Topaz Video AI Alpha v3.4.0.1.a

Another bug: after encoding a (longer) video, the app becomes very unresponsive, taking seconds after mouse clicks to react, sometimes even totally hanging with the need of a forced shut down.

EDIT: I only saw now that there’s a 3.4.0.1.a (as the App didn’t show an update available and it’s not even clearly mentioned in the first post - just when you look at the threads title you get an idea that there might be an update.

So all posts above are for the 3.4.0.0.a. I’ve yet to test the 3.4.0.1.a. But then, looking at the change log there weren’t any fixes for the Mac version anyways?

3.4.0.1.a still hangs after bigger encodes with Iris (on MacOS 13.4)

Mac Studio M1 Ultra 64GB RAM Ventura 13.5 - I’m seeing zero difference in rendering speed with this Alpha via the Iris model compared to the released 3.39 version, always around 10.4fps. I’m using a 1080p source, no upscale, just Iris Auto, noise = 1, Recover Original Detail = 25.

I’ve even tried the trick I used on previous betas (yes I know this is an Alpha!) which was to clean all the crud away using good old trusty Onyx, but it didn’t make a difference this time.

Thought I’d let you know @suraj as Iris speed improvement on Ventura was one of the items to test.

Is it just me or is there a green-tinge to the “after” preview?

See above: Iris Speedup is only there with SD and only up to 2x upscales.
FHD to 4K even is considerably slower.

Ah, yes, I had this in the previous beta, too.

BTW I submitted an idea to have all outputs selected automatically when you start an export:

Do you happen to know the version that this started occurring for you? We are working on the preview experience, this is a large project and we do not have a release date at this time.

Thank you for confirming that issue 1 is still occurring!

I cannot recreate this issue, can you share a screen recording?

Hi @ida.topazlabs ,

I posted that on the 3.3.6.0b thread. I never saw a 3.3.5, so I was on 3.3.4 until that (.6). Here’s what I said, and a link. BTW: It’s still happening on 3.4.0.1 alpha. Thanks for looking into it.

When getting to the 2nd to last frame in the preview, the cursor stays there and there is a > 30second wait for that last frame and the preview to start playing.

I do not want to blame your hardware, however, I am wondering if this is a combination of the hardware and the size of the video or the filter used, etc, I cannot reproduce this issue on an M1 Mac or on a 3060 RTX.

I get this when previewing. It’s done processing, but the preview is stuck at generating preview. ( it only happened one time, but thought I’d let you know )

1 Like

Hi Ida,

It’s happening on 720x486 ProRes source. And anything else I use on it.

I included the HW info to let you know. It’s actually almost state of the art save for the graphics card. I thought there could be some part of that, so I got tired of putting it off, and now have this RTX 4070, just delivered 10 minutes ago from FedEx. I have some other things to do first but should have this up and tested by tomorrow.

I’ll let you know…

:blush:

Gene

2 Likes

It seems like this new version fixes some of the performance regressions I noticed recently.
Can be connected to driver and/or TVAI version.

720x480

Topaz Video AI  v3.3.1
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.945 GB (Adrenalin 23.5.2)
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 720x480
Benchmark Results
Artemis		1X: 	91.29 fps 	2X: 	48.37 fps 	4X: 	12.85 fps 	
Proteus		1X: 	82.43 fps 	2X: 	45.30 fps 	4X: 	13.06 fps 	
Gaia		1X: 	57.15 fps 	2X: 	40.13 fps 	4X: 	16.34 fps 	
4X Slowmo		Apollo: 	141.18 fps 	APFast: 	262.84 fps 	Chronos: 	52.53 fps 	CHFast: 	97.76 fps

Topaz Video AI  v3.3.10
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.941 GB (Adrenalin 23.7.2)
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 720x480
Benchmark Results
Artemis		1X: 	78.59 fps 	2X: 	46.23 fps 	4X: 	14.78 fps 	
Proteus		1X: 	72.46 fps 	2X: 	37.01 fps 	4X: 	14.02 fps 	
Gaia		1X: 	55.34 fps 	2X: 	37.75 fps 	4X: 	19.29 fps 	
4X Slowmo		Apollo: 	127.41 fps 	APFast: 	262.03 fps 	Chronos: 	53.57 fps 	CHFast: 	90.16 fps 	

