I’ve been reading the forums for weeks including the benchmarks, but I cannot make a decision.
My primary interest is upscaling SD material, DVD or lower.
I couldn’t find benchmarks with 480x360 or 640x480 for both cards and 3.5.x
What I’ve read so far is that the cards are quite close in TVAI performance.
I also read about RDNA3 and Tensor cores and that there is room for improvement- which may come with 4.x
I used to play games occasionally, but I haven’t had time for it this year.
Regarding games the 4080 could be preferred over the 7900, but the 7900 has more memory.
The RTX costs 20% more than the RX, which is alright if the performance is 20% better.
I only buy graphics cards once in a few years, so I may sound a bit confusing.
But I want to avoid that I run into the thought “Damn, I should have bought the other card”.
He already has the second card.
And yes, the internet is full with reports of failing RTX especially 4090 cards - which seems very unfortunate considering the cost of those but that’s the disadvantage if you choose something at the highest end, reaching the limits of what’s actually doable, I guess.
A pimped to the max race-car V12 motor won’t run as reliable as your typical 200hp Diesel, either…
I used to have those black screens watching youtube, etc. at least once a week with my GTX 1080 Ti on Windows 10. usually when watching videos or streams or using GigaPixel AI on a large number of photos (bulk process).
The only way out when those black screens occurred, was to hard reset my machine (push the “reset” button - equivalent to power off/on) . I even reinstalled (clean install - Format and install) windows 10 about a year ago but still those continued to happen. I suffered with this thing for almost two years.
I installed Win 11 (clean install) about 2 months ago and since then it stopped.
I was surprised to see that the 3090s and 80s were mentioned as having problems.
My 3080ti has had no issues with anything since I got it on its launch day.
In fact, I bought an RX 6900 XT just for running TVAI on it. In the month or so that I had it, it only successfully processed a tenth of all the videos I tried on it. I eventually got tired of it and bought a 4070 ti to replace it. I have processed a few hundred videos on it without any issues.
I use it for the same thing the opening poster said they want to use TVAI for.
My suggestion is to stay away from the 7900 and get the 4080.
Main difference between these two manufacturers is - NVIDIA produces AI-oriented processors, AMD - “advanced GPU”. It means, if you workflow is not based on constatnt AI-oriented processing - you cna choose AMD as cheaper and slower alternative. But if your workflow based on AI-processing - there is no good alternative in consumer segment to NVIDIA’s AI-processors, despite bugs in drivers.
7900 XTX has no normal tensor processor at all. It very suitable for gaming and episodic AI-related work. But it is still unacceptable for serious usage in AI-flow.
So each one can choice according his needs.
Actually, I see no NVIDIA issues on multiple machines I have. But I’m agree - more complains should be about cards that most selling (just a quantity matter).
Yet another time my opinion is - card should be choose according workflow planned to execute. For single task it is possible use even modern CPU only. But for hard works AI-processors should be choosed - not HW-accelerators for AI tasks…
They have said that they are AI-oriented, so it is a focus of theirs. But, it is important to know that this only makes a difference if the AI processing you intend to use is made on their AI frameworks.
TVAI does have Nvidia specific implementations of most of their models. The non-Nvidia specific implementations run slower for sure, but it’s not impossible for Topaz to make AMD specific implementations that will yield speed improvements. Nvidia forced AMD to copy them, so they are behind at the moment. If you look back at VEAI 2, AMD cards were faster than the Nvidia cards.
I feel like it’s a little misleading to take those marking terms to make it sound like their products do different things.
If I remember correctly, the Gaia model does not have an Nvidia specific implementation.
Actually - not. If you take a look of AMD’s specification for XTX you will find only HW-accelerators. And more - it is weaker than old NVIDIA’s Turing cores.
Any way RX - is a cheaper solution for gamers and for whom occasionly make processing without time-critical requirements.
For more intensive workflows you need to use any AI-processor.
So choice should be according real requirements.
About implementation - it is not a matter of card architecture. For TopazLabs the same architecture can be used successfully (for TVAI), but have multiple issues in “brother” product (for TPAI).
One of the big selling points of the Apple silicon M1 and M2 is that they have Neural Cores for AI tasks. It’s the same story. They only make a difference if you use the Apple AI frameworks.
AMD might be farther behind then I thought. I can also see AMD not wanting to reinvent the AI framework.
So, finally, we’re staying at the same point - choice of AI rpocessing unit depend on requirements and workflow.
If you are gamer and make episodical AI-processing - HW-accelerators are your choice.
If AI-p;rocessing is your business - you need AI-processors.
It sounds to me like you are saying that because Nvidia was the first to be implemented in TVAI, it will be the only one. Because their cards are currently faster, they are somehow AI cards, and the others not. That is false.
All the GPUs from all the companies (Once the drivers get worked out) can run TVAI. There are very few people that report buying and using industrial GPUs on these forums. The ones that I can remember, expressed disappointment that they didn’t work or that they were slower than the ‘Gamer’ GPUs.
I’m just trying to make it clear that Nvidia is on top right now, but that that may not always be the case. Don’t go buying some GPU based on dubious marketing terms that make it sound like the competition doesn’t have—or even have plans for—similar offerings.
Completely wrong. At the moment consumer market divided by NVIDIA’s AI-processors and HW-accelerators by other companies. But may be two year later it will be other state.
So selection for right now is obvious and depend on you requirements.
And it is not related to T%L products. Many companies now got the same race. So company which can use most powerfull HW available will be winner for nearest time (until HW market change)
No, you’re wrong. Just because card of manufacturer A is somewhat slower than the top model of B this doesn’t mean that it can’t be a valid choice at all.
performance/cost ratio is also important, for some performance/watt as well.
Plus, and this is IMO a really important point: If manufacturer B has the fastest card at the moment but has a history of driver issues leading to crashes and hardware defects (RTX 4090 series, anyone?) and the GPU of manufacturer A does the same just a bit slower but reliable it might even be the preferable choice.
I really like the vivid discussion here. This is evidence that my decision is not easy if you have an eye on money.
Maybe owners of RTX 4080 (4090) and RX 7900 XTX could post built-in benchmark results for resolutions below HD, by chance?
But it is exactly what I said - you should select HW according your requirements and workflow. If you are gamer and only sporadically make AI processing - cheaper but weaker cards are your choice. If you are envolved in massive AI-processing - you need normal AI-processor even from consumer segment
