Videoproc 7 / IA Gen 2 / HQ / GOP 24 (17 fps = GPU 100%)
Topaz Video 5 THM12 / AHQ12 / GHQ5 (2.3 fps = GPU 97%)
If I don’t force details, I got absolute no stars visible on bulb
The problem is they’ll put out a beta version, then sometimes literally a week or less later release it as a public version. a week for a beta test is useless, especially when 99% of beta testers aren’t in it to do free labor, we just want a new toy to play with.
IMO they need two things:
Also got this out of Topaz
1109 fps is huge ^^ and please let us tune HEVC … Main @ 6.2 => You need a Cray to display this correctly
“Low”, “Medium” and “High” is for general users. Considering licence price, we should get tunable things (like VSDC 
)
Agreed. A “release” version should be a stable build based on an already nearly stable beta that has had some period of intensive bugfixing with no new features being added. Buggy functions disabled if necessary. Nothing labeled “experimental.”
Releases with new features should probably happen no more often than quarterly. Anything in between should be bugfixes only.
The current scheme is really a lot of alphas and betas, with nothing that can justifiably be called a release.
The main problem is everybody doing the same thing
No one want to fail the IA run
See ChatGPT how “dumb” he became self learning with humans
I dont think that is the right input bc the data from space is very different, the output of JWST is very noisy, it was designed in 2003?
I can spot the technology behind it, looks like a raw output from canon in the 2007 era.
If JWST woud be based on actual technology its output would be much better.
I don’t know clearly where to take data, but JWST is all but noisy, artifacts are only to its structure with 18 separate pannels. Also NASA/ESA delivers RAW data, untouched, scientists do their owns tuning. Allmost all Hubble shots are computer enhanced whish is easy because taken in visible spectrum.
Also near infarred is all but accurate wave length for detailed image, but it vastly compensate it with its 6.5m main mirror and its berylium coating.
Try to compare Orion nebula from Hubble / JWST, there is a world between them
Voyager 1 is still able to send data with 8Ko of memory and a CPU that is just able to run a car button of now, but those data are however priceless 
But this is another discussion
Exactly what i said.
JWST is based on Technology in 2003 and Hubble is based on technology in 1978.
So if JWST would be based on tech from 2013 it would deliver a decade better images.
Project started in 1989, nirCAM is 2013 tech
Hubble is at most with all updates @ 0.043’ rez (2002 upgrade)
JWST is @ 0.032’ rez
Anyway, we (I 
) do like space frames training ^^
Rieke, M. et al. 2003, SPIE, 4850, 478
NGST NIRCam Scientific Program and Design Concept
Perhaps this is why my benchmark on my MBP M2 Max came in so low this time around. I just complained about it in the benchmark thread 
Opened!
Hello everyone,
We are currently investigating some performance differences in 5.1, we plan to have an updated build released soon. Thanks for your reports.
These really should be getting worked out in the betas. Isn’t that why you opened them up to everyone, to get more beta testing input you could resolve before putting up releases?
For all we know there were last minute changes made to the release version that caused this and nobody would have been able to discover it until now.
But yes, this should have been caught before it was approved for release.
Fix incoming! Apologies for dismissing your initial report, I was on Win version which does not have the same issue.
I was told by AI (ChatGPT) that Topaz AI enhancer was the best.
If it’s not the case, I will estimate this is an AI conspiracy
The rise of Skynet !
The only way to know that would be if people running this release version who had also tested the betas could report on how the release compares with its betas.
