Cpu limitations? Ryzen 7 1700 only utilizing 60% of cpu during rendering.. No Hardware Acceleration!

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

Solved by

Apr 27, 2018 01:55

I do weekly videos for the Young Ladies Operator Net a ham radio thing for us ladies and I have a Ryzen 7 1700 with two NVME drives for editing work.. We originally record using OBS then drag the files from the 8TB cold storage drive into the first NVME drive then write to the second NVME drive.. Here is some screen shots of my issues.

First the CPU only utilizes 37% what does it take to hammer the cpu and actually speed up the process?

Second the videos are about 1hr 32 minutes long each week and it always seems to take about 30 minutes to process however I would think since this computer is dedicated rendering box that we could improve this by about 20% My gut tells me that this software is having issues knowing how much to push to the cpu to get rendered.. Anythoughts on how to unlock my software to use it to its fullest?

https://drive.google.com/file/d/1TLXAwbfpuO6J1NVriu1EkV7kLwztipRF/view?usp=sharing

https://drive.google.com/file/d/18pU94pBJQ6W7r5jbMQfO4UqpGgiHY_hD/view?usp=sharing

What can I do to help speed things up and take advantage of my processing power?

This was check marked however is now uncheckmarked but during the process to render I did not choose enable Hardware Acceleration..

https://drive.google.com/file/d/12fnN1vbINkRrb9bz4G0A-P6V_otlJNz8/view?usp=sharing

Streaming PC
CPU Specs:

Operating System: Windows 10 Pro 64-bit
Processor: AMD RYZEN 7 1700 8-Core 3.0 GHz (3.7 GHz Turbo) Socket AM4 65WProcessor
Overclock: 3.8GHzAir CooledCase
Cooling: 4 x120mm fansPC
Case: V31 Thermaltake Case
Memory:Graphics Card: EVGA GeForce GTX 1060 SC GAMING, 6GB GDDR5
Motherboard: ASRock X370 Killer SLI/ac AM4 AMD Promontory X370 SATA 6Gb/s USB 3.0 HDMI ATX AMD
Power Supply: EVGA 750 Watt
Keboard and Mouse
Hardware: Logitech K400r
Mouse: Razer DeathAddder ModelRZ01-0084 Gaming
MouseMouse Pad: Rosewill RGMP-700 Extended PRO Gaming Mouse Pad
Headphones: Tascam th-200x
Microphone: Logitech Webcam C270
Microphone: Lyx Pro SMU-1
Headphone and Mic Default: Plantronics Rig 800HD
Monitors: 40 Inch Hisense TV and 25 Inch Acer IPSORICO
7 Ports USB 3.0 PCI Express Card – Interface USB 3.0 5-Port Add On Cards and 2 Rear USB3.0 Port Express Card for

PC DesktopHard Drives
2 NVME Drives
C:\ Main OS Samsung 250GB 750 EVO SSD SATA III 2.5″ MZ-750250BW
D:\ Storage Western Digital Red 8TB

This message was edited 7 times. Last update was at Apr 29. 2018 04:40

cyberfun3

Newbie Joined: Mar 13, 2018 10:15 Messages: 18 Offline

Apr 27, 2018 12:29

Looks like your video card is working well, since you have low cpu usage? On produce did you enable Hardware accelleration?
What is your video card? AMD or Nvidia?

On my setup, 1hr mixed video and still images, to 1080p, 50mbps, MP4, h264 is about 40-45 mins, with HA enabled. My cpu is high at around 80-95%, gpu engine is low at 10-30%, using GT 730, on old cuda driver. AMD FX8300 8core 3.3Ghz,16GbDDR3
256Gb SSD,GTX1050 2Gb.Nvidia 391.24

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

Apr 27, 2018 13:38

Quote Looks like your video card is working well, since you have low cpu usage? On produce did you enable Hardware accelleration?
What is your video card? AMD or Nvidia?

On my setup, 1hr mixed video and still images, to 1080p, 50mbps, MP4, h264 is about 40-45 mins, with HA enabled. My cpu is high at around 80-95%, gpu engine is low at 10-30%, using GT 730, on old cuda driver.

