Question At what point to AMD CPUs outperform Intel CPUs in Adobe Premiere using H264?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Bavor

Member
Nov 11, 2001
82
18
81
Right now I have a i7-6700K at 4.6 GHz. Most of my video editing is in Adobe Premiere Pro for YouTube at 2K and 4K resolution. My i7-6700K was significantly faster than the R7-1700X system I tried for video encoding with the YouTube 4K preset. Depending on the video, the 1700X took 30% to 50% more time to encode the same video.

If I wanted a Ryzen 3000 series CPU for my new system, does anyone know what model would be fast enough so that I won't be encoding video at a slower speed than the 6700K that used the iGPU to speed up encoding?
 

Bavor

Member
Nov 11, 2001
82
18
81
Yeah I kinda speed read through the thread but seems this is the point.

At the least what you were doing is an apples to oranges comparison. You posed the OP as a CPU vs CPU question, but seems the whole time GPU acceleration was the difference. Nothing wrong with taking advantage of GPU acceleration but it's not the only product with an integrated GPU, plus a dedicated GPU (from anyone) can also provide acceleration. And as has been stated many times GPU accelerated encoding is generally considered inferior to pure CPU, even if it's fine for many most people. In fact if you're looking for the fastest GPU acceleration then even a low end CPU or GPU (with the newest GPU acceleration) might be up with the fastest?

I do appreciate the effort you've gone to running benchmarks and providing data to the community, but the CPU vs CPU benchmark without GPU acceleration should have been the fundamental starting point. Then you can consider GPU acceleration and the convenience/quality of an integrated or dedicated GPU. And again: iGPU acceleration is super convenient and fast and good enough for most people, but not apples to apples compared to pure CPU.

Its not apples to oranges for my use case though. That's the point. I create H.264 video and upload it to YouTube. Many other people do the same thing. Adobe Premiere and Media Encoder use Intel's iGPU to speed up encoding. I'm trying to find something newer from AMD that will also be an improvement in video encoding speed in my use case. I'm just trying the encoding once without the iGPU it to see how much the iGPU speeds up encoding. Other people said using the iGPU cuts their video encoding time in half.

Sure, there is apparently some quality differences, but nobody seems to be able to tell unless they look at the video frame by frame or side by side frame by frame. YouTube's limit on bandwidth for video probably has something to do with that. I know another YouTuber who has at lest 50 times the followers I have. He switched from CPU only encoding to using the iGPU and even asked in one of his videos if anyone can tell a difference in his videos and when he changed his encoding method because he switched to a new method. Nobody could pinpoint the video where he changed to encoding with the iGPU on the Intel CPU. Maybe if I was making nature documentaries people could tell the difference, but in tech and computer gaming related videos nobody seems to be able to tell when watching the videos.

Have a look at clocks while doing this,at default settings a K CPU will use the same TDP for HTT on or off which means a bit of difference in clock speeds.

The CPU is locked to 4.6 GHz all cores. The speed fluctuates between 4,597 and 4,601 MHz. The speed didn't change when hyperthreading was disabled. I have the TDP increased in the UEFI/BIOS to a setting that the CPU probably wouldn't be able to hit without sub ambient cooling.
 

richaron

Golden Member
Mar 27, 2012
1,357
329
136
Its not apples to oranges for my use case though. That's the point. I create H.264 video and upload it to YouTube. Many other people do the same thing. Adobe Premiere and Media Encoder use Intel's iGPU to speed up encoding. I'm trying to find something newer from AMD that will also be an improvement in video encoding speed in my use case. I'm just trying the encoding once without the iGPU it to see how much the iGPU speeds up encoding. Other people said using the iGPU cuts their video encoding time in half.
Sorry bro it's 100% apples to oranges. And in this case a great example of software hiding complexity. I have no problem if you're fine with GPU encoding, in fact I like the idea. But I will make sure it's understood what's actually going on, you are throwing GPU acceleration into the CPU vs CPU encoding question. And thus we open the door to other options such as other CPUs with iGPUs or simply a dGPU upgrade as an answer to your badly phrased or ambiguous OP.

I think it's great you're running benchmarks and I'm looking forwards to more. But I need to make it clear that the CPU vs CPU benchmark (no acceleration) is the baseline. Then we can consider how much faster and convenient a GPU is, and what your options are. Partially because if you are asking us (including me) for advice about something involving GPUs then the answer may be outside the scope of your OP.
 
