How much hardware power do you need to run Video AI? I spent unseemly sums of money to do a thing, to build a thing, and this is all the thing can do! Point blank matter of fact..... The simple truth that none of these reviewers ever show or tell, is that nothing currently on the market can run Video AI at a reasonable performance level at a reasonable price point. Yeah I'm a touch irritated about such low end performance from such top end gear, but you can't argue these numbers, these are provable repeatable facts! If any still doubt me, well then build my exact system below and see for yourself. My full PC build below (may be affiliate links): Intel Core i9-13900K -- amzn.to/3IV6bkL EK Nucleus AIO CR360 -- amzn.to/3TE2oNG MSI GeForce RTX 4090 -- amzn.to/43zmqO7 MSI MPG Z790 Carbon Motherboard -- amzn.to/3PBsyQc G.SKILL Ripjaws S5 Series DDR5 RAM 96GB -- amzn.to/3TEoHD6 SAMSUNG 990 PRO (x2) -- amzn.to/3IVRT3n Solidigm P44 Pro -- amzn.to/3xfms1u WD_BLACK SN850X -- amzn.to/4a9JDst MEG Ai1300P Power Supply -- amzn.to/49jz30O Phanteks Enthoo Pro Full Tower -- amzn.to/3PHM3X7 Topaz Video AI -- www.topazlabs.com/topaz-video-ai Give a 👍 and subscribe 🔔 if you'd like to see more videos like this and thanks for watching 😃!
bro thank you for this video, one of the only people coming from this angle compared to everyone else just talking about games that I don’t care about lol. That aside can you do a video on how long it takes to upscaled a 3 to 5 minute video from 1080p to 4K using your 4090 with both Proteus and Artemis?
There is little difference in my opinion between any of the upscaling AI models, aside from the amount of manual control available. The total amount of time depends on the frame rate of the video and your choice of a single or a double enhancement. A typical 24 fps video using a single enhancement for upscaling, I'll normally run between 8 and 11 fps. So worst case scenario for me on a 24p video is 3 seconds of render time for every 1 second of upscaled video. A 30 or 60 fps video will naturally take longer to render. A typical 1.5 hr long movie normally takes 4.5 - 5 hours to render if only using a single enhancement. But if you use a double enhancement as shown in this video here, that movie took 20 hours if memory serves. In hind sight, I should have just rendered the video 2 separate times with a different model each time. It would have only taken half the amount of time in total. It's just a lot of work to upscale video like this, every program capable of doing the job just takes forever to do it. And too, the more you watch the clock the longer it will take! Cheers 🍻
@@Finite-Tuning IMO that’s actually really good, currently on my iMac (i9 9900k 48GB ram, Radeon 575x GPU) I’m only able to do a 2X upscale on 480p content with proteus and doubling frame rate, that takes about 3 and a half hours. Upscaling to higher than 720p takes 8 hours minimum, and 1080p to 4K I’m looking at 24+ hours. I’m currently looking to build a high end pc with a 13900k CPU, 64GB ram, and a 4060ti super with 16gb vram, from the research I’ve done the extra vram doesn’t increase render times but it does give you room to run multiple instances at once (depending on the settings of course).
@@zenvultra: Spend your money on the GPU if this is what you are trying to do with a computer. GPU matters more then CPU from my experience, and I'll spell it out real quick so you know where I'm coming from. I have all Intel based systems, I've run Topaz Video AI on all of them. If using the processor alone to render/upscale video, .8 fps is a good number on an i9 13900K! My laptop 12th gen i9 is far slower, then my desktop 9th gen i7 is even slower still. But when using the GPU's, I have a desktop 2070, a laptop 3080ti, and a desktop 4090. By far the 4090 produces the best render times, followed by the 3080 then the 2070. The GPU is were you should focus the bulk of your money "if" video upscaling/rendering is your primary task. But this is just a really hard job for any hardware, so always buy the absolute most hardware you can afford. If you need to save money for another month to afford it, then do that! Because buying less than the best you can afford only leads to regret. I bought the best of the best at the time on the market, and I'm still disappointed in the performance! But I have no regrets about getting the best thing available at that time to do this job. Cheers and good luck 🍻
Wow, I was expecting a 4090 to render much faster than that. My RTX 2080 (laptop version) took 22 hours for Conan the Barbarian to upscale/ remove noise / convert to 60fps with Topaz. It was well worth the wait, as the final results were amazing. And that video is only 1080p! So I'd imagine that a 4k conversion would take extremely long on my laptop. I'm building a new PC with a 4080, and I'm really hoping that I'd get faster results, combined with a 16 core processor. Hopefully I won't be disappointed when I test it out. Although I'm perfectly okay with 1080p and a faster rendering time that should at least be half the time of rendering with my current situation. Thanks for the informative video!
