Question Which AMD processor?

[NINaudio]

I'm looking to build a new PC. My most intensive use scenario right now is using Handbrake to backup my movie collection. So I'm looking for the fastest AMD cpu for that under $500. This leaves me looking at 3900x, 5800x (when available), or 5600x. Are the more cores in the 3900X going to give me the most benefit in this use scenario in most cases?

 

[SamMaster]

Anandtech encoding tests

According to the link I found here, it appears that for Handbrake, the 5800X is slightly faster then the 3900X. Of course, that's with a 30 second search effort using a single source, so I encourage you to research more to get a bigger picture.

 

[NINaudio]

Yeah, I've used the bench, but I was more wondering what people's real world experiences have been. In the 4k handbrake encoding threads in the forum, it appears that the 3900x is faster than the 5800x.

 

[Hitman928]

It depends on the encoder used, resolution, and settings. If you're using x264 with 720p sources and the fast preset in Handbrake, the 5800x will be faster as it won't use even all of the 5800x available cores effectively so the fewer, faster cores will be better. If you're doing x265 with 4K sources and (probably) just about any preset, the 3900x will be faster. With x265 and especially 4K, handbrake will be able to use more than the 5800x's 8 cores effectively allowing the 3900x to perform better. This is all general guidance though and can depend even on what's happening in the source video.

Edit: Added some details.

 

[NINaudio]

It depends on the encoder used, resolution, and settings. If you're using x264 with 720p sources and the fast preset in Handbrake, the 5800x will be faster as it won't use even all of the 5800x available cores effectively so the fewer, faster cores will be better. If you're doing x265 with 4K sources and (probably) just about any preset, the 3900x will be faster. With x265 and especially 4K, handbrake will be able to use more than the 5800x's 8 cores effectively allowing the 3900x to perform better. This is all general guidance though and can depend even on what's happening in the source video.

Thanks, I do most of my encoding from bluray to 1080p x264 with an RF of 14 using very slow and DTS-HD audio passthrough. It's a custom preset I've made based off of the HQ 1080p30 surround preset. I'm more concerned with the picture quality being better than outright speed of the encoding process. What are your thoughts based on this info?

All in all on my current desktop, an encode of a 2 hour movie takes around 10 hours. Using the same settings on my laptop (AMD 5 Ryzen 4600H), the encode takes about 4 hours.

 

[Hitman928]

Thanks, I do most of my encoding from bluray to 1080p x264 with an RF of 14 using very slow and DTS-HD audio passthrough. It's a custom preset I've made based off of the HQ 1080p30 surround preset. I'm more concerned with the picture quality being better than outright speed of the encoding process. What are your thoughts based on this info?

All in all on my current desktop, an encode of a 2 hour movie takes around 10 hours. Using the same settings on my laptop (AMD 5 Ryzen 4600H), the encode takes about 4 hours.

1080p puts it in the gray area a bit. Do you think you may move to 4K sources at some point in the future with this machine?

Also, an RF setting of 14 is probably not good. It will give you great picture quality, but your file sizes have to be huge. Pretty much to the point that I question if it is worth re-encoding the source at all? An RF setting at about 20 with a good 1080p source will drastically reduce your file size for a very minimal difference in visual quality, if you can notice it at all.

 

[AnitaPeterson]

Thanks, I do most of my encoding from bluray to 1080p x264 with an RF of 14 using very slow and DTS-HD audio passthrough. It's a custom preset I've made based off of the HQ 1080p30 surround preset. I'm more concerned with the picture quality being better than outright speed of the encoding process. What are your thoughts based on this info?

All in all on my current desktop, an encode of a 2 hour movie takes around 10 hours. Using the same settings on my laptop (AMD 5 Ryzen 4600H), the encode takes about 4 hours.

RF of 14?
At this point, you're treading into NVENC territory.
You might as well use a $90 nVidia GPU and run the encoding on it.
The size will be similar, and you'd cut encoding times to 30% of what you're seeing today, no matter what CPU you use.

 

[NINaudio]

1080p puts it in the gray area a bit. Do you think you may move to 4K sources at some point in the future with this machine?

Also, an RF setting of 14 is probably not good. It will give you great picture quality, but your file sizes have to be huge. Pretty much to the point that I question if it is worth re-encoding the source at all? An RF setting at about 20 with a good 1080p source will drastically reduce your file size for a very minimal difference in visual quality, if you can notice it at all.

4K is in my future, but probably not for another year or two.

PQ is fantastic. I've tried it at higher RF's before and there was too much artifacting and banding in dark scenes for my tastes. I basically ripped specific test scenes at various RF until I found where there was minimal banding/artifacting visible on my TV and left it there. It typically cuts file size down by almost 2/3 compared to the straight rip, so it is still saving a significant amount of room. I'd have to doublecheck when I get home, but I think I have the encoder set to run "very slow". Like I said, I'm concerned with maximum PQ, and I was under the impression that running it in software through handbrake gave better picture quality than running it in hardware through NVidia? Or has that changed recently?

