TMPGEnc Software and Graphics Cards

[Robert Simandl]

Not sure whether to post this here or in a hardware topic area, figured I'd try here first since I'm looking at hardware to specifically run TMPGenc software (Smart Renderer, Video Mastering Works, Authoring Works).

My current PC is a FIRST generation (as in over 10 years old) Core i7 and looking to upgrade to Ryzen 3900X and X570 mobo. Already bought those, looking for a video card that will work well with this combo for video editing and will provide hardware acceleration for TMPGenc software. No gaming whatsoever.

I see lots of info for nVidia CUDA, but very little on TMPGenc using hardware acceleration for AMD video cards. Is that even a thing?

Anyway, looking for suggestions on video cards for use in hardware acceleration for editing with the TMPGenc Smart Renderer, Video Mastering Works, and Authoring Works applications, either nVidia or AMD. Not particularly loyal to either brand, just looking for what works best with this software and my above mentioned new hardware.

Moderators if this belongs in a hardware topic area, it won't hurt my feelings if you move it.

Thanks!

[Snoopy329]

I build a PC about a week ago using the Ryzen 3600, B450 mobo and Geforce 1650Super and works great for editing 4k video with TMPGenc.

[Robert Simandl]

Thanks!

[mbextreme]

Hi:
A quick question to those who uses TMPGenc programs (I like them very much, Smart Renderer in particular). I have a computer with i7-4790 and GTX 1050Ti. I for sure have checked "use Nividia for decoding" but when monitoring with CPUID, I can see that the program uses almost 100% of all CPU cores, and only a few % of the GPU. Thus it takes a long time on any decoding.

Anything I am missing here?

Thanks in advance.


[Attachment 63012 - Click to enlarge]

[poisondeathray]

Hi:
A quick question to those who uses TMPGenc programs (I like them very much, Smart Renderer in particular). I have a computer with i7-4790 and GTX 1050Ti. I for sure have checked "use Nividia for decoding" but when monitoring with CPUID, I can see that the program uses almost 100% of all CPU cores, and only a few % of the GPU. Thus it takes a long time on any decoding.

Anything I am missing here?

Thanks in advance.

Most likely explanation is that the majority of that CPU% is consumed by CPU Encoding, not Decoding . The bottleneck is usually CPU encoding, not decoding

Decoding takes very little CPU% for most formats , maybe a few % . Certainly not 90-95%. So offloading the that small decoding task to GPU will not gain much speed in terms of final output speed

[mbextreme]

poisondeathray, thank you for the prompt reply.

I did type too fast. It should have been "use Nividia for H.264/H.265 ENCODING".
The snap shot above is the final output ENCODING phrase after editing.

I thought the GPU for sure is suppose to take some load off the CPU and make encoding faster.

Any insights?

[MeNeedVidHelp]

poisondeathray, thank you for the prompt reply.

I did type too fast. It should have been "use Nividia for H.264/H.265 ENCODING".
The snap shot above is the final output ENCODING phrase after editing.

I thought the GPU for sure is suppose to take some load off the CPU and make encoding faster.

Any insights?

I personally don't use the "use Nvidia" for encoding. I personally find my encoding times faster without it. What encoding exactly are you getting?

[mbextreme]

In Smart Renderer, most of the times "Master" clips are only re-encoded at the trimming/joining points. The rest are all straight copying, when outputting, so it is very fast.

There are times when mixed with some "Rescue" clips that must be totally re-encoded. I found this process is almost like real time; meaning a 1 hour file would take 1 hour to re-encode.

That's why I asked why it makes no differences whether I choose to use Nvidia or not. I thought the GPU is suppose to help it faster in the encoding process.

[mbextreme]

For comparison, this image shows the same file re-encoding by AVC Any Video Converter.

This program uses the GPU 100%, and re-encodes the same file in 1/10 of the time.


[Attachment 63020 - Click to enlarge]