Just a heads-up. I can’t speak to the performance difference, but Nvidia staff have confirmed that the unusually high PCIe load is a reporting bug on 5000 series gpus. They said the API used has not been updated for this generation of cards.
Can you tell me please what fps are you getting with 4070 Ti on the same clip on 1080P and with 4K output resolution on starlight mini? Thanks in advance.
Man. 4070 Ti and 4070 Ti Super are NOT the same. Ti Super has the same amount of VRAM as the 5070 Ti, and more… Ti Super is pretty much a 4080 Mini. It is very close to the 5070 Ti, and perhaps even faster for Topaz Video at this point… Which is the main point of this thread.
My neighbor has a an ASUS 5070 Ti OC Prime. Maybe I’ll stop nagging him about swapping it for my sweet ASUS 4070 Ti Super TUF OC, if the latter one is actually faster for what I use it for - mainly Starlight Mini. 
at the moment i can only test on RTX 4070 (Ryzen 3900X) and RTX 4070 Ti SUPER / RTX 5070 Ti (both Ryzen 7950X).
On 1080p or 4K?
i think he used my preset, which scales up from 1080p to 4K using NYX and other enhancements.
1 Like
Thank you for replay. That’s good to know. I just cancel my order on RTX4070ti and I order RTX 4080 for 800usd second hand. I think this is better choice for starlight mini as it’s faster and it have 16gb of ram instead 12gb. It’s for sure jump in performance compare to my old 3070 with 8gb ram.
1 Like
I can confirm that is a very good choice indeed - I was thinking about getting the RTX 5070Ti as well but I ended up going for the RTX 4080 Super for about 778 Eur second hand and I am very happy with it, even upscaling 360p or 480p SD videos using Starlight Sharp to QHD gives me 0.4 FPS which is usable.
2 Likes
Good to hear that and thanks for the info. I’m glad you found 4080 super for that price. The market in Sweden are not that great and the prices are high even for second hand market. This was cheapest 4080 card I could find. It’s zotac rtx 4080 trinity oc. We will see the card will perform with my video files in starlight mini. I hope I will see similar jump in performance like you wrote.
Hi,
The i7 isn’t the problem.
The 10th generation is the problem: PCie 3!
(The 11th generation initially has PCie 4)
→ Of course, the motherboard must also support this.
RAM should be super fast.
Kind regards
1 Like
FP4 isn’t used anywhere.
With FP4 the precision would drop significant.
Seems to me there are actual issues with 5070 Ti compared to 4070 Ti Super….
They say they will fix it in a future driver. I say “Sure!”
i hope they will. I am curious how the higher RTX 50 GPUs perform, compared to the RTX 4070 Ti SUPER and if a RTX 5080 or 5090 is able to match or outperform the 40s GPU.
What PCIe version do you run again?
i have tested both GPUs on PCIe 4.0 (B650 Chipset) and PCIe 5.0 (X670E, X870E). I am also testing the standard RTX 4070 on a PCIe 3.0 Mainboard, which can get about to the same speed as the RTX 5070 Ti using my specific Topaz preset.
1 Like
I can run your test on my 4070 Ti Super on my PCIe 3 system a bit later….
And I think that’s at least part of the problem.
In general the 5070 Ti does perform quite well in TV(AI). But with Topaz software there are quite often inconsistencies leading to some models/resolution/GPU combinations underperforming (see e.g. Iris on Apple Silicon) or seldomly overperforming (Hyperion on AppleSilicon).
Also, higher resolutions show little dependency of the sheer number crunching power but increasingly to the whole system performance and (V)RAM throughput.
So, upscaling from SD to (F)HD will show big differences between, say, a 4060 and a 4090 but upscaling 4k material only very little.
yes, i do think that especially NYX is heavily responsible for the difference in performance. However the RTX 5070 Ti is about 40% slower than the RTX 4070 Ti SUPER on exactly the same system with this preset, which i think is quite a lot.
1 Like
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