Ok to answer your question Hardware Acceleration has been disabled so we do not belive the gpu was being worked at all.. However its an Nvidia GTX 1060 for more info you can visit my website and under gaming and equipment is my hardware information for both of my systems.. This unit is the streaming pc.. so its lorentedforddotcom

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

Apr 27, 2018 13:46

Quote Looks like your video card is working well, since you have low cpu usage? On produce did you enable Hardware accelleration?
What is your video card? AMD or Nvidia?

On my setup, 1hr mixed video and still images, to 1080p, 50mbps, MP4, h264 is about 40-45 mins, with HA enabled. My cpu is high at around 80-95%, gpu engine is low at 10-30%, using GT 730, on old cuda driver.

So i might have lied.. I looked and it appears that Enable OpenCL technology to speed up video effect preview/render was check marked.. However Hardware Acceleration was not check at the bottom when i started the reneder.

So now this has been unchecked as well..

https://drive.google.com/file/d/12fnN1vbINkRrb9bz4G0A-P6V_otlJNz8/view?usp=sharing

JL_JL

Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline

Apr 28, 2018 20:23

lorentedford, probably a few things going on. Your CPU load linked pic shows you are using AMD SMT technology, so the most you can "hammer" your cpu as you desire is about 50% based on WIN load charts as the metric. SMT takes one physical proc and shows load for both threads on the proc (2 graphs per proc), however, the proc can only achieve 100% load so the sum of the two threads at most can be 100%. If you want to see higher load, disable SMT in BIOS and your WIN load metric will then be ~74% vs 37%, however, more than likely no significant change in encode times, probably a little longer encode times but you'll get that CPU "hammer" feeling.

Your other linked pic shows you are producing to a wmv file for PD uploading, so any current generation video card can't be involved in the encode process as they don't support wmv PD encoding.

The specifics of your input media format from OBS was not mentioned, if your main objective is similar quality but reduce the encode speed from the 30 minutes, consider:
1) If you want wmv output, consider using the GPU to decode the timeline if for the input media format decoding is supported by the GPU and PD (pref > HA > decoding)
2) Consider using SVRT if an applicable format is available for your input media, and then upload that produced file format to share outside of PD (Produce > Intelligent SVRT)
3) Consider using a H.264 profile and your GTX1060 to hardware encode vs CPU encoding. CPU load will drop substantially so you won’t get that "hammer" feeling but the encode process will be significantly faster. (Produce > Fast video...technology: > Hardware...encoder) Upload the produced file outside of PD.

Jeff

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

Apr 28, 2018 23:30

Quote lorentedford, probably a few things going on. Your CPU load linked pic shows you are using AMD SMT technology, so the most you can "hammer" your cpu as you desire is about 50% based on WIN load charts as the metric. SMT takes one physical proc and shows load for both threads on the proc (2 graphs per proc), however, the proc can only achieve 100% load so the sum of the two threads at most can be 100%. If you want to see higher load, disable SMT in BIOS and your WIN load metric will then be ~74% vs 37%, however, more than likely no significant change in encode times, probably a little longer encode times but you'll get that CPU "hammer" feeling.

Your other linked pic shows you are producing to a wmv file for PD uploading, so any current generation video card can't be involved in the encode process as they don't support wmv PD encoding.

The specifics of your input media format from OBS was not mentioned, if your main objective is similar quality but reduce the encode speed from the 30 minutes, consider:
1) If you want wmv output, consider using the GPU to decode the timeline if for the input media format decoding is supported by the GPU and PD (pref > HA > decoding)
2) Consider using SVRT if an applicable format is available for your input media, and then upload that produced file format to share outside of PD (Produce > Intelligent SVRT)
3) Consider using a H.264 profile and your GTX1060 to hardware encode vs CPU encoding. CPU load will drop substantially so you won’t get that "hammer" feeling but the encode process will be significantly faster. (Produce > Fast video...technology: > Hardware...encoder) Upload the produced file outside of PD.

Jeff

Great thought on SMT but i don't think it exactly works that way.. Please watch this video and tell me your thoughts.

