How many are experiencing crashes or frozen computer?
I’ve had so many frustrating crash aborts with 3.x that I sometimes use 2.6.4, which has also had crashes. Nothing I try seems to have any affect on this instability. I even get computer aborts when attempting command line. Lately, crash aborts have been the norm, usually after a long 1 of 2 processes completions. The fact that this has also happened with 2.6.4 means that whatever is going on, it was present in that version as well, so cannot blame the 3.x versions alone. I have submitted a lot of logs to devs, but today I searched and found several threads here about crashes, so clearly I am not alone. I’m on Mac Intel with a powerful 2020 system and graphics card. I see some who have experienced crashes are on PC. No doubt there is enough hardware variation represented where it doesn’t make sense to point the finger at hardware. One would hope that frequent crashing would be at the top of the bug hunt list. Is it?
I may be one of those odd one outs. I have 3 computers all with different hardware and not one is crashing with 3.0.9. I generally run Proteus auto and restrict my upscale to around 2x. I also run 2 to 3 separate instances on each computer at the same time. I only need Proteus for my work cases and Auto in 3.0.9 is better balanced now(for me) then it has been in the past. Most of my upscales are 480P to 1080p or 720p to 1440p with some 1080p to 2160p and 45 minute to 3 hour videos. I preprocess poor quality videos and all videos are post-processed. Output settings are ProRes 422 Lt and copy the audio. By restricting my processing like this I don’t even use the preview. This method always gets me better upscaling than I can get with non-ai methods. That’s all I want from this software.
I don’t do 4x upscale but I will go from time to time from 480 to 2160 but in 2 steps. I find I get better quality results by splitting the upscale. I don’t have a need for 4k videos personally. IMO a waste of drive space for my use. Others can have needs I don’t have.
So for your tracking of problems here is a list of my computers. And just to be clear I never process to or from an external drive. That’s just a disaster waiting to happen IMO. I’ve been running internal SSD’s for the last 8 years. External is for storage or possible playback only.
I don’t have any TV’s. I only have 1080p 27 inch, a 1440p 27, a 4K 28 and a 4k 34 inch monitor that I view videos on. The 1080p is connected to a 4K Fire TV Stick. The 2K is just a spare now. The 28 inch 4k is on a KVM switch between a Ryzen 5700G with 64GB ram, RX 6600, 256GB NVMe and 1 TB SSD with Windows 10. The other computer on the KVM is a Ryzen 5600x with 32GB ram, RTX 3060(12Gb), a 1TB EVO 970+ NVMe and a 1TB Crucial P5+ NVMe running Windows 11. The last is my main computers that runs 3 plex servers 2 of which are in virtual machines(VMware Player). They can all run at the same time but mostly don’t. This computer has a Ryzen 3900x with 64GB ram, RTX 3060(8Gb) Ti, a 2 TB EVO 970+, 4TB SSD, 2-2TB SSD, and 1TB SSD. I have 60+TB of external storage that is on a switched power relay system so the drives are only powered and spinning when I need them. This computer is running Windows 11. All computers are connected over gigabit ethernet, all drives are shared and I can control any one from the other using TightVNC.
One other thing to note I have separate UPS’s on all computers and a separate UPS for the networking. Since none of my machines have spinning hard drives I do power off my computers if not in use. I see no reason not too. The AIO’s and fans are easy replacements if I ever have that need. Also all computers are in well ventilated areas and not sitting on the floor. They also get cleaned every few months.
For me this version seems to work even slower than the prior version.
It looks like V AI isn’t using my GPU at all, mostly using CPU and a tiny fraction of what my card can handle (newest generation Nvidia ; checked the advanced stats in TaskManager) - but crawling around with ~1-1.5FPS while using almost none of my resources.
In a test V AI actually wanted to take around 8 hours to format a 7 minute video. That’s ludicrous in my eyes.
I hope there are plans to increase the performance with new GPU’s considerably, otherwise I might as well get my money back.