I know, so did I. Hens my disappointment. but I don't do anything in 1080 anymore, so 4k to the moon..... and back when we all have 8k oh my! It is here and now just most don't use it.
@@Dlyfex.aep7: Try actually watching the video, read the description and the parts list. Then understand how dead wrong you are! Nothing on the consumer market is currently better hardware than what I currently have. There is no "better" hardware! And no, processing time depends MUCH more on GPU then it does on CPU. I have 3 generations of Intel and Nvidia products that easily prove this fact.
𝐈'𝐦 𝐭𝐫𝐲𝐢𝐧𝐠 𝐚𝐧 𝐀𝐈 𝐬𝐮𝐦𝐦𝐞𝐫𝐲 𝐞𝐱𝐩𝐞𝐫𝐢𝐦𝐞𝐧𝐭 𝐡𝐞𝐫𝐞. 𝐋𝐞𝐭 𝐦𝐞 𝐤𝐧𝐨𝐰 𝐢𝐟 𝐲𝐨𝐮 𝐥𝐢𝐤𝐞 𝐭𝐡𝐢𝐬 𝐬𝐨𝐫𝐭 𝐨𝐟 𝐭𝐡𝐢𝐧𝐠 𝐚𝐧𝐝 𝐈'𝐥𝐥 𝐢𝐧𝐜𝐥𝐮𝐝𝐞 𝐢𝐭 𝐦𝐨𝐫𝐞 𝐨𝐟𝐭𝐞𝐧. Video Summary: The speaker demonstrates that even with high-end hardware (RTX 4090 and i9 3900K), Topaz Video AI only achieves 2.2 FPS when upscaling videos, highlighting the immense power requirements for video processing. Highlights 🎮 High-End Hardware: The speaker’s setup includes an RTX 4090 and i9 3900K. ⚡ Performance Issues: Achieving only 2.2 FPS while processing video. 🎥 Video Noise: Upscaling old films to remove noise is a primary challenge. 📊 Proof of Performance: Demonstrates real-time processing to prove claims. 🔥 Heat Generation: The powerful setup generates significant heat. 💻 Editing vs. Gaming: Emphasizes the difference in demands between video editing and gaming. 🛠 Custom Solutions: The speaker uses specific settings to optimize video quality. Key Insights ⚙ Power Limitations: The RTX 4090 and i9 3900K struggle to deliver high frame rates in video processing, showing that even top-tier hardware has limitations in demanding tasks like video upscaling. 🔋 📈 Real-World Performance: Despite advanced technology, the speaker only achieves minimal frame rates, underscoring the inadequacy of current hardware for intensive video editing tasks. 📉 🖼 Importance of Quality: The need to reduce noise in older films highlights the challenges of maintaining visual quality in upscaled content, which many consumers may overlook. 👁 ⏳ Time Consumption: The lengthy rendering times for video processing indicate that video editing is a time-intensive task, often requiring significant patience and resources. ⌛ 🔥 Thermal Management: The setup’s heat generation requires substantial cooling solutions, emphasizing the physical demands of powerful computing systems. ❄ 🎮 Different Workloads: The disparity between gaming performance and video editing frames per second highlights the unique challenges faced by content creators compared to gamers. ⚔ 🏗 Custom Settings Needed: The speaker’s need for tailored settings in Topaz Video AI reveals that achieving optimal results often requires fine-tuning and adjustments beyond standard presets. ⚙ Cheers 🍻
Hey! Have you made a test with a 1080p to 1080p movie? I want to buy a 4090 laptop (ik its power is way more low than a desktop 4090) but I'm curious how many fps would it render if you would try to render a low res input video to 1080p with an upscale model or with a denoise model like artemis or iris. Could you try this for me, please? I mean, upscale a 320p/480p video to 1080p and see how many fps it renders.