Last edited:

Bavor

Member
Nov 11, 2001
82
18
81
Sorry bro it's 100% apples to oranges. And in this case a great example of software hiding complexity. I have no problem if you're fine with GPU encoding, in fact I like the idea. But I will make sure it's understood what's actually going on, you are throwing GPU acceleration into the CPU vs CPU encoding question. And thus we open the door to other options such as other CPUs with iGPUs or simply a dGPU upgrade as an answer to your badly phrased or ambiguous OP.

I think it's great you're running benchmarks and I'm looking forwards to more. But I need to make it clear that the CPU vs CPU benchmark (no acceleration) is the baseline. Then we can consider how much faster and convenient a GPU is, and what your options are. Partially because if you are asking us (including me) for advice about something involving GPUs then the answer may be outside the scope of your OP.

My OP was very clear, "If I wanted a Ryzen 3000 series CPU for my new system, does anyone know what model would be fast enough so that I won't be encoding video at a slower speed than the 6700K that used the iGPU to speed up encoding?"

I didn't say I was looking at other Intel CPUs with iGPUs. I didn't say I'd be changing my current RTX 2080 GPU. I didn't say I was looking to replace my RTX 2080 with another GPU. Also, the R5 2400G and R5 3400G don't have the ability to use their onboard Vega 11 GPU to speed up H.264 video encoding in Premiere. Different GPUs such as a RTX 2080 ti or RTX Titan probably won't show a significant difference in video encoding or be worth the expense based on my experience with different GPUs and H.264 encoding. Previous tests done by others have shows that different Intel and AMD GPUs show a minimal change in encoding time. The difference between a GTX 1070 and RTX 2080 ti when enabling CUDA acceleration can be as little as 10-20 seconds on a 10 minute 4K video. The cost of a Quadro card isn't likely to show a good return on investment to speed up video encoding either based on how little difference there is on lower end GPUs.

The baseline is my current computer with my current CPU. What I'm comparing it to are my options for a new system with a CPU from AMD. I'm not asking for GPU upgrade options. Nowhere did I ask about replacing my RTX 2080. Nowhere did I ask about newer Intel CPUs with iGPUs. That's why my current system with the 6700K and iGPU acceleration is the baseline for the comparison.

This ins't an Intel vs AMD thread. If I was looking at newer Intel CPUs, then the answer would be 9900K or KS for my use case no questions asked. Its not looking into what's the difference when I hinder my current system's performance by not using all available features compared to the newest CPUs thread.
 

Hitman928

Diamond Member
Apr 15, 2012
5,324
8,015
136
My OP was very clear, "If I wanted a Ryzen 3000 series CPU for my new system, does anyone know what model would be fast enough so that I won't be encoding video at a slower speed than the 6700K that used the iGPU to speed up encoding?"

I didn't say I was looking at other Intel CPUs with iGPUs. I didn't say I'd be changing my current RTX 2080 GPU. I didn't say I was looking to replace my RTX 2080 with another GPU. Also, the R5 2400G and R5 3400G don't have the ability to use their onboard Vega 11 GPU to speed up H.264 video encoding in Premiere. Different GPUs such as a RTX 2080 ti or RTX Titan probably won't show a significant difference in video encoding or be worth the expense based on my experience with different GPUs and H.264 encoding. Previous tests done by others have shows that different Intel and AMD GPUs show a minimal change in encoding time. The difference between a GTX 1070 and RTX 2080 ti when enabling CUDA acceleration can be as little as 10-20 seconds on a 10 minute 4K video. The cost of a Quadro card isn't likely to show a good return on investment to speed up video encoding either based on how little difference there is on lower end GPUs.

The baseline is my current computer with my current CPU. What I'm comparing it to are my options for a new system with a CPU from AMD. I'm not asking for GPU upgrade options. Nowhere did I ask about replacing my RTX 2080. Nowhere did I ask about newer Intel CPUs with iGPUs. That's why my current system with the 6700K and iGPU acceleration is the baseline for the comparison.

This ins't an Intel vs AMD thread. If I was looking at newer Intel CPUs, then the answer would be 9900K or KS for my use case no questions asked. Its not looking into what's the difference when I hinder my current system's performance by not using all available features compared to the newest CPUs thread.

