You need direct x support otherwise the software will not start to preview or export.
When i set my quadros to TCC (Tesla Compute Cluster Mode) TL software does hang.
eGPU works great with VEAI, with no performance penalty, as far as I have experienced.
Hi, Iāve just upgrade my two computers to use RTX cards after agonisingly long waits on my NVIDIA P2000 when restoring video.
One is an AMD 6 core with an RTX 3080 Ti card. The other is an i7 6 core with an RTX 3060Ti card. The improvement is astonishing. I also fitted 2 x 1TB NVME cards - one to each computer. This helps the transfer to disk time enormously. So I render out to SSD using deinterlace and upscale from 720x576 to 3072x2304 at 50 FPS and am getting 0.33 sec per frame render time on the Intel computer with the 3060 Ti card. I have every respect for the work that you are doing on the software, and will try the idea of two versions running simultaneously on my RTX 3080 Ti computer, only as the RTX 3060 Ti seems to be getting better performance. I realise the note you posted was in January but notice only sporadic 100% use of the GPU core capability on the RTX 3080 Ti compared to the 100 % near continuous use on the RTX 3060 Ti.
1 Like
The transfer to a ready made file is time consuming on the CPU as well as the GPU I think. You must have a fast disk to save to. I render out to 8 bit tif sequences then it is the fastest. I am getting 0.33 seconds per frame (3 frames per second) with my RTX 3060 Ti (uncooled) on my i7 6 core - and thatās de-interlacing and 400% upscaling. Looks marvellous and goes through nicely. Iām rendering out to a 1TB NVME which is the fastest to use for storing. Envious of the speeds that youāre getting on your RTX 3090 and i9!!!.
1 Like
Not sure if Iām doing something wrong, but Youtube doesnāt appear to support a 704p resolution so the highest option available is 480pā¦ so YT is downscaling your upscale to basically the same size as your sourceā¦
But your piece does raise an interesting suggestion - a model that could take the āold video styleā and clean/modernise, and potentially vice-versa, would be a very interesting item.
Iām using Artemis High Denoise on a 1920x1080 @ 29.97 fps with 18,291 frames using a GTX 960 GPU and ETA is 4 hrs: 14 mins, 1.75 sec/frame which is 52.56x over source frame rate This speed is way too slow for a 10 min, 10 sec video. A 90 min video would take forever. Will a GTX 3060 reduce the processing time by 3x to get at least 0.583 sec/frame?
Thanks a lot for those details! I am currently using an RTX Quadro A4000 (Ampere card, think a 3070 but with 16GB of RAM, triple the amount of tensor cores of 3070 but 1/3rd of the amount of ray tracing cores), currently running VEAI uses up between 4-8GB of VRAM depending on the model and the scaling size, it never went above 50% (I havenāt tried upscaling to 8K though), and Graphics_1 utilization is always somewhere between 30%-45%, also depending on the model and the upscaling settings.
I have a few questions:
- Do you still see support for cards with more than 12GB on the horizon?
- Are the extra tensor cores expected to add anything to performance vs. a regular 3070 for example?
- Since you were supporting CUDA before, canāt you add an options in the settings to switch between CUDA and DirectML? Or does DirectML have the same performance anyway?
- How is your progress with RTX 3000 series optimizations in general, when you say āsuperior performanceā, how much of an increase are you expecting in general?
- Are the optimizations usually part of the model itself or the inferencing software? Do you think one day we could be training our own models and then using them within VEAI?
Thanks a lot for the great work!
### My benchmark Result
Here is my practical benchmark of conversion time.