Frame rate always depends on exactly what you're doing and how much compute power you have available. If I'm just converting a file format, I can do 1000+ fps in 1080, but if I'm trying to denoise/enhance for example, I can do anywhere from real time rendering all the way to 100's of fps, it all depends on the settings I use. Keep in mind though, with low power draw comes low performance results. All the laptops I've seen with a 4090 cost around $3500. For that same money you can build a desktop with a 4090 and it will have much greater processing power and cooling ability, not to mention upgrade ability. I splurged on my system somewhat wastefully and just barely hit $4,000. Some of my components went down in price now while others have gone up, but I could build it again for about 3500 today just by choosing slightly different parts. Point is, I would never spend that kind of money on a laptop when the same money can get you so much more performance in a desktop. But you can expect around about half of my results with a laptop 4090. Not a good bang for the buck in my opinion, but the more compute power you have the faster you can process data. 1080 is generally pretty easy work for a 4090, at least by comparison to 4k.
Good question, I've wondered the same about many systems. But without the actual hardware to test, a guess is all I can offer. In my honest opinion, I don't think an actual liquid nitrogen cooled super computer could do the same thing showed in this video, at anything better than 20 some frames per second! This type of video work is just the hardest thing to do here in the real world. I don't think there is anything on the market or elsewhere that can do the this job in real time. This job is the hardest thing I ask of a PC, and they all cry under this real world work load!
Impressive difference in the result. Too bad about the processing time but hopefully as tech improves and AI improves, it'll rapidly get faster! Have you considered the high end Apple hardware? At the very least it might be less space, heat, noise, and power requirements?
It takes pure compute power to do this type of work. As far as I'm aware, Apple does not have anything comparable to what I already own, let alone anything more powerful. And I would never pay that Apple premium regardless. But when a single GPU finally posses the power of six of these 4090's, maybe then this task won't be so painful. This video demo was a worst case scenario, usually I run between 8-11fps on a typical upscale and enhancement session. Running a dual enhancement as shown in the video really takes a toll.
Maybe look into NVIDIA VSR or AI filters in MPC-HC or MadVR to denoise/upscale spatially and temporally in real-time? You can capture the video output with a capture device, and would probably get very similar results to what you get with Topaz or other offline AI upscaling software.
My goal is to fix the source material so that it can be replayed on anything without the need of further enhancement. I'm sure AI de-noise and upscaling will become more standard in TV's and the like, my LG already does some, but it still leaves the source lookin like dog shit. I'm really just trying to get the original as good as it can be, so that everything can just play it with no need to further enhance. I do use MadVR to replay some movies, JRiver and Power DVD are my two main players. Cheers 🍻
I showed exactly what I was doing in the video! But to upscale a typical 1.5 hour movie to 4K will take about 5 hrs with my hardware. Of coarse that number will change depending on your settings. If you do a double enhancement like I showed, that movie took almost 20 hrs if memory serves.
@@BlaugranaMumbai: Yes, of coarse it can. But it just might take 2 days, maybe more depending on your settings. Video upscaling with noise removal and enhancement is just really hard work for any hardware.
@@BlaugranaMumbai: Just shy of a day. That's a little better than I expected at 1 minute per hour render times. I can normally hit between 8-11fps with my hardware depending on my setting. The demo I showed in this video was an extreme case. This stuff is just really hard work for any hardware regardless of settings though. So good luck. Cheers 🍻
Without a serious GPU to go along with this program, you can expect .01 to .08 FPS to run the same task. The program gives you the option to choose which GPU that you want to use, if I only use the built in GPU on my i9, It would take 1 year and 3 months to render this movie (basically, sarcastically)! This program requires serious GPU power for this kind of work load, and a 4090 is the best of the best on the market, and you've seen all that it can do. Doing a double enhancement like I did on 4K video is pretty much a worst case scenario though. Other settings and features of the program would be less intensive. Most 4K videos I work on render at 8 FPS. It's a great program, but requires a liquid nitrogen cooled super computer to run a task like this any faster.
