I also had a major performance hit in this release (Mac). Downgraded for now. Sounds like it’s being worked on so hopefully the fix is coming soon.
5.1.1 was just released and should fix the performance issue.
Did AMD fix RDNA3??
So … if I well understand, the only usefull feature will be “reserved” to ? … We even don’t know
I am “a professional”, at least as much as Disney or Warner seeing their release … That thing said, I can therefore benefit of it ?
Why adverting it in 5.1 if the purpose is not turning 99 % of paid users furious ?
There was actually “somehow” multiple GPU efficiency with 5.0. Have I to understand that I bought 4x 4090 for being kicked off the right to use this functionnality ?!?
… Or in other words, are we, all paying ingeneers and devs (imo a relative huge amount) for giving majors functionnalities that even don’t care at any moment untill now from at least 10 years ?
Any 4K release of legendary movies ? No … we count them on one hand fingers, and I even don’t speak with remastering, that are not.
Any effort for localisation audio ? No … We still are with DD 5.1 in 99% of stuff, where as 11.4 are sold by tons to what I call “Netflix cornflakes eaters”
Is there any official position with it ? In order to be clear with a 300$ refund or not, or if I am a bit stupid and miss something important.
Regards.
1 Like
Hello everyone,
To clarify some of the discussion around the multi-GPU mode selection added in version 5.1:
-
Video AI Pro is a new license type that will be fully launched in June. We’ll have more information to share on pricing and availability of the Pro license soon.
-
Existing functionality in Video AI will remain available to users working with Individual licenses.
-
The “Single video” mode that is available to all users is the same “All GPUs” mode from previous versions of Video AI. It has now been updated to work with TensorRT models, and this is also available to all users.
-
The “Multi video” mode is designed to optimize utilization of many GPUs during large batches of renders. This mode is much more likely to be used by professional organizations and large teams of editors in a server environment because of the overall decline in popularity and software support for multi-GPU desktop systems.
3 Likes
Why does it stop when I do multiple preview landers at the same time? I don’t understand
1 Like
Multi video is not going to be of much use without better control of GPU selection.
Single video is already not usable effectively with more than two GPUs on a system because the only options are to pick one GPU or all of them, and the problem will be even worse in multi video if it isn’t possible to choose which GPU to use with each video.
BTW, server-based video production teams don’t usually use multi-GPU systems. They use distributed processing render farms made up of multiple systems networked together. This is something that would actually be worth a “Pro” product: a server system to control and preview enhancement and multiple clients installed on networked systems that the server would feed frames to for processing and then pull back and reassemble into a finished export.
1 Like
Wanted to confirm that the previews and preview bar missing in 5.1 is fixed in 5.1.1… but that said 5.1.1 is SO SLOW that it is now effectively unusable. As a test I grabbed 4k copy of Starship Troopers because it has crazy grain. No resize, aion v1 to 50fps, and nyx v3 relative to auto. This normally runs at about .7-2.3fps. It is now .2fps. A simple 2 second preview is taking 15 minutes.
Guys, I cannot say this enough, stop changing your gui every 10 minutes and fix these jokes of speed!
5 Likes
Existing functionality in Video AI will remain available to users working with Individual licenses.
Does that mean the regular (non-pro) version will not be developed in the future? Or, that both pro and regular versions will have parallel development with different features, bug fixes and better, improved models for both versions?
Continued model improvement is what I need for my work. Currently the existing ones are barely adequate + rendering speed needs to be fixed so it works like before. I can’t constantly update hardware with every TPAI version change.
Running a test render now. The frame rate is back to where it was before. Thanks for the fix.
the so called multi gpu feature is also a joke. Where is the sense than not even a single pair of cpu and gpu can be maxed out ? I dont see it.
1 Like
I feel less alone, was on the way to patch my 4x 50% workload on RTX 3070 with 5.0 feeling I was mainly not competent.