 

[Hitman928]

4K is in my future, but probably not for another year or two.

PQ is fantastic. I've tried it at higher RF's before and there was too much artifacting and banding in dark scenes for my tastes. I basically ripped specific test scenes at various RF until I found where there was minimal banding/artifacting visible on my TV and left it there. It typically cuts file size down by almost 2/3 compared to the straight rip, so it is still saving a significant amount of room. I'd have to doublecheck when I get home, but I think I have the encoder set to run "very slow". Like I said, I'm concerned with maximum PQ, and I was under the impression that running it in software through handbrake gave better picture quality than running it in hardware through NVidia? Or has that changed recently?

Hardware encoders have improved drastically over the years, but software will give you significantly less file sizes at the same picture quality and I don't think any of them can provide the basically lossless quality you are trying to achieve. If you are planning on 4K in the somewhat near future using the same machine you are building now, I would go with the 3900x as going to 4K will extend encode times significantly and will also be able to use the extra 50% mores cores to help run the encode as quickly as possible.

 

[Hitman928]

Here's a nifty site I found where you can compare quality settings and file size comparisons.

x264 vs x265 vs VP8 vs VP9 (examples) | mattgadient.com

Image and video samples of a source encoded via x264, x265, VP8, and VP9.

mattgadient.com

 

[NINaudio]

Hardware encoders have improved drastically over the years, but software will give you significantly less file sizes at the same picture quality and I don't think any of them can provide the basically lossless quality you are trying to achieve. If you are planning on 4K in the somewhat near future using the same machine you are building now, I would go with the 3900x as going to 4K will extend encode times significantly and will also be able to use the extra 50% mores cores to help run the encode as quickly as possible.

Thanks so much for the advice, it is greatly appreciated. I guess the other alternative is to go with what's available now, like a 5600x and upgrade later when parts are available to the 5900x?

 

[Hitman928]

I still question if you can actually see a difference between RF14 and RF20 on a good 1080p source. The Handbrake documentation even mentions that you probably shouldn't use RF lower than ~19 for DVD sources and that higher quality sources should use higher RF values. I'm not trying to tell you what to do, it's your computer/videos do you what you want, but maybe pick a 30 second clip or something and try a blind test of different RF values where you don't know which is which and see if you can really spot the RF14 video. But again, it's up to you.

Yes, you could start with a 5600x and move to a 5900x when it's more available. It will be way faster than your desktop CPU now. If you really don't care about encode times, then you might want to start using x265 as well, it's a more modern encoder and will produce smaller files at the same quality settings (but take longer).

 

[NINaudio]

I still question if you can actually see a difference between RF14 and RF20 on a good 1080p source. The Handbrake documentation even mentions that you probably shouldn't use RF lower than ~19 for DVD sources and that higher quality sources should use higher RF values. I'm not trying to tell you what to do, it's your computer/videos do you what you want, but maybe pick a 30 second clip or something and try a blind test of different RF values where you don't know which is which and see if you can really spot the RF14 video. But again, it's up to you.

Yes, you could start with a 5600x and move to a 5900x when it's more available. It will be way faster than your desktop CPU now. If you really don't care about encode times, then you might want to start using x265 as well, it's a more modern encoder and will produce smaller files at the same quality settings (but take longer).

I care somewhat, in that I'd rather it not take 10 hours to do a 2 hour movie if it could be done in say half the time on a newer processor.

We're critical customers here. My wife can see the difference when I went from 16 to 14 in a dark test scene from a blu ray, she used to work in video editing. It's quite noticeable when watching in the living room. I've always had a fairly critical eye, it's both a blessing and a curse. lol Was a blessing when I used to do quality control for cosmetics where visual blemishes were of major concern. It is a curse when you hand wash your car and can never unsee that stupid scratch...

I'll play around with the RF settings again once I get the new system set up and see if anything has changed.

edited to add: there are some definite, yet subtle, differences between 16 and 20 that I could observe in that link you provided me with. It seems most evident to me in shadow details and facial details as well.

 

[Thunder 57]

Here's a nifty site I found where you can compare quality settings and file size comparisons.

x264 vs x265 vs VP8 vs VP9 (examples) | mattgadient.com

Image and video samples of a source encoded via x264, x265, VP8, and VP9.

mattgadient.com

Very cool, thanks!

 

[fleshconsumed]

5900x would be ideal if you could stretch your budget by $50 and if you could find one in stock. I convert 1080p blurays to HEVC/x265 using 7mbit 2-pass, audio gets converted to AAC LC VBR 5 (custom ffmpeg build). The tests where 5800x matches or exceeds 3900x are somewhat misleading because it depends on how well that particular test utilizes multiple cores. Higher core CPUs will often let your run multiple encodes in parallel without losing a lot of performance. For example if I run single encode on my 3900x with my preferred settings I get about 0.9x speed, but if I run 2 encodes I get about 0.7-0.8x speed each. So if I run a single encode I can process 2-pass movie in about 4 hours, but if I run 2 encodes in parallel I can encode 2 movies in 6 hours. I'm looking to see if I can upgrade my 3900x to either 3950x/5900x/5950x in the future whenever we get more availability/used prices come down. My recommendation still stands at around $500 you should either wait for 5900x or get a used 3950x (in major metro areas you can find them around 520-570), if you can't go above $500/can't wait then get 3900x.