I wasn't aware Davinci had any AI abilities, but I know some OFX (AI) plugins exist. I have tried to use Davinci, really really tried, but I just can't wrap my head around the thing. I'm Vegas pro for life, nothing else makes any damn sense to me. I have several AI enhancing tools now, they all do the same basic thing as Topaz Video AI and they all run at snail speed. I have some that can upscale to HDR and that's even worse! I wish I could use Davinci, but that thing just makes no sense to me.
@@Finite-Tuning It uses AI for some things but I do not think the Spatial and Temporal noise reduction is technically AI but it does use the GPU a lot. I saved some older Shoots I shelved for being too noisy. ua-cam.com/video/akPJJ_ggaVc/v-deo.html
@@garypranzo9334 : Vegas pro has noise reduction as well, and then my Boris FX plugins take that to much higher levels. Topaz does something far beyond just removing noise. It can create missing/needed pixels to reconstruct the damaged areas. It's not the only kid in town, but it is one of the very best at doing what it does. I have 4 different AI enhancing programs currently, and none of them use 100% cpu/gpu 100% of the time. They all run slower than the speed of smell and they all do relatively the same job with varying levels of success. Video editing is just hard work. Cheers 🍻
You should have rendered the video in 1080p first, then render it to 4k, tat way it would have done it faster, ai is preety new in the market and our present hardware is not that capable right now
You ain't doing any kind of enhancement in 4K with those numbers! 720 is nothing, I can do over 1000fps with that child's play! 1080 is still nothing, I can do better than real time renders. Do as shown in the video, then watch your system grind to a halt.
How much hardware power do you need to run Video AI?
I spent unseemly sums of money to do a thing, to build a thing, and this is all the thing can do! Point blank matter of fact..... The simple truth that none of these reviewers ever show or tell, is that nothing currently on the market can run Video AI at a reasonable performance level at a reasonable price point.
Yeah I'm a touch irritated about such low end performance from such top end gear, but you can't argue these numbers, these are provable repeatable facts! If any still doubt me, well then build my exact system below and see for yourself.
My full PC build below (may be affiliate links):
Intel Core i9-13900K -- amzn.to/3IV6bkL
EK Nucleus AIO CR360 -- amzn.to/3TE2oNG
MSI GeForce RTX 4090 -- amzn.to/43zmqO7
MSI MPG Z790 Carbon Motherboard -- amzn.to/3PBsyQc
G.SKILL Ripjaws S5 Series DDR5 RAM 96GB -- amzn.to/3TEoHD6
SAMSUNG 990 PRO (x2) -- amzn.to/3IVRT3n
Solidigm P44 Pro -- amzn.to/3xfms1u
WD_BLACK SN850X -- amzn.to/4a9JDst
MEG Ai1300P Power Supply -- amzn.to/49jz30O
Phanteks Enthoo Pro Full Tower -- amzn.to/3PHM3X7
Topaz Video AI -- www.topazlabs.com/topaz-video-ai
Give a 👍 and subscribe 🔔 if you'd like to see more videos like this and thanks for watching 😃!
bro thank you for this video, one of the only people coming from this angle compared to everyone else just talking about games that I don’t care about lol. That aside can you do a video on how long it takes to upscaled a 3 to 5 minute video from 1080p to 4K using your 4090 with both Proteus and Artemis?
There is little difference in my opinion between any of the upscaling AI models, aside from the amount of manual control available. The total amount of time depends on the frame rate of the video and your choice of a single or a double enhancement. A typical 24 fps video using a single enhancement for upscaling, I'll normally run between 8 and 11 fps. So worst case scenario for me on a 24p video is 3 seconds of render time for every 1 second of upscaled video. A 30 or 60 fps video will naturally take longer to render.
A typical 1.5 hr long movie normally takes 4.5 - 5 hours to render if only using a single enhancement. But if you use a double enhancement as shown in this video here, that movie took 20 hours if memory serves. In hind sight, I should have just rendered the video 2 separate times with a different model each time. It would have only taken half the amount of time in total. It's just a lot of work to upscale video like this, every program capable of doing the job just takes forever to do it. And too, the more you watch the clock the longer it will take!