I am now using 5.1.1 giving it 3 models to eat, 1 GPU is 100% or at most 95%+5% all others are FULLY sleeping. So for me this patch is “majors” will get true multi GPU, you paid users will get “cheated multi”
1 Like
I can confirm the performance problem for my MacBook Pro M2 Max has been resolved in 5.1.1. I’ve posted updated benchmark results and things are back to normal.
Something has been disturbing since version 5.1.0. Here’s the problem. The interface shifts a little to the right each time the program is opened. And there’s nothing you can do to correct the problem the next time you open it.
What’s more, if I zoom in and out in the taskbar and reopen the window, it redraws to the right.
4 Likes
Let me disagree with this:
- Video AI Pro is a new license type that will be fully launched in June. We will have more information to share about pricing and availability of the Pro license soon.
I hope we get smart information so that I can ask for a full refund within 30 days. I wait for the condition like a kitten.
- Existing Video AI features will remain available to users working with individual licenses.
Wrong: things have turned out in a strange way with this new version. The behavior is no longer (at all) the same. users can’t post tons of essays, because the encoding is SO long and therefore not multiple GPUs. We need to explore new ways of “Why this version with these models has this behavior”
- “Multi Video” mode is designed to optimize the use of many GPUs during large batches of rendering. This mode is much more likely to be used by professional organizations and large editor teams in a server environment due to the general decline in popularity and software support for multi-GPU desktop systems.
Fake. There have never been so many GPUs on the market since the disappearance of ETH miners. All these people are aiming for a new use
- AI? Yes: Renting GPU power, profitable? Absolutely not, the market is too young and AI is “evil” mainly because of the press.
- Mining crappy crypto coins? Yes: Due to the need to be profitable with millions of GPUs mainly Nvidia, Too many projects, too many scams, too many “memes”, unpredictable
Applications exist, but mainly on Linux
Origin of your “many” numbers ? Mine : Jensen Huang CEO, that said: “50% of GPUs between 2018 and 2022 were sold to miners, which is why we introduced LHR.” So… What are these millions of Nvdiia GPUs doing now ?
So … What to do? Convert our crappy BluRays paid at full price to Mickey because this industry is now dangerously concentrated?
You want a denoised “Straship Troopers”? → Dream, it will never come out.
More generally : You want a good film at the cutting edge of technology, those from the 90’s to 10’: → Dream, it’s not profitable for the studios.
What is now profitable is to sell series at very high prices and announce, to the great dismay of payers, that it is being discontinued for reasons of profitability (example: Constallation). Plus, all these streamers are putting the “HD”, “Atmos” label on really mediocre stuff. Apple even invented Dolby Atmos on 2 channels…
We are tired of this way of looking like idiots. Many of us are waiting for the cutting-edge Video AI enhancer: you, the opportunity to break these insane barriers.
The profitability for the professional that you seem to be targeting is not the same as ours, it starts at 500%. We would love to get such a feature for the $300 per year asking. We don’t need the majors to become even more majors on our backs (in case they don’t give you a single fund to develop, otherwise I would change my mind)
TensorRT models
Very good news! But actually…eh how are they used? Many of us are wondering about multi dveenu of mono, so we would be very happy to have a tutorial or a development expert explain things. I’m really ready to learn.
What we can also see is that even with great efforts with AMD RDNA3 and the new ones to come, they are only fit to be destroyed 
It’s a shame their market shares continue to increase
Very sincere respectful greetings
3 Likes
Sorry what are you saying ?
Key to prove what ?
I know what h265 is and is not lossless
1 Like
We’ve removed a section of replies that had shifted away from release discussion, and we’ll be focused on topics related to Video AI 5.1 for the rest of the thread.
7 Likes
This and Photo AI both have major issues with the window sizing after restoring from minimized.
All I want is software that can remember it’s last window state. If it was maximized before minimizing or closing, it needs to return to that state!
2 Likes
Can you post this in the bugs section too please? Otherwise it may never get fixed.
1 Like
It would be great that when I reduce the window size via the mouse it would stop getting smaller at the smallest size where all controls still fit into the windows as it does in Gigapixel and Photo AI.
