exact same as you
From what Iām seeing, nothing renders faster than 0.2fps on the new release. I reverted back to 3.0.12 and rendered a 10-second clip; itās processing at 0.8fps. Not a barn-burner but itās 4 times faster. Iām opening a support ticket to see if the techs can help me. My GPU is nothing special (Nvidia GTX 1650) but I wouldnāt expect performance to be 1/4th what it was under the previous release.
After over a day using the updated version, Iām pleased by the improvements in speed. As noted by others, v. 3.1 makes greater utilization of my workstationās CPU/GPU. When running more than one process at a time, the software seems to prioritize the shortest clip first, and my station is pretty much maxed out running the program.
One caution, my settings from v. 3.0 for the AI parameters are now all too strong; at least for the Proteus model.
I would like to mention that in Germany these ālicensesā are handled different. In Germany even the highest court often explained that for most users the software has the same sort of value than a physical thing, and in fact of this they said that we do not only buy a copy with some rights of usage. in that example they said: āyou buy āMicrosoft Wordāā. and with the buy all rights of usage at this single piece of product is being handed over to the new owner who is of course also legitimated to sell his software someday.
1 Like
From my testing Chronos Fast is 4x slower on GTX cards on this release than 3.0.12. Are you using Chronos Fast by chance?
FWIW, I am currently getting 7.6fps with my RTX 3060 (12GB) in v.3.1. Hopefully, support will come through for you.
So far Iāve used version 2.6.4, the other versions 3.0.x were all slower.
The 3.10 version is not as fast as the 2.6.4 version,
it is faaaaaaaaaaaaaaaaster! (approx. 2x as fast). Very good!
My system: Ryzen 5 2400g, RTX 2060.
What I noticed: If I minimize the GUI, then about 10% less CPU is used! The GUI is always flickering slightly, it seems like it is constantly repainting and using some system resources to do so.
Previously āGraphics_1ā (Task Manager) was used, now CUDA for about 50%. Possibly there is still room for improvement, the GPU still doesnāt seem to be fully utilized.
1 Like
I have a GTX 1660, and the speed is mediocre for me too.
OK. Fair point. Bleeding edge. But there should be reasonable standards even for latest tech. I mean if you really want to make an argument for a bleeding edge being a privilege to use. I can make an counter argument that being beta testers for Topaz team is equal payment. Bugs for feedback. Feedback for bugs. Fair deal in my view.
On the other hand if its going to be charged at premium rate as finished commercially ready product, than it should reflect the advertisement as such. Iām not being controversial here. That is how the practice was done until fairly recently. In it in recent times with digital version of software only and subscription models that trend has shifted in rapid release cycle to a captive audience. There was rapid release cycle in the past. But it was called, alpha, beta, beta 2, beta 3 etc. We have the same thing today, except the beta part is pronounced silent. Bleeding edge technology is fine, but reliably should be a priority.
1 Like
Or you can look at it other way. Old perpetual license meant you were buying a finish product. Now you get one 1 year license to participate in beta program. Seriously, there should be no company that releases a new major program for at least a year, because that is how long it takes to actually make a product that is out of the beta phase. Releasing a version 3, rewritten from scratch is not a major new release its a beta program version 1 of a new product that is far from prime time ready. As the Topaz stated officially the old version up to version 2.6.4 was benefiting from longer testing and optimization process. Version 3 is beta in all but name. And its really more like version 1 of a new program inspired by what they have done with Topaz Video Enhance up to version 2.6.4.
The fact that Topaz team build the old version with some limitations that they realized will limit progress forward and therefore decided to rewrite the app from scratch is a cost for the Topaz, it should not be passed down to consumer. Once version 3 or however you want to call it has been properly tested and is stable, they could release it. Otherwise you are not paying 1 year for upgrades, you are paying 1 year to participate in beta program.
5 Likes
Mediocre is being generous. To me itās unusable.
1 Like
āOnlyā 7.6fps??? 
The best I get is 12.5fps with my RTX3060 (12GB), upscaling 1 video 2x (1280x720 ā 2560x1440) with preset Artemis/Proteus, GPU load about 80%, GPU memory at 2.7/12GB.
However when I do same thing with 1 video upscaling 1,33X (1920x1080 ā 2560x1440) speed drops to 4.34fps with same preset as above.
To run 2 videos in parallel with the RTX3060 is possible but not recommended, because processing time is more than double the time whe running both in sequence.
Iāll test more odd upscale ratios in order to find out which scaling perfoms best.
Is Chronos Fast used if youāre not changing the FPS rate? The Frame Interpolation filter (the only place I see Chronos Fast) is only active if I change the FPS. I do use that if Iām changing the frame rate and it didnāt seem unduly slow on the prior release.
Thanks; I hope so too. Iām not feeling too good about having just paid for a license renewal to a release I canāt use. I know this is very complex software but I feel like Iām beta testing it. Itās not exactly inexpensive.
Chronos Fast is only used if you change the frame rate.
Thatās what I thought and Iām not doing that.
Ok I should add that if you are using a 4k source and changing the frame rate by 2x you will see itās more than 4x slower than on the 3.0.12 release - specific to GTX cards.
This is all true. The problem with TVAI is not that is was being developed, but that they released it too early. Then compounded their mistake by dropping support for v2.6.4 (VEAI). They should have kept v3 to the BETA TESTER group.
Or you can look at it the other way. 
Eventually (now) TVAI has overtaken VEAI in performance (drastically even). And after a few months of growing pains, I finally no longer feel the need to go back to using VEAI. Going from 1.8 fps to 7 fps is quite a stretch to ignore.
However, TVAI is still very much beta. Stabilization is outright broken, as well as motion deblur. Those 2 options should have been disabled for the release version.
All-in-all, I feel Topaz took a goodwill hit with the v3 stunt. Their commercial guys were clearly not in sync with the tech folks. I still feel confident in the dev team, though, as I see them working very hard on improving things ā and they actually have, significantly. So, Iām still happy with my decision to extend my licence for another year (back in October).
7 Likes
I would love to know what changed between 3.0(.12) and 3.1 to give such a big performance boost tbh. Wonder if thereās more things in that specific area that can be done too, that will just take more time
My cursory understanding of what suraj explained (donāt shoot me if I am wrong), was that they reworked the entire Tensor core logic (primarily to fix performance on 40xx cards); and in the process actually initially caused a major performance drop on 3080 series card (or had their Tensor cores flat-out not work at all). Then, with the advent of 3.1.0.4.b, they finally worked out the kinks. And also simply optimized TVAI to take much better use of your CPU and GPU.