PC: Lenovo ThinkCentre M75q-1
CPU: AMD Ryzen 5 PRO 3400GE @
Memory: 16GB
Graphics: On APU, AMD Radeon Vega 11 Graphics (video mem 2GB)
360p ā 720p approx 0.30 sec/frame
480p ā 960p approx 0.35 sec/frame
720p ā 1440p approx 2.00 sec/frame
1080p ā 2160p approx 2.50 sec/frame
Actually the above results are almost same on the follwoing configuration, too:
PC: Lenovo ThinkCentre M75 Gen2
CPU: AMD Ryzen 7 PRO 4750GE @3.10GHz 8cores 16threads
Memory: 16GB
Graphics: On APU, AMD Radeon Vega Graphics (video mem 2GB)
### ROI for separate video card ??
I want to know if these conversion time can be shorten to 1/2 or 1/4 if I buy the high performance video card like RTX3080, RTX3060 (with another video card extensable PC) investing about $1,000.
If not, I donāt want to spend money much for slight covnersion speed-up. Anyone had the answer for this ?
Thanks,
1 Like
@footofwrath Make sure your video gets high bitrates. If lower than youtube expects, youtube downsamples your video to save their storage.
Iām not sure if my software is bugged, but I have a Ryzen 5950X, RTX 3080, and 32GB RAM @ 3600mhz. Iām also using studio drivers.
But I feel that with all this power, the upscaling would be faster?
Iām using the Artemius Medium quality and upscaling 720 Hour-long videos to 1080p and itās doing so at 0.15 sec a frame. Is this good?
When I check my resource manager CPU use is only 25-30%, and the GPU 3-5%, am I missing something?
Iām using the āAll GPUSā option, but it doesnāt seem to be using my full resources.
1 Like
Thanks for linking to that thread, my questions were actually meant as a follow-up on these exactly. I just replied to another reply by mistake.
Here are some speed comparisons between my desktop pc with an Intel Core i7 8700k 6 core + RTX 2070 Super and my laptop with an AMD Ryzen 7 5800U 8 core + AMD Radeon Graphics RX Vega 8 iGPU using Artemies low quality V.13 on a 640 * 480 video upscale by 200 %:
RTX 2070 Super: 0.1 sec/frame
RX Vega 8: 0.3 sec/frame
Intel Core i7 8700K: 1.1 sec/frame
AMD Ryzen 7 5800U: 1.5 sec/frame
The RTX is three times that fast than the RX Vega 8, but the RTX alone weighs 1.4 kg and consumes 215 W of power. The whole laptop with the RX Vega 8 weighs slightly under 1.1 kg with a TDP of 15 W (may consume some more watts on full power).
1 Like
Just a quick one for those on a tight budget, wanting to have a reasonable fast card for smaller projects:
The GCN 3.0 FURY (X, Nano, etcā¦) have 4096 Cores, pack around 8 TFLOPS of FP32 and are equiped with HBM Memory, offering a very high memory bandwith. 4GB of VRAM doesnĀ“t seem to much, but the card runs on FP16 just the same in VEAI, 1080p input totaly is workingā¦
These cards (specially the Fury X watercooled) perform VERY good in VEAI for the price at the moment. Around 200 Bucks depending on where you live and buy barely is enough for a bigger GEN9, smaller GEN10 or RXā¦ Card from AMD - all of them being significanntly slower than the FURYā¦
DONĀ“T Compare Gaming benchmarks, Cuda Epxeriences, etcā¦ the only usecase where the difference is comparable, is OPENCLā¦ In Gaming and most other usecases, the FURY are not better than similar priced NVIDIA - but in VEAI (Using DirectML) - they shineā¦
Sorry, no numbers at the moment, postet them on facebook, maybe IĀ“ll find them againā¦
Use free version of Da Vinci resolve to piece everything togetherā¦ lets you maintain the original audio with little other work to be done. And easily builds clips from images to get the upscale going even faster - you can just out to TIF and let the final render in Resolve do the rest. Free version up to 4K render. 
Heh that wasnāt my pointā¦ the guy I quoted, chaossky, showed a video upscaled from like 420p to 704p, but YT doensāt have a 704p option, so when it renders it back he only gets 480p publishedā¦
1 Like
Is there a guide for RTX 3000 series yet? And are we able to make proper use of e.g 3950x and 3090 w/ PCIe 4.0 SSD?
Or still need to run two instances?
As far as i know its two instances.
