I’m considering updating my Topaz Video AI from my current 2.6.4 to the latest 3.3.5 for $119.
But, I’m not sure if my rig will support it.
These are my current PC specs…will they support version 3.3.5, or do I need to update my CPU and/or GPU card?
Processor Intel(R) Core™ i7-3770 CPU @ 3.40GHz 3.40 GHz
Installed RAM 32.0 GB
System type 64-bit operating system, x64-based processor
Edition Windows 10 Home
Version 21H2
OS build 19044.3086
Experience Windows Feature Experience Pack 1000.19041.1000.0
NVIDIA GeForce 9600 GT
Total Available Graphics Memory 4095 MB
Dedicated Video Memory 512 MB
Will updating to 3.3.5 decrease the time if takes to render? Will upgrading my hardware decrease the time?
Currently a 12 second, 9.32 MB DVD(480p) clip took 6 hours to upscale to 1080p
I’m guessing the 9600 GT can’t even run version 2.6.4 (cpu processing would explain the slowness), so I wouldn’t expect any improvement on 3.3.5.
Also I wouldn’t spend much money upgrading your current hardware, unless you can find a very good deal on a second hand pascal gpu, assuming your motherboard and power supply can support it.
I am less concerned about you CPU, but your GPU.
I myself have the same CPU generation as you do (well, maybe with a little bit more salt. I got the Ivy Bridge-E CPU) with NVIDIA 1080 TI and I am running the v3.x, it’s not the fastest processing but it runs OK, no complaining.
for sure it runs faster then v2.x on my hardware.
I would suggest finding yourself a 2nd hand 3080 card. I think you would be happy with the results in general. best compromise price to performance in my opinion. instead of buying a whole new rig.
My guess with the new GPU, you would be getting with that 12 sec 480p video upscaled to 1080p roughly 30 sec - 1 min.
if you want, upload this 12 sec video of yours and I can test run with TVAI v3 and tell you how long it took me to process , then you will have an idea of how much of a speed boost you would be getting if you upgrade your GPU, as we are running on roughly the same CPU, motherboard & memory generation. the only difference really is the GPU.
That way you would be getting an indication on the speed boost if you change the GPU on your current rig.
You mean upload the original DVD clip,not my rendered clip? I found out that my motherboard won’t support any higher CPU(I’m maxed out). I would need to get a new motherboard(basically a whole rig), which I can’t afford now. So, it sounds like upgrading my GPU will decrease render times. Can anyone recommend a ‘upper-medium’ quality/priced GPU that would work in my rig. I’m currently using a 'constant rate factor of 10…should I increase that to ??? to increase speed?
I do render whole movies, that takes time. Got a faster CPU and Topaz 3.x seems to be faster with Proteus.
But now there is Iris and does much better jobs on some files.
36h with Proteus or 5.5 days with Iris…(M2 Mac) for one 90min movie
I chose the better quality. Topaz gets better and faster, but new models with much more benefits cost time.
Choose if you want a speed up with old settings or better quality with new things.
correct, upload the original 480p clip. that way i can run an upscale test and provide you the time it took me to process it.
We are not looking into upgrading your CPU, as this would be diminishing return in case of TVAI.
we are looking into upgrading your graphics card. since i got the same rig as you are apart from the graphics card, once i test your clip and provide you the time it took me to upscale it, you would have an idea of what performance boost you would be expecting by upgrading your graphics card only…
Ah yes, that’s what I thought you meant, just wanted to make sure. Ok, attached.
I’ve been running some experiments with various settings to compare render time and output quality.
Here’s the settings for the first two attempts:
#1 Artemis High Quality, crop to fill frame off, grain off, CRF 15, 7.11 sec/frame
started 8:40AM 7-23-23 ended 9:15AM 40MB 35 mins
#2 Artemis High Quality, crop to fill frame off, grain off, CRF 10, 8.20 sec/frame
started 9:24AM 7-23-23 ended 10:03AM 78.5MB 39 mins
My first attempt about 4 days ago with grain took 6 hours, making a 116MB file. I don’t see any difference between the rendered files with grain, or no grain.
I tried to use the ‘interlaced’ models, but the ‘keep audio’ was greyed out.
When you run your test, can you list the various settings/options that you used?
Hmmm…I don’t see anyway to attach a file…do I need to post it to Google Drive, then post the link here? OK, found the ‘upload’ button, but it says ‘not allowed’
In the meantime, I’ve experimented with various combinations of settings…Artemis Low, Med, High Quality, grain on, off, CRF from 1-15, noise, high compression…for my 12 sec 9.32MB DVD clip, they all rendered at about 7.14 sec/frame, taking about 35 mins, ranging from 40MB to 254MB…and THEY ALL LOOK THE SAME. The 40MB file looks as good as the 254MB one.
I’m currently running a Gaia High Quality CRF 10, no grain…but it’s running at 51.01 sec/frame, with over 3 hours to go. The split screen progress doesn’t look very promising.
