• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Which new cpu processor for encoding and decoding video.?

Donnie Fisher said:
As I understand it though, the ampere cards have essentially the same nvenc engine setup as the previous turing series - you may as well get a turing card ... and they are the same setup across the whole family ... so a 2060 will have the same nvenc processing unit as a 2080ti (for the nvenc part only that is. )
Click to expand...
That’s assuming you can still buy those cards. I think with the pricing of last gen cards you are no better off buying new. Better resale value etc.

intel’s quick sync is fast but it’s quality is just not there. I doubt things would have improved that much even in 11th gen. I don’t think that’s intel’s design priority. They package these features into their processors to tick boxes as opposed to really offer something tangible.
 
cpu encoding quality is better for lower file sizes i find. but for quick transcodes or streaming while doing other stuff the gpu is good.
 
Spartacus1972 said:
I read that Handbrake doesn't scale beyond 6 cores for encoding.
Click to expand...

It does scale:
119149.png
 
Spartacus1972 said:
I'm not sure what you're trying to show there. Because the bottom of the chart is an 8core cpu, the next 2 up are 6 core, and 5th from the top is an 8 core cpu.

All that tells me, is the obtained score is nothing to do with the number of cores.
Click to expand...

What is tells you is the score is based on multiple factors, of which core are one. Are you going to look at the 5900X vs 5950X results and say cores have nothing to do with it?
 
Spartacus1972 said:
I'm not sure what you're trying to show there. Because the bottom of the chart is an 8core cpu, the next 2 up are 6 core, and 5th from the top is an 8 core cpu.

All that tells me, is the obtained score is nothing to do with the number of cores.
Click to expand...

2700X limited by its AVX2 implementation which was improved in zen2. https://www.anandtech.com/show/14525/amd-zen-2-microarchitecture-analysis-ryzen-3000-and-epyc-rome/9

That's why a 3600 is better for encoding HEVC or h264 than a 2700X.
 
To answer the original question, the best consumer-grade CPU with reasonable availability within your budget is the Ryzen 9 5950x based on benchmarks.

I wanted to highlight a couple of points regarding GPU-based encoding, given previous answers:

NVENC is great for what it is, however the end result is not as good as pure CPU-based encoding. CPU encoding, while slower than GPU, (especially with high quality settings) will give you either better quality or smaller size or to a degree both. The difference is quite noticeable, especially if you have a large collection. People that are quality focused would never go in Handbrake and just click on the "Fast" preset and get it over and done with. There can be a significant amount of tweaks with several options. While NENV compares very well to Fast/Very Fast preset encodes, it does not do so with the more advanced or custom presets.

I would suggest using CPU encoding to rip original media in order to extract the best image quality possible in the smallest file-size possible. You could then use GPU-encoding for the "delivery" of content (for example Plex server) by making use of "hardware transcoding" features. This way you can retain the best "original" content possible and take advantage of quality improvements as time passes. If your original output is lower quality to begin with, no matter how good hardware transcoding gets over the next few generations, it will always be your limiting factor. Re-ripping your entire set of original media in a few years to counter this issue may not be an attractive option, especially if you have a decent-sized collection.

This becomes particularly obvious when we start talking about 4K content where storage space savings or output quality differences can be substantial.

Also, as per above responses, core count plays a role but comparisons are relevant only within the same architecture. Comparing a 12-core Opteron 6174 from 2010 with a Ryzen 3900X based on core count makes no sense when trying to extrapolate performance/scalability.

Even "small" differences in FPS in encoding performance can add up. 20fps difference across different core-count CPUs may not be linear, but it will mean a significant saving of encoding time from what would be your ETA - Such differences add up quickly for a 2hr movie for example. If you encode a movie a week it doesn't really matter I guess, but if you do so 16 hrs a day (or more if automated) then that's a different story.

Ultimately your mileage may vary based on your priorities (for example if you value time over quality or cost) and that can be different from everyone.
 
I know this isn’t answering your question, I just want to throw it out there as I went through the same process. Have you considered just increasing HDD space and then keeping the raw rips?

I played around with encoding for a bit, it was loads of work, loads of time, and you always know the quality isn’t going to be as good as the original.

I just use makemkv and store the raw file, done. You just need a player that can deal with the high bw rips I use and Apple TV and Infuse for this.

If you really want to encode then I’d go with a >=1650 GPU, will be considerably faster than even a 1K CPU.
 
SkyTerran said:
It does scale:
119149.png
Click to expand...

i think that chart explains a lot for my situation.
iv gone from my video encoding machine which uses a xeon e3-1225 v3 (haswell 4 core 4 thread @ stock clocks (3.4ghz all core stock)) to a ryzen 1700 overclocked to all core 3.6ghz 8 core 16 thread.
using handbrake and doing a encode test same file and settings on each machine the intel one averages 6.2fps and the ryzen is averaging 11.6 fps. my xeon is pegged at 100% cpu usage the ryzen is bouncing around 85-92%.
was hoping to see a bigger increase from the ryzen but looking at that anandtech bench test i can see you dont get decent scaling as cores increase. can see in the chart the 5800x gets 69fps and the 5950x gets 98fps. granted clock speeds are down on the 5950x but in an ideal world id expect the 5950x to get to 125fps or more.
 
Last edited:
Tgese huge computationsl tasks are supposed to be where cloud services excel. Ie running the encoder in AWS or Azure. Transferring files to a single vCPU instance, spinning up and encoding using many vCPU, then using a single vCPU for the file transfer back. But it needs lots of patience for those long file transfer times.
 
diminishing returns. i doubt 64 threads would be double the speed of 32 threads. the more threads that get added the greater the losses are.
 
R4z0r said:
I know this isn’t answering your question, I just want to throw it out there as I went through the same process. Have you considered just increasing HDD space and then keeping the raw rips?

I played around with encoding for a bit, it was loads of work, loads of time, and you always know the quality isn’t going to be as good as the original.

I just use makemkv and store the raw file, done. You just need a player that can deal with the high bw rips I use and Apple TV and Infuse for this.

If you really want to encode then I’d go with a >=1650 GPU, will be considerably faster than even a 1K CPU.
Click to expand...

Infuse 6 looks interesting. I've been thinking about the Apple TV when I move to 4k. It seems to be one of the best streaming boxes along with the Nvidia Shield. I want something that will cope with all the HDR formats and Dolby Atmos.
 
You must log in or register to reply here.
Back
Top Bottom