 

[MrTeal]

I care somewhat, in that I'd rather it not take 10 hours to do a 2 hour movie if it could be done in say half the time on a newer processor.

We're critical customers here. My wife can see the difference when I went from 16 to 14 in a dark test scene from a blu ray, she used to work in video editing. It's quite noticeable when watching in the living room. I've always had a fairly critical eye, it's both a blessing and a curse. lol Was a blessing when I used to do quality control for cosmetics where visual blemishes were of major concern. It is a curse when you hand wash your car and can never unsee that stupid scratch...

I'll play around with the RF settings again once I get the new system set up and see if anything has changed.

edited to add: there are some definite, yet subtle, differences between 16 and 20 that I could observe in that link you provided me with. It seems most evident to me in shadow details and facial details as well.

I don't know if it's because I mostly watch movies on what was a very nice TV for black levels at the time (Panasonic ZT60), but I find even a lot of relatively high bitrate rips look terrible because the artifacts on blacks just pop out so much. I'm far from an expert, but when I'm ripping my BluRays I just crank the settings up. I'd rather toss another 10TB drive in the NAS than deal with it.

 

[NINaudio]

Thanks all. I'm going to go with a 5600X for now and upgrade to the 5900X later on when they're available for a reasonable (ie. retail) price.

 

[NINaudio]

Well, got everything all assembled today. It's a huge improvement, about 3.5x faster than previously with the same settings. Thanks so much for the help!

 

[sdifox]

So just remux without compression? Storage is cheap.

Never mind you already got it xd.

 

[guachi]

With an RF of 14 I just wouldn't even bother to use Handbrake. Rip the movie without compression.

 

[NINaudio]

With an RF of 14 I just wouldn't even bother to use Handbrake. Rip the movie without compression.

It typically cuts file size down by almost 2/3 compared to the straight rip.

Maybe you missed this part....

 

[ElFenix]

if you're finding that you're not using all your CPU time, open another handbrake instance and run another encode.

kidna disappointed no one pointed out our own extensive handbrake benchmarking we've done here on the board.

Question - Handbrake 1.3.3 - Benchmark your System - New benchmark criteria

Back in 2018 ElFenix posted the original Handbrake Benchmark Thread. I thought it was cool so I started updating the thread with a results table. With his permission I'm posting an update to that thread. If you want to participate we're going to tighten up the testing methodology (see below)...

forums.anandtech.com

Benchmark your computer @4K with Handbrake 1.1 and H265! NOW WITH ZEN3!

edit: now with HB 1.3.3 results and zen 3 is a monster: https://forums.anandtech.com/threads/benchmark-your-computer-4k-with-handbrake-1-1-and-h265.2544492/post-40376657 new testing regime: If you're up for it let's try it with a more rigorous approach. Instructions below if you want to...

forums.anandtech.com

 

[NINaudio]

if you're finding that you're not using all your CPU time, open another handbrake instance and run another encode.

kidna disappointed no one pointed out our own extensive handbrake benchmarking we've done here on the board.

Question - Handbrake 1.3.3 - Benchmark your System - New benchmark criteria

Back in 2018 ElFenix posted the original Handbrake Benchmark Thread. I thought it was cool so I started updating the thread with a results table. With his permission I'm posting an update to that thread. If you want to participate we're going to tighten up the testing methodology (see below)...

forums.anandtech.com

Benchmark your computer @4K with Handbrake 1.1 and H265! NOW WITH ZEN3!

edit: now with HB 1.3.3 results and zen 3 is a monster: https://forums.anandtech.com/threads/benchmark-your-computer-4k-with-handbrake-1-1-and-h265.2544492/post-40376657 new testing regime: If you're up for it let's try it with a more rigorous approach. Instructions below if you want to...

forums.anandtech.com

I mentioned that I went through them in post #3.

While a good tool, I really wanted to get input from people ripping movies with more focus on quality then using the basic fast presets.

 

[ElFenix]

I mentioned that I went through them in post #3.

While a good tool, I really wanted to get input from people ripping movies with more focus on quality then using the basic fast presets.

those threads aren't bench.

 

[fleshconsumed]

I still think 3900x would have been better than 5600x. Yes, Zen3 is 25% faster per core than Zen2, but 3900x has twice the cores.

I just picked up used 3950x for $400 (heck of a deal, I know) on facebook marketplace, and it's been a really nice improvement over 3900x. When I ran 2 simultaneous encodes on 3900x my CPU was getting pegged and the system was bit slow to respond. With 3950x I get 20% faster encodes, I can run one VM, and my system is still way more responsive. Really happy right now.