After looking at more links, it seems like if you match the quality of the iGPU output by reducing the CPU quality then possibly even the latest 8 core CPUs from intel or AMD would be able to match or beat your current iGPU speed. Unfortunately I don't have the hardware to actually test this out but I'll repeat what I said earlier in the thread, if you're happy with the quality from the iGPU and you're only doing one video at a time, it doesn't seem like an upgrade to a more powerful CPU would be worth it unless you want it for other reasons as well.
 
  • Like
Reactions: spursindonesia

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
The baseline is my current computer with my current CPU. What I'm comparing it to are my options for a new system with a CPU from AMD. I'm not asking for GPU upgrade options. Nowhere did I ask about replacing my RTX 2080. Nowhere did I ask about newer Intel CPUs with iGPUs. That's why my current system with the 6700K and iGPU acceleration is the baseline for the comparison.

If you're happy with the performance of your current system, I see no reason to upgrade yet. Wait until at least the 4000-series Ryzens.

My OP was very clear, "If I wanted a Ryzen 3000 series CPU for my new system, does anyone know what model would be fast enough so that I won't be encoding video at a slower speed than the 6700K that used the iGPU to speed up encoding?"

It seems to me you've already answered your own question with the 3800X benchmarks. Or you could take the easy way, and buy a 3950X. The AMD encoding monster. Intel-wise that would be the 9900K(S), as you point out yourself. You can't really go wrong with either of those, since both are top-notch CPUs.

Either there is a finite budget, or money is no object. To give further recommendations, we'll need to know which it is.
 
  • Like
Reactions: spursindonesia

Bavor

Member
Nov 11, 2001
82
18
81
I know Adobe has improved scaling in Premiere over the past few years, but I'm not sure if adding more cores and threads will help as much as I thought with H.264 encoding after seeing how the 6700K does without using the iGPU compared to the 3800X.

i7-6700K with iGPU: 14 minutes 59 seconds
i7-6700K without iGPU: 21 minutes 12 seconds
r7 3800X:14 minutes 37 seconds

6700K encoding without the iGPU:
t0PwNcJ.jpg


Doubling the cores/threads doesn't come near to doubling CPU only encoding performance.

After looking at more links, it seems like if you match the quality of the iGPU output by reducing the CPU quality then possibly even the latest 8 core CPUs from intel or AMD would be able to match or beat your current iGPU speed. Unfortunately I don't have the hardware to actually test this out but I'll repeat what I said earlier in the thread, if you're happy with the quality from the iGPU and you're only doing one video at a time, it doesn't seem like an upgrade to a more powerful CPU would be worth it unless you want it for other reasons as well.

I'll have to mess around with the encoding settings to see what settings I need to change to do that. I haven't tried it in the past because of all the different options and how many different variables there are to change.

If you're happy with the performance of your current system, I see no reason to upgrade yet. Wait until at least the 4000-series Ryzens.

It seems to me you've already answered your own question with the 3800X benchmarks. Or you could take the easy way, and buy a 3950X. The AMD encoding monster. Intel-wise that would be the 9900K(S), as you point out yourself. You can't really go wrong with either of those, since both are top-notch CPUs.

Either there is a finite budget, or money is no object. To give further recommendations, we'll need to know which it is.

At this point, I'll probably wait to upgrade. I was hoping a newer system with more cores, better IPC, faster RAM, and newer hardware would greatly reduce video encoding times and also deliver higher FPS in gaming now and in the future also with being more "future proof" since I have very limited upgrade options with my current system compared to newer AMD hardware. My only CPU upgrade in a 7700K. I can only use on NVMe SSD in my current system without add in cards. Even then, the limited PCI Express lanes may become an issue when using files from both drives. I've also seen in some 3950X reviews that some game engines still don't work well with 16 cores and show reduced performance and more stuttering and lower minimum and 1% low FPS.

I sold off some computer hardware I wasn't using along with the PC I had connected to my TV and thought between that and selling most of my current system's parts after building a new system that the upgrade costs wouldn't be high. If I'm going to need a $750 CPU to maybe show improvements then i'll wait and see what is available next year.

Maybe I should just learn how to use other video editing software that has better CPU core/thread scaling then look at upgrade options. I've been using Premiere because its the software that I learned to edit videos on and I get the creative cloud subscription for a very low price through a former university I attended.
 

