64 gpu. Both run at same speed and takes seconds. Ditch the mac dude :p
Unfortunately, with the youtube compression, we can’t see a lot of things in the test…
Yes, I know, that is an extra-annoying issue….
Seems like Grax continues to be a work in progress as it’s always kind of been. It’s tough to lose nebula detail like you showed a little past half way in. I still like NXT while linear with a small amount like 30 or 40 and then some Topaz at the very end so I can eliminate any remaining color or lum noise. Doing it in two smaller phases seems to be more efficient at least for me.
I do exactly the same.
Thanks for this explanation. I use both tools.
its wrong to compare a basic pc, there both done in seconds on my pc using cuda ;)
Hey Sascha. QQ. How long does it take to denoise, let's say a 24-megapixel image on a Mac? In PC with NVidia and CUDA GPU acceleration it takes 20-30 seconds. Without GPU acceleration it might take up to a minute.
With NXT it takes me less than 10 sec, and more than a minute for GraXpert
On my 14” M3 Max MacBook Pro with 128GB of RAM and the maximum number of CPU and GOU cores, I ran the PixInsight benchmarks, and it came out around 22,000. I think that shows that PixInsight is not well optimized for the Mac, when the top performing machines are in the 40k range for the benchmark.
But, I don’t know of a way to benchmark GraXpert versus NoiseXterminator. That would be an interesting test.
I think Graxpert denoise is supposed to be used on linear data, and Noixe XT is to be used on stretched.
No, that is not accurate. NXT should as stated by Russ preferably be applied in linear stage, but can also be used in non-linear stage