Topaz Video AI Alpha  v3.4.0.1.a
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.941 GB (Adrenalin 23.7.2)
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 720x480
Benchmark Results
Artemis		1X: 	94.12 fps 	2X: 	47.84 fps 	4X: 	14.04 fps 	
Proteus		1X: 	83.37 fps 	2X: 	42.90 fps 	4X: 	13.38 fps 	
Gaia		1X: 	62.58 fps 	2X: 	42.14 fps 	4X: 	15.10 fps 	
4X Slowmo		Apollo: 	141.57 fps 	APFast: 	278.99 fps 	Chronos: 	55.61 fps 	CHFast: 	106.44 fps 	

1920x1080

Topaz Video AI  v3.3.1
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.945 GB (Adrenalin 23.5.2)
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	20.78 fps 	2X: 	09.80 fps 	4X: 	02.36 fps 	
Proteus		1X: 	19.19 fps 	2X: 	08.69 fps 	4X: 	02.28 fps 	
Gaia		1X: 	10.73 fps 	2X: 	06.95 fps 	4X: 	02.55 fps 	
4X Slowmo		Apollo: 	17.19 fps 	APFast: 	38.65 fps 	Chronos: 	13.89 fps 	CHFast: 	21.66 fps

Topaz Video AI  v3.3.10
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.941 GB (Adrenalin 23.7.2)
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	16.93 fps 	2X: 	08.03 fps 	4X: 	02.34 fps 	
Proteus		1X: 	14.85 fps 	2X: 	07.30 fps 	4X: 	02.31 fps 	
Gaia		1X: 	09.65 fps 	2X: 	06.25 fps 	4X: 	02.69 fps 	
4X Slowmo		Apollo: 	26.74 fps 	APFast: 	46.93 fps 	Chronos: 	11.57 fps 	CHFast: 	19.14 fps 	

Topaz Video AI Alpha  v3.4.0.1.a
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.941 GB (Adrenalin 23.7.2)
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	22.85 fps 	2X: 	11.09 fps 	4X: 	02.83 fps 	
Proteus		1X: 	20.27 fps 	2X: 	09.30 fps 	4X: 	02.73 fps 	
Gaia		1X: 	10.90 fps 	2X: 	06.95 fps 	4X: 	02.70 fps 	
4X Slowmo		Apollo: 	18.80 fps 	APFast: 	49.20 fps 	Chronos: 	14.52 fps 	CHFast: 	22.63 fps 	

Thank you for posting these details benchmarks. I’m glad the performance is back to original 3.3 levels.

Hi @suraj & @ida.topazlabs - One thing I have noticed on this Alpha (Mac Studio Ultra 64GB macOS 13.5) is that no matter what I throw at it, TVAI produces these weird lines on the output, see attached screenshot. It doesn’t produce these lines using exactly the same video with exactly the same settings via the latest official 3.3.10 release. Doesn’t matter what the input resolution video is nor the output resolutions, I’ve tried all resolutions just to check… same thing. This is via the Iris model. So you know! :grinning:

So, after some more time with the 3.4.0.1.a, all on the Mac Studio M2 Ultra 24c/60c with 64GB RAM running MacOS 13.4:

-the fill up all RAM bug at RAM setting of 100% has been resolved
-speed is considerably up on SD and 2x upscales with Iris on AppleSilicon at least with a high GPU core count.
-Unfortunately, speed has also drastically decreased for 4x upscales
-Previews seem to work fine and also quite immediately for both Preview and actual Encoding tasks
-there still is a mismatch of the preview time point that differs a few frames from the source. See example image:

To be further examined:
-after longer encodes the 3.4.0.0a tended to become unresponsive, taking several seconds after mouse clicks to respond and often even has to be forced shut down - have to see if it is the same with 3.4.0.1.a

1 Like

Can you provide (a clip of) the source video. I never had such here…

Hi @jo.vo

I’ve sent you a private message as per your request for a clip of the source file I’m using.

Have you noticed the following? If I set my Max Memory to 10% all is well. If I set it to anything else, 40%, 100% whatever, then it slows down the processing by exactly half! Maybe it’s just my setup but I thought I’d ask as yours isn’t too far off mine you know?!

1 Like