Today was the first time I processed more than one file at a time. I had left the max processes setting at the default of 2. I loaded 8 files into V AI 3.0.8 and noticed it was processing 2 files at a time; then I walked away from the computer. When I later returned I saw that it had restarted because of BSOD crash. The crash (as reported by the Who Crashed utility) was VIDEO MEMORY MANAGEMENT INTERNAL. V AI was the only application running. I’m no expert on debugging BSOD dumps but I thought I’d post this just in case anyone else has seen the issue. I’ve processed several files, one at a time, though this version without any issues. My GPU is an Nvidia 1650, 12GB. I’ve never had this issue on my PC (Windows 10-64 64GB).
A 1650 only has 4 GB of on board Ram. Now it might be shared system ram to move the total video ram to 12 GB but that will be very inefficient and with extreme heavy use could cause Windows to crash. Windows is in charge of swapping video ram with system ram. You can see the same thing happen with games that need more memory then is on the GPU. The second screen shot is my RTX 3060 Ti. It has 8GB of on board ram. You can see that in the Dedicated GPU memory. My system has 64GB of system ram. You will notice that task manager is showing 32GB of shared GPU memory. That is the amount of system ram that Windows is willing to use should my DGPU memory is not enough. Then Task Manager lists GPU Memory. That is the total of shared and dedicated memory. You don’t really want to ever being shared memory with a modern Graphics program. It will slow the machine to a crawl and if you max it out it will crash. This shared memory is a hold out from very old hardware when DGPU’s had very little and very slow onboard ram. Some processors(APU) that have GPUs built in could possibly benefit also from shared memory.
Bottom line you 1650 is a limited GPU and don’t try to process to many files at the same time.
i was never able to export auto crop stabilisation footage without crash, could you please make it work in the next update and also provide a detailed tutorial in order to make it work.
Also would be great to know with what config it work .
MikeF: Sorry about that; it is a 4GB GPU; I don’t know where I pulled 12GB from. I will stick to rendering one file at a time as I did with version 2.6.4 which processed multiple files serially. Thanks for getting back to me.
Here’s a little thing that might help some with crashes as soon as you check a preview or export. Make sure if you are copying the audio stream that is is compatible with the video container. I normally don’t use videos that have opus audio streams but this particular video was a YouTube capture. YouTube uses opus as the encoding for videos. Well Opus is not compatible with .mov containers in Topaz. If you are using the copy function and not transcode VAI will crash. I’ve run into similar error messages with other programs stating some audio streams cannot be used in say an MKV or MP4 container. Other programs don’t crash but give you a warning. In fairness to Topaz those programs have well over 10 years in this field.
So if you get a crash as soon as you go to preview or encode verify the container is compatible with your audio or just transcode the audio and replace it with the original afterwards.
UPDATE: So I may have found a work-around to crashing and/or the sometimes hazy color block patterns that seem to result when using Stabilization:
I cut up my 55 minute SD video (old B&W and color mix, choppy) into smaller clips of less than 5 minutes. These I am able to process! However, some of the clip motion confuses Stabilization, so I went further and isolated specific clips. This seems to resolve the computer crashes I was experiencing, which makes me wonder if the overall video length was too much before, perhaps indicating a memory or graphics card overload issue(?). And whenever the Stabilization becomes confused resulting in the output becoming some color moire pattern, I just isolate that in its own clip and process it separately. Later, I reassemble the clips in Final Cut Pro.
Bottom line: for me, my crashes were due to the length of the video, which in turn, may have been causesing a graphics card memory issue. Cutting the video up and subsequently processing smaller clips has resolved the crashing!
Topaz Video AI’s latest update logged me out. I am trying to log back in using browser and now Topaz is telling me that I do not own any product when I purchased Video AI on Nov 22, 2022. Anyone else had this issue with the latest update?
The option to select render order of upscaling, stabilization and interpolation. In some cases, it makes sense to do it in a specific order. At the moment its doesn’t make it clear what order it does it if you select 2 or 3 of these.
I absolutely agree, but until they do fix it, this is the only way I can use the app.
Recently, I did a search for the term “crash” in both the TVEA and TVAI video forums. One user concluded that the Devs had been made aware of the issue but, to-date, have not indicated they are going to address this. So, for me, cut-n-splice R us for the time being.