I am a bit dissappointed with these results. Is there anything in the settings/ files I can do beyond the obvious options that are in the Topaz Video UI (preferences). I saw somewhere that there was something strange going on with Intel and how it interacts with the program?
So I found someone in Reply #165 on this thread that has almost the same specs as you. They also have a i9-13900HX but with only 32GB of RAM. Same GPU though and same OS.
Their results are very similar to yours so I think that may be normal for your setup.
I donât know what is possible, but it would be great to see if Topaz can seek to optimise the program because Iâm running the best specs you can find on a laptop. Imagine the way things go, it will only get worse as they increase the power of these AI models.
Comparing your 1080p numbers to my 5950X/ 3080Ti desktop rig, your 1X numbers look low, 2X numbers look good for a laptop, and your 4X numbers are actually better than my rig.
My understanding is that more upscaling leans on the CPU more, so I think your CPU is doing very well.
The 4090 mobile variant is based on the same die used in the desktop 4080, but has a much lower power limit than the desktop 4080. So, depending on the GPU power-limit / cooling of your particular laptop, you could end up with GPU performance as low as a 4070 / 4060Ti (I think there is a range in which different laptop manufacturers can set different power limits for different laptops)
I ran the benchmark module on 4.0.8 while watching the output of Asitop. I noticed that the ANE (neural engine) was engaged in every benchmark test (to varying degrees) EXCEPT for Iris 2X! Does that explain the notoriously poor performance of Iris 2X for 1080 input file?
Is this a clue that the devs can use to address the issue on modern Mac hardware?
A few days ago, I received a warning email from CableMod about the 90-degree adapter I had purchased for my new RTX 4090 GPU. Theyâre doing a recall.
This is a quote from a recent article on Tomâs Hardware:
âNvidia and its board partners do not cover damage to GPUs that use third-party power adaptors. Before the recall was announced, CableMod offered to repair affected GPUs whose RMA was refused because of their adapters. CableMod is known to send these impacted units to the repair shop NorthridgeFix.â
If youâre using a CableMod angle adapter on your RTX 40x0 you may have a problem like this.
This page CableMod Angle Adapter Recall
Aside from ensuring that the connector is properly seated and appropriate bend radius precautions are taken - you could also consider TDP limiting the card to 450W or so. It really is a disgusting situation and the 12V-2x6 situation is not going to be much better IMO.
Hi I did not build my workstation for gaming or video editing, but got a drone a while back and am now getting into editing video. Recently I tried to convert a time lapse to 60 FPS so that I could slow it down for a smoother look and my processing time to export a short clip was 1 hour and 37 min. I was monitoring CPU and GPU utilization and all 10 cores were idle but GPU was maxed out at 100%
I ran the benchmark test and I donât really know how to read it. Can someone help me understand this?
Benchmarking is really most useful as a speed comparison between different system configurations, or different models on the same system. The fps values it returns usually donât correspond to performance with real exports.
Thank you⊠my CPUID monitor told me that the bottleneck is the GPU. Iâm just not sure what gains I might see if I upgraded⊠and given that I wonât be using Video AI a lot, Iâm really not sure if it would be worth it.
If you get a newer more powerful GPU, that CPU would be able to keep up with it. I agree that it wouldnât really be worth it for you. Unless you really want to use that computer more often, and running TVAI makes it unusable.
I had a GTX 1060. It would usually take 12 hours to do a 45 minute show with the model and settings I was using. My other computer with an RTX 3080 ti would take about an hour on the same video.
From what I see people posting, GPU seems to be the most common limitation for most desktop systems. Iâm running an export right now that has my GPU in the 95-100% range while my CPUs are 0-5%. Already saving my pennies for a new GPU in 2024.