Cheers 🍻
@@Finite-Tuning IMO that’s actually really good, currently on my iMac (i9 9900k 48GB ram, Radeon 575x GPU) I’m only able to do a 2X upscale on 480p content with proteus and doubling frame rate, that takes about 3 and a half hours. Upscaling to higher than 720p takes 8 hours minimum, and 1080p to 4K I’m looking at 24+ hours. I’m currently looking to build a high end pc with a 13900k CPU, 64GB ram, and a 4060ti super with 16gb vram, from the research I’ve done the extra vram doesn’t increase render times but it does give you room to run multiple instances at once (depending on the settings of course).
@@zenvultra:
Spend your money on the GPU if this is what you are trying to do with a computer. GPU matters more then CPU from my experience, and I'll spell it out real quick so you know where I'm coming from.
I have all Intel based systems, I've run Topaz Video AI on all of them. If using the processor alone to render/upscale video, .8 fps is a good number on an i9 13900K! My laptop 12th gen i9 is far slower, then my desktop 9th gen i7 is even slower still.
But when using the GPU's, I have a desktop 2070, a laptop 3080ti, and a desktop 4090. By far the 4090 produces the best render times, followed by the 3080 then the 2070. The GPU is were you should focus the bulk of your money "if" video upscaling/rendering is your primary task.
But this is just a really hard job for any hardware, so always buy the absolute most hardware you can afford. If you need to save money for another month to afford it, then do that! Because buying less than the best you can afford only leads to regret.
I bought the best of the best at the time on the market, and I'm still disappointed in the performance! But I have no regrets about getting the best thing available at that time to do this job.
Cheers and good luck 🍻
Wow, I was expecting a 4090 to render much faster than that. My RTX 2080 (laptop version) took 22 hours for Conan the Barbarian to upscale/ remove noise / convert to 60fps with Topaz. It was well worth the wait, as the final results were amazing. And that video is only 1080p! So I'd imagine that a 4k conversion would take extremely long on my laptop. I'm building a new PC with a 4080, and I'm really hoping that I'd get faster results, combined with a 16 core processor. Hopefully I won't be disappointed when I test it out. Although I'm perfectly okay with 1080p and a faster rendering time that should at least be half the time of rendering with my current situation. Thanks for the informative video!
I know, so did I. Hens my disappointment. but I don't do anything in 1080 anymore, so 4k to the moon..... and back when we all have 8k oh my! It is here and now just most don't use it.
It depends on cpu get a better cpu and ram
@@Dlyfex.aep7:
Try actually watching the video, read the description and the parts list. Then understand how dead wrong you are!
Nothing on the consumer market is currently better hardware than what I currently have. There is no "better" hardware! And no, processing time depends MUCH more on GPU then it does on CPU. I have 3 generations of Intel and Nvidia products that easily prove this fact.
𝐈'𝐦 𝐭𝐫𝐲𝐢𝐧𝐠 𝐚𝐧 𝐀𝐈 𝐬𝐮𝐦𝐦𝐞𝐫𝐲 𝐞𝐱𝐩𝐞𝐫𝐢𝐦𝐞𝐧𝐭 𝐡𝐞𝐫𝐞. 𝐋𝐞𝐭 𝐦𝐞 𝐤𝐧𝐨𝐰 𝐢𝐟 𝐲𝐨𝐮 𝐥𝐢𝐤𝐞 𝐭𝐡𝐢𝐬 𝐬𝐨𝐫𝐭 𝐨𝐟 𝐭𝐡𝐢𝐧𝐠 𝐚𝐧𝐝 𝐈'𝐥𝐥 𝐢𝐧𝐜𝐥𝐮𝐝𝐞 𝐢𝐭 𝐦𝐨𝐫𝐞 𝐨𝐟𝐭𝐞𝐧.
Video Summary:
The speaker demonstrates that even with high-end hardware (RTX 4090 and i9 3900K), Topaz Video AI only achieves 2.2 FPS when upscaling videos, highlighting the immense power requirements for video processing.
Highlights
🎮 High-End Hardware: The speaker’s setup includes an RTX 4090 and i9 3900K.
⚡ Performance Issues: Achieving only 2.2 FPS while processing video.
🎥 Video Noise: Upscaling old films to remove noise is a primary challenge.