Bavor

Member
Nov 11, 2001
82
18
81
I ran one last test using the 3600 MHz RAM from the 3800X PC in the 6700K system. I couldn't get the RAM to run at 3600 MHz on my 6700K system. However it ran at 3333 MHz without issue on my 6700K system. Doing the same video encoding, it took 14 minutes and 41 seconds, so no significant difference between that and the 3800X system with the 3600 MHz RAM.

i7-6700K with iGPU: 14 minutes 59 seconds
i7-6700K with iGPU and 3333 MHz RAM: 14 minutes 41 seconds
i7-6700K without iGPU: 21 minutes 12 seconds
r7 3800X:14 minutes 37 seconds

Faster RAM is cheap and I can get that then wait to see what AMD and Intel have available next year.
 

Staples

Diamond Member
Oct 28, 2001
4,952
119
106
A little off topic but is it possible to use quick sync in Premier if you are using a discrete video card? (I use another editor and have never used Pr)
 

Bavor

Member
Nov 11, 2001
82
18
81
My 3950X arrived today. So far its slower with H.264 video encoding than my i7-6700K was. Its also slower than the 3800X was.

i7-6700K with iGPU: 14 minutes 59 seconds
i7-6700K with iGPU and 3333 MHz RAM: 14 minutes 41 seconds
i7-6700K without iGPU: 21 minutes 12 seconds
r7 3800X:14 minutes 37 seconds
R9 3950X: 15 minutes 10 seconds

The 15 minutes 10 seconds time was after some UEFI/BIOS tweaks to increase the speed and increase the power limits. Those tweaks also raised the Cinebench R20 score by about 300 points. I think the issue is that Premiere doesn't properly support enough threads to take advantage of the 3950X and most of the cores of the 3950X are running at 3.5 GHz or slower during the encoding process. I tried encoding again with Ryzen master running so I could watch the core speed. With the 3800X, all of the cores stayed at 4.1+ GHz during the entire encoding time. I'll try an all core overclock on the 3950X to see if that helps with encoding time. Premiere has definitely increased its ability to use more threads, but its not to the point where the 3950X has any advantage over CPUs with less cores it seems.

x9QRe3l.jpg
 

Bavor

Member
Nov 11, 2001
82
18
81
So sorry. You better wait until....

What can I do / pay to take that mess off your hands :)

Its all within the return windows of Amazon fortunately. I should have gone with the 9900K even though its a dead end platform. I had a chance to get a 9900KS at MSRP recently and passed on it.
 

Kedas

Senior member
Dec 6, 2018
355
339
136
Have you tried disabling one die so basically turn it in a 3800X or disable SMT to reduce the number of threads provided?

It clearly isn't using al your CPU power, you may want to use a better program, or be patient with their slow development for (AMD) high core count support.
Does it use the MKL library? you can try that 'fix' also.

edit: Can you start 2 projects at the same time? (to increase the number of threads used)

(My PC will be ready in the next few days, assuming all parts work)
 
Last edited:

Zstream

Diamond Member
Oct 24, 2005
3,396
277
136
I think you’re missing out. With that many threads available, you could likely multitask without any loss, or degradation.
 

Bavor

Member
Nov 11, 2001
82
18
81
Have you tried disabling one die so basically turn it in a 3800X or disable SMT to reduce the number of threads provided?

I've run the same encoding test on a 3800X system that used the same motherboard, 3600 MHz RAM, and a similar RTX 2080 video card. It only saved me a few seconds over a 6700K system using 3333 MHz DDR4 RAM. I was hoping double the cores/threads of a 3800X would give a significant improvement. Also, scrolling through the timeline and playback at a faster speed while editing are a lot smoother with the 6700K than the 3800X.

It clearly isn't using al your CPU power, you may want to use a better program, or be patient with their slow development for (AMD) high core count support.
Does it use the MKL library? you can try that 'fix' also.

I learned video editing on Premiere Pro and I'm sure I could learn to use other software, but I don't do video editing professionally, so I was hoping not to have to learn other video editing software since Adobe has been updating Premiere to use more CPU cores/threads. I tried the MKL fix with no real change in encoding time.

edit: Can you start 2 projects at the same time? (to increase the number of threads used)

(My PC will be ready in the next few days, assuming all parts work)

I usually only do one video every few days, so having the ability to encode one and edit the other or encode two at the same time doesn't help much unfortunately.