I ran the tests with all the Models and it upscaled it very fast for me.
With VEAI 2.6.4 i took me 1m 6s to upscale it to 1080p using Artemis HQ - Auto (I did not test the others with VEAI - the others i tested with TVAI v3.3.5 - see below), twice the time it took me with TVAI v3.3.5. I encoded it as h264 CRF21 in VEAI. in TVAI it was encoded using h265 which supposedly slower then h264. Even then, TVAI v3.3.5 beat the VEAI v2.6.4 more than 100% speed.
Here are the time it took for each one of the AIs in TVAI v3.3.5
My Specs: Processor: Intel(R) Core™ i7-4930k CPU @ 3.40GHz (Ivy Bridge-E) RAM: 32.0 GB DDR3 SDRAM Quad Channel 1600MT/s 9-9-9 (CL9 / CAS# 9.0) OS: Windows 10 Enterprise Edition (64-Bit) 22H2 GPU: NVIDIA GTX 1080 Ti (11Gb VRAM)
As for the GPU: in Gaming the performance of the GTX1080 Ti is in par with the RTX 3060.
In TVAI (v3.x) the 3060 I think would perform better as the 3060 has Tensor Cores & RT Cores and mines doesn’t (1080 Ti, only has CUDA cores) and TVAI utilized those RT/Tensor Cores to its benefit. CUDA cores count are the same for both 1080 Ti & RTX3060 (3584 CUDA Cores). 3060 also comes with GDDR6 while 1080 Ti is GDDR5x. so my bet that you would get better results with a RTX3060 then i am getting with my GTX 1080 Ti
As you can see the only major difference between your computer and mines are: the GPU and the Topaz version (v3.3.5)
If you want to check the difference between the outputs, I suggest using the Video-Compare tool.
Wow, using VEAI 2.6.4 sure was fast and a small file. I’m going to run a test on my system using CRF 21 and see how long it takes. When you hover your mouse over the ‘Constant Rate Factor’ question mark, it says ‘lower than 17 produces better quality’. This is why I was using CRF of 15, or 10. But, it looks like using 21 is just as good?
Yes, our systems are almost identical, except for GPU’s, but I wonder if my Edition Windows 10 Home Version 21H2 makes a difference, compared to your enterprise 22H2? I’m not sure what kind of RAM I have, but yours is probably better?
So, you would recommend me upgrading to a RTX 3060? Will my system support it? What’s the difference between RTX 3060 and RTX 3060 TI? Which one should I consider?
I’ve tried using the ‘video compare tool’…is this the one that splits the screen into 4 screens with 4 different ‘models’, or the split screen slider that lets you see the original video, and the upscaled video?
In H.264 and H.265, CRF ranges from 0 to 51 (like the QP). 23 is a good default for x264, and 28 is the default for x265. 18 (or 24 for x265) should be visually transparent; anything lower will probably just waste file size. Values of ±6 will result in about half or twice the file size.
I never use sub 20 crf. usually i am between 23-26.
You can read more about it in this Post.
I can assure you this would have zero impact the OS version. Enterprise just means it could be managed remotely by your IT staff and some other features mainly for organizations.
RAM is also insignificant at those levels as we are at the same generation. and anyway mostly is processed through the GPU VRAM, assuming you would have a newer generation of GPU (1080 or above).
I would try and aim for 3080 if possible. if not I assume the TI might be better, but that is beyond my knowledge as i own none of them. I believe the 30 series would fit your system. better double check though.
Thanks for the fast reply, and that CRF link…loads of info. I’m going to research those RTX 3060 and RTX 3060 TI GPUs, prices, and if they will work in my system. My baseboard is MSI, B75A-G43 (MS-7758) 2.0…I think this is from when I built my PC in 2013…time to upgrade!
IMO it’s time for a new computer.
I have an RTX 3050 and I’m happy. As for me, it’s fast at a low price. Today I would buy an RTX 4060 (Ti?) and that is what I would recommend to you if you are considering buying a 3060 Ti.
@Kadet@tkmops 4060 performs in about the same as the 3060, because of the cut down memory bus, lower GPU Cores and PCIE-Link x8 (PCIE Lanes) unlike the 3060 which is x16. this is especially significant to older computers with PCIE Gen.3. our mate has a PCIE Gen.3…
RTX 3060 Ti outperformed the 4060 Ti in many titles and on others it was in Par. in almost all cases 3060 Ti outperformed the 4060. the 3060 out performed the 4060 on some titles as well.
all the critics on Youtube like Lines Tech Tips, Gamers Nexus, Hardware Unboxed, JayZTwoCents, etc. are all adamant “stay away from this card”.
the problem is that he is on a tight budget. so a GPU replacement would give him the best boost money can buy with his limited budget. that is why we were only talking about GPUs.
Thanks for all the responses, everyone. Still trying to find out which cards my system will support. Since my motherboard is so old(from 2013), I’m probably limited to which cards will work. I’ll post back when I get more info.