📊 Proof of Performance: Demonstrates real-time processing to prove claims.
🔥 Heat Generation: The powerful setup generates significant heat.
💻 Editing vs. Gaming: Emphasizes the difference in demands between video editing and gaming.
🛠 Custom Solutions: The speaker uses specific settings to optimize video quality.
Key Insights
⚙ Power Limitations: The RTX 4090 and i9 3900K struggle to deliver high frame rates in video processing, showing that even top-tier hardware has limitations in demanding tasks like video upscaling. 🔋
📈 Real-World Performance: Despite advanced technology, the speaker only achieves minimal frame rates, underscoring the inadequacy of current hardware for intensive video editing tasks. 📉
🖼 Importance of Quality: The need to reduce noise in older films highlights the challenges of maintaining visual quality in upscaled content, which many consumers may overlook. 👁
⏳ Time Consumption: The lengthy rendering times for video processing indicate that video editing is a time-intensive task, often requiring significant patience and resources. ⌛
🔥 Thermal Management: The setup’s heat generation requires substantial cooling solutions, emphasizing the physical demands of powerful computing systems. ❄
🎮 Different Workloads: The disparity between gaming performance and video editing frames per second highlights the unique challenges faced by content creators compared to gamers. ⚔
🏗 Custom Settings Needed: The speaker’s need for tailored settings in Topaz Video AI reveals that achieving optimal results often requires fine-tuning and adjustments beyond standard presets. ⚙
Cheers 🍻
Hey! Have you made a test with a 1080p to 1080p movie? I want to buy a 4090 laptop (ik its power is way more low than a desktop 4090) but I'm curious how many fps would it render if you would try to render a low res input video to 1080p with an upscale model or with a denoise model like artemis or iris. Could you try this for me, please? I mean, upscale a 320p/480p video to 1080p and see how many fps it renders.
Frame rate always depends on exactly what you're doing and how much compute power you have available. If I'm just converting a file format, I can do 1000+ fps in 1080, but if I'm trying to denoise/enhance for example, I can do anywhere from real time rendering all the way to 100's of fps, it all depends on the settings I use. Keep in mind though, with low power draw comes low performance results.
All the laptops I've seen with a 4090 cost around $3500. For that same money you can build a desktop with a 4090 and it will have much greater processing power and cooling ability, not to mention upgrade ability. I splurged on my system somewhat wastefully and just barely hit $4,000. Some of my components went down in price now while others have gone up, but I could build it again for about 3500 today just by choosing slightly different parts. Point is, I would never spend that kind of money on a laptop when the same money can get you so much more performance in a desktop.
But you can expect around about half of my results with a laptop 4090. Not a good bang for the buck in my opinion, but the more compute power you have the faster you can process data. 1080 is generally pretty easy work for a 4090, at least by comparison to 4k.
Same. 7800X3D + 4090 this is so stupid. And I'm using Studio Driver
I’ve been looking for some Contant like this for a while, so do you think a Quattro card would do better maybe a thread ripper platform
Good question, I've wondered the same about many systems. But without the actual hardware to test, a guess is all I can offer. In my honest opinion, I don't think an actual liquid nitrogen cooled super computer could do the same thing showed in this video, at anything better than 20 some frames per second! This type of video work is just the hardest thing to do here in the real world. I don't think there is anything on the market or elsewhere that can do the this job in real time.
This job is the hardest thing I ask of a PC, and they all cry under this real world work load!
Impressive difference in the result. Too bad about the processing time but hopefully as tech improves and AI improves, it'll rapidly get faster!
Have you considered the high end Apple hardware? At the very least it might be less space, heat, noise, and power requirements?
It takes pure compute power to do this type of work. As far as I'm aware, Apple does not have anything comparable to what I already own, let alone anything more powerful. And I would never pay that Apple premium regardless. But when a single GPU finally posses the power of six of these 4090's, maybe then this task won't be so painful. This video demo was a worst case scenario, usually I run between 8-11fps on a typical upscale and enhancement session. Running a dual enhancement as shown in the video really takes a toll.
Maybe look into NVIDIA VSR or AI filters in MPC-HC or MadVR to denoise/upscale spatially and temporally in real-time? You can capture the video output with a capture device, and would probably get very similar results to what you get with Topaz or other offline AI upscaling software.