I think you’re missing out. With that many threads available, you could likely multitask without any loss, or degradation.

I haven't had that be a major issue. With the 6700K system I have no trouble gaming, streaming, and recording at the same time. The newer version of NVENC means that the CPU load is very low for recording and streaming and with the bandwidth of Twitch and YouTube for streaming nobody can tell the difference. Despite AMD's marketing BS, nobody streams at lossless 4K 60+ FPS, so you actually don't have a significant difference in streaming with a R7, R9, or i7 CPU. The simple fact is there aren't any streaming platforms that support lossless 40K 60+ FPS uploads, or none that are free or low cost.

Also, I haven't noticed any improvement in day to day use with the 3950X over the 6700K so far. Its not like having Chrome open while playing a game and using Teamspeak taxes the system. With only doing a few videos a week, having increased ability to multitask while encoding a video isn't worth the cost of a new system. With the improvement the 3800X had over the 1700X in video encoding, I was hoping something with twice as many cores and threads would significantly reduce encoding time. The 1700X took about 5 minutes longer (19-20 minutes vs 14-15 minutes) than the 6700X encoding a ~9 minute 4K H.264 video. The 3800X was able to match the 6700K and even be slightly faster. I was hoping the 3950X would be an even bigger improvement. I saw the usage of each core/thread in task manager with the 3800X and assumed incorrectly that the workload on each core/thread would scale across the cores of the 3950X in a similar manner. Instead Premiere gives less of a workload to each core/thread and encoding actually takes longer.
 
  • Like
Reactions: Arkaign

moinmoin

Diamond Member
Jun 1, 2017
4,956
7,675
136
There are only 6 threads with regular usage above 70%, so 16 cores appear to be overkill for your use case, never mind 32 threads. If you keep the 3950X my suggestion would be to set CPU affinity for Premiere Pro to the 8 real cores of the better CCD of the two. This can be adapted once should Adobe ever support more than 8 threads for encoding.
 
  • Like
Reactions: spursindonesia

Bavor

Member
Nov 11, 2001
82
18
81
There are only 6 threads with regular usage above 70%, so 16 cores appear to be overkill for your use case, never mind 32 threads. If you keep the 3950X my suggestion would be to set CPU affinity for Premiere Pro to the 8 real cores of the better CCD of the two. This can be adapted once should Adobe ever support more than 8 threads for encoding.

I tried several different combinations of setting affinity for Premiere and for Media Encoder. The best results with the 3950X are the same(within 2 seconds) encoding time as the 6700K.
 

Kedas

Senior member
Dec 6, 2018
355
339
136
According to these test you should see an improvement between 3950X and 3800X of at least 10%
Also the intel CPUs don't increase much going from 8 to 16 cores.
 

Zstream

Diamond Member
Oct 24, 2005
3,396
277
136
I won’t quote the post, as it’s a pain in mobile, but you’re methods of multitasking is much different than mine I guess. I have YouTube playing a video, game running, outlook up, chrome up, and a few VM’s at the same time, and a few other programs like Lightroom, paint.net, discord etc..

Maybe return the chip and get three or four monitors, so you can really multitask.
 

Bavor

Member
Nov 11, 2001
82
18
81
According to these test you should see an improvement between 3950X and 3800X of at least 10%
Also the intel CPUs don't increase much going from 8 to 16 cores.

I saw that before and was hoping for bigger performance improvements. The difference is probably due to different effects and different types of source material. I was hoping that the 3950X would make doing longer videos less time consuming.

It doesn't appear that anything is wrong with my system. All other benchmarks I run are average above average results for a 3950X system. For example, I'm getting 9,400+ on Cinebench R20 and ~4100 on Cinebench R15. My Firestrike Physics score is above 32,000. TImeSpy CPU score is about 14,000. My NVMe SSD is getting 5,000+ MB/s read speed and and just under 4,800 MB/s write speed in CrystalDiskMark 7.0.0 x64.
 

Bavor

Member
Nov 11, 2001
82
18
81
Gamers nexus latest video on their new Threadripper system explains some of what i was experiencing. Under certain workloads, their 8080k was faster than their latest threadripper system due to the effects used. Results are still variable depending on use case.