Topaz Video AI Beta 6.2.0.0.b

After installing this BETA with Resolve plugin, my Davinci Resolve 19 will no longer open.
I looked at the Davinci logs and this is what it shows.
0x0000cc44] | BtCommon | INFO | 2025-03-30 13:11:46,209 | Memory config: reserved=18401M pinned=8000M log=0
[0x0000cc44] | BtCommon | INFO | 2025-03-30 13:11:46,209 | Using default pooled memory manager
[0x0000845c] | LeManager | INFO | 2025-03-30 13:11:46,211 | 521, 29
[0x0000171c] | BtCommon | INFO | 2025-03-30 13:11:46,211 | BtResourceManager Process Thread Started
[0x00003534] | OpenFX | INFO | 2025-03-30 13:11:46,274 | OFX: loading com.TopazLabs.FrameInterpolation.BETA

==========[CRASH DUMP]==========
#TIME Sun Mar 30 13:11:46 2025 - Uptime 00:00:00 (hh:mm:ss)
#PROGRAM_NAME DaVinci Resolve Studio v19.1.4.0011 (Windows/MSVC x86_64)
#BMD_ARCHITECTURE x86_64
#BMD_BUILD_UUID d512ce47-d15b-4a20-8aa4-3be7e78493e5
#BMD_GIT_COMMIT 755196be3de2c6b815963568c83e02811d250cbb
#BMD_UTIL_VERSION 19.1.4.0011
#OS Windows
After uninstalling the beta Resolve works again.

I had the same problem and just removed the beta plug-ins from the common/ofx folder.

1 Like

I am confused, is this local Starlight experiment already included in this beta or not (and if not, why mention it?)

And does this new beta include the tensorrt optimizations from 6.1.1.1.a.trt or do we need to remain on that beta for 50 series GPUs?

According to the release note it’s back end prep, so most likely they’re looking for new bugs that might be traceable to that before they try hooking a local Starlight model onto it.

1 Like

Not yet, sometime this Spring.

Exactly what @topaz257 says.

Can look at my post just above yours where I’ve also linked my memory tuning. Now compare my x4 benchmarks with rtx 5080 vs any rtx 5090. My 4x speed with rtx 5080 beats any 5090 unless they’ve also tuned their memory to run a lot faster.

My memory tuned with those timings and speed require a dedicated fan to stay below the temperature where they start to throw errors (measured SPB Hub Temperature of >50ºC means that the memory modules themselves are probably around 57-58ºC, as the SPB Hub sensor is located on the motherboard). With a fan pointing at my memory, they reach their max SPB Hub Temperature of about 40C after 10-15min of memory stress testing, which is some 10C below the threshold of when they start to run the risk of throwing errors.

When my memory wasn’t as tightly tuned, but still ran with the same voltage, tREFI and tRFC (the most heat generating timings) and without a fan, they ran at 65-70C and were fine, as with looser timings they can handle higher temperatures.

Just noticed your mITX config: extra impressive. Which case do you have? Considering a SFF build (waiting on local Starlight requirements)…

I had the same problem, so I removed the beta plugins from the common/ofx folder.
From my experience, I think the official version will have the same problem.
I can’t install the official version of TVAI 6.2.0.

Ncase M2 Grater

1 Like

Some good news here, 6.2.0.1.b has reduced the performance hit compared to 5.x from 30% down to about 10%.

Now, how about that last 10%?

1 Like

This is all great news, I am hoping to see some new generative enhancement-models running locally soon.

By the way, wow is the new fast video player supporting HDR coming along for MacOS? I feel like forever ago since you stopped including the “Mac version coming soon” line in the change log. Have you discontinued work on this or is it still being worked on?

This is something I have been looking forward to as it’s just tedious to have to render a preview and open in an external video player to check the results. Especially when you need to do a lot of small fine adjustments to make it look just right.

Ran some Benchmark on 6.2.0.1, comparing M1 Max and M4 Max. Copied below. When a new section for 6.2 is available I will also copy them. I mostly see almost 2x improvement! APFast is a notable exception !

Topaz Video AI Beta v6.2.0.1.b
System Information
OS: Mac v15.05
CPU: Apple M1 Max 32 GB
GPU: Apple M1 Max 21.333 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 07.48 fps 2X: 04.20 fps 4X: 01.48 fps
Iris 1X: 08.11 fps 2X: 02.83 fps 4X: 00.86 fps
Proteus 1X: 06.96 fps 2X: 04.70 fps 4X: 01.37 fps
Gaia 1X: 02.53 fps 2X: 01.77 fps 4X: 01.20 fps
Nyx 1X: 02.22 fps 2X: 02.72 fps
Nyx Fast 1X: 04.82 fps
Rhea 4X: 00.55 fps
RXL 4X: 00.58 fps
Hyperion HDR 1X: 29.09 fps
4X Slowmo Apollo: 08.30 fps APFast: 33.00 fps Chronos: 02.57 fps CHFast: 04.60 fps
16X Slowmo Aion: 08.56 fps

Topaz Video AI Beta v6.2.0.1.b
System Information
OS: Mac v15.04
CPU: Apple M4 Max 36 GB
GPU: Apple M4 Max 27 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 12.38 fps 2X: 07.27 fps 4X: 02.52 fps
Iris 1X: 13.49 fps 2X: 04.78 fps 4X: 01.60 fps
Proteus 1X: 14.14 fps 2X: 07.17 fps 4X: 02.57 fps
Gaia 1X: 03.96 fps 2X: 02.89 fps 4X: 02.09 fps
Nyx 1X: 03.70 fps 2X: 04.30 fps
Nyx Fast 1X: 10.19 fps
Rhea 4X: 00.89 fps
RXL 4X: 00.95 fps
Hyperion HDR 1X: 50.42 fps
4X Slowmo Apollo: 14.69 fps APFast: 33.66 fps Chronos: 04.81 fps CHFast: 08.32 fps
16X Slowmo Aion: 13.76 fps

Comparing against 6.1.0, my 3080 ti is mostly back to the same speeds. There are a few that are still slower according to the benchmark. Nyx is the worst at 6.44 fps compared to 9.74 fps.

Ok, Does anyone have a solution for when MPV starts “acting up” out of nowhere (again)?

This transition to mpv cost me multiple months of updates, a few v6 builds were running, but now my computer just shuts topaz down whenever i try to load a video into to GUI and i am about 100% sure this is an MPV problem because i dowloaded the windows PLEX client (which also uses MPV and also has a ton of people who have a ton of issues with mpv being “very finicky”)

Why can’t we just get an option to DISABLE mpv ? This “feature” is litereally ruining the user experience… I paid for a year license and more than 6 months of it has been downtime due to mpv issues (espseciallly the early ones)

It’s incredibly ridiculous that i have to keep defaulting back to old versions of this program every time mpv acts up

GRANTED – this hasn’t happened in a while, but even once is ridiculosu … having it crop back up randomly is just stupid to have this as part of the program… you guys have to know it is a very finicky thing especially when dealing with a large range of computers.

We don’t NEED mpv … We NEED topaz to work

1 Like

Unfortunately I still can’t convert my ProRes HQ mov videos to h265 mp4. With AMD Ryzen 9 7900 u 7950 CPUs in combination with RTX 4090 or RX 7900GRE u. 7900XT the process stops at less than 10% completeness. With i9 9900 and Rtx 2080 ti it works .

Here is the logfile:
logsForSupport.zip (55.9 MB)