My goal is to fix the source material so that it can be replayed on anything without the need of further enhancement. I'm sure AI de-noise and upscaling will become more standard in TV's and the like, my LG already does some, but it still leaves the source lookin like dog shit. I'm really just trying to get the original as good as it can be, so that everything can just play it with no need to further enhance.
I do use MadVR to replay some movies, JRiver and Power DVD are my two main players.
Cheers 🍻
I don't know what you're doing but I only need it for upscaling my 1080p 30fps videos to 4K, how much time do you think it will take with an RTX 4090?
I showed exactly what I was doing in the video! But to upscale a typical 1.5 hour movie to 4K will take about 5 hrs with my hardware. Of coarse that number will change depending on your settings. If you do a double enhancement like I showed, that movie took almost 20 hrs if memory serves.
@@Finite-Tuningcan i5 13gen , with 2070gtx can do this job? Doesn’t matter if it takes 2 days.
@@BlaugranaMumbai:
Yes, of coarse it can. But it just might take 2 days, maybe more depending on your settings. Video upscaling with noise removal and enhancement is just really hard work for any hardware.
@@Finite-Tuning thanks man. Finished my first 21 min video and took exactly 21 hrs
@@BlaugranaMumbai:
Just shy of a day. That's a little better than I expected at 1 minute per hour render times. I can normally hit between 8-11fps with my hardware depending on my setting. The demo I showed in this video was an extreme case. This stuff is just really hard work for any hardware regardless of settings though. So good luck.
Cheers 🍻
I need this same review with a Mac M2 or M3
Without a serious GPU to go along with this program, you can expect .01 to .08 FPS to run the same task. The program gives you the option to choose which GPU that you want to use, if I only use the built in GPU on my i9, It would take 1 year and 3 months to render this movie (basically, sarcastically)! This program requires serious GPU power for this kind of work load, and a 4090 is the best of the best on the market, and you've seen all that it can do.
Doing a double enhancement like I did on 4K video is pretty much a worst case scenario though. Other settings and features of the program would be less intensive. Most 4K videos I work on render at 8 FPS. It's a great program, but requires a liquid nitrogen cooled super computer to run a task like this any faster.
M1 Mac ultra 2.5-3 fps - M1 Pro 1.2fps - M1 Mac mini .8 fps
Crazy stuff lol
I have same setup and use Topaz and Davinci resolve studio.
Davinci is better for this job and will utilze the 4090 better.
I wasn't aware Davinci had any AI abilities, but I know some OFX (AI) plugins exist. I have tried to use Davinci, really really tried, but I just can't wrap my head around the thing. I'm Vegas pro for life, nothing else makes any damn sense to me. I have several AI enhancing tools now, they all do the same basic thing as Topaz Video AI and they all run at snail speed. I have some that can upscale to HDR and that's even worse! I wish I could use Davinci, but that thing just makes no sense to me.
@@Finite-Tuning It uses AI for some things but I do not think the Spatial and Temporal noise reduction is technically AI but it does use the GPU a lot.
I saved some older Shoots I shelved for being too noisy.
ua-cam.com/video/akPJJ_ggaVc/v-deo.html
@@garypranzo9334 :
Vegas pro has noise reduction as well, and then my Boris FX plugins take that to much higher levels. Topaz does something far beyond just removing noise. It can create missing/needed pixels to reconstruct the damaged areas. It's not the only kid in town, but it is one of the very best at doing what it does. I have 4 different AI enhancing programs currently, and none of them use 100% cpu/gpu 100% of the time. They all run slower than the speed of smell and they all do relatively the same job with varying levels of success. Video editing is just hard work.
Cheers 🍻
You should have rendered the video in 1080p first, then render it to 4k, tat way it would have done it faster, ai is preety new in the market and our present hardware is not that capable right now
Nothing new about what I'm doing here. I have years worth of invoices to show how much I've already paid!
17m video to 3 minutes not bad
You ain't doing any kind of enhancement in 4K with those numbers!
720 is nothing, I can do over 1000fps with that child's play! 1080 is still nothing, I can do better than real time renders. Do as shown in the video, then watch your system grind to a halt.
that's too long!