It might take awhile for it to upload and render but what your going to see is in the first half 10 minute render failed which has been brought down to one minute and twenty three seconds and the restart in real time continued.. At the end you will see cinebench running where all 16 threads are pegged at 100% usage.. I really think its some thing in Cyberlink please direct me to understand more about the SMT and exactly why it has to be disabled for cyberlink power director.. Here is what i understand from wiki and AMD concerning links will be below.

https://en.wikipedia.org/wiki/Simultaneous_multithreading

Here is some third party testing of similar system.
http://www.hardwaresecrets.com/does-disabling-smt-on-a-ryzen-7-cpu-improves-performance/

Here is the conclusion of the testing.
http://www.hardwaresecrets.com/does-disabling-smt-on-a-ryzen-7-cpu-improves-performance/7/

https://www.amd.com/en/products/cpu/amd-ryzen-7-1700#product-specs

From everything i have gathered so far I was leaning towards possibly a software related issue and utilzing the newwer Ryzen 7 1700 cpu.. Maybe it could be the cross ccx latency not sure what is causing things to not use the entire cpu resources..

Here is an article i found on Cross ccx latency.
https://www.pcper.com/reviews/Processors/AMD-Ryzen-and-Windows-10-Scheduler-No-Silver-Bullet?page=1

Again not sure if this applies to this application.. I don't know or understand what all is under the hood using Power Director 15.

JL_JL

Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline

Apr 29, 2018 07:27

Did you try with SMT off and see your load and encode time difference?

Did you try using your GTX1060 to encode or SVRT for your wanted speed increase? For at least what little you showed of your timeline, simple transcoding, no real editing, SVRT and/or GPU encoding would perform very nice.

Glanced at several of the links and most implied the same as what I said, SMT will pretty much be a tossup on performance, a little less a little more. Unaware how cinebench relates to reality, they may be fully subscribing 16 threads. The single proc load was good ~10% as one would have guessed. I don't think it's a Ryzen specific PD issue, Intel with Hyperthreading would behave the same with PD. For Hyperthreading it's maybe a 5-10% encoding benefit depending on encoding formats utilized with cpu encoding.

At ~2:26 in your posted video you have shadow files being generated by a secondary PD process, orange movie clip in left bottom corner of media library 2018-04-26 18-41-37... clip. Controled with pref > General. This adds additional CPU load vs the encoding. Hard to guess how much as it was PD15 release version specific. Best to turn off or at least let it finish (green movie clip icon in the corner) prior to any form of PD benchmarking encode performance. About the time and latter, your "unloaded" CPU load is maybe 20-30%, very likely the shadow file generation load.

Jeff

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

Apr 29, 2018 07:40

Quote Did you try with SMT off and see your load and encode time difference?

Did you try using your GTX1060 to encode or SVRT for your wanted speed increase? For at least what little you showed of your timeline, simple transcoding, no real editing, SVRT and/or GPU encoding would perform very nice.

Glanced at several of the links and most implied the same as what I said, SMT will pretty much be a tossup on performance, a little less a little more. Unaware how cinebench relates to reality, they may be fully subscribing 16 threads. The single proc load was good ~10% as one would have guessed. I don't think it's a Ryzen specific PD issue, Intel with Hyperthreading would behave the same with PD. For Hyperthreading it's maybe a 5-10% encoding benefit depending on encoding formats utilized with cpu encoding.

At ~2:26 in your posted video you have shadow files being generated by a secondary PD process, orange movie clip in left bottom corner of media library 2018-04-26 18-41-37... clip. Controled with pref > General. This adds additional CPU load vs the encoding. Hard to guess how much as it was PD15 release version specific. Best to turn off or at least let it finish (green movie clip icon in the corner) prior to any form of PD benchmarking encode performance. About the time and latter, your "unloaded" CPU load is maybe 20-30%, very likely the shadow file generation load.

Jeff

From my past experience with a 1 hr 32 minute video under hardware encoding it typically takes about 38 minutes to complete with the Nvidia GTX 1060..

Cpu does seem to be faster by about almost 10 minutes but its interesting that its not really using the full cpu load of the system but rather caping its self to a strick percentage..

I am going to do some playing around later on turning off SVRT is not yet an option i can do because i use xsplit for streaming and it utilizes almost a constant 80% with hyper threading on if i was to turn hyper threading off i would be able to stream at my 1920x1080 @ 60fps with Slow encoding preset.

JL_JL

Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline

Apr 29, 2018 08:52

Quote From my past experience with a 1 hr 32 minute video under hardware encoding it typically takes about 38 minutes to complete with the Nvidia GTX 1060..

That does not sound right, your video showed H.264 1920x1080 16Mbps encode, my GTX1070 would take under 20 minutes for that. GTX1060 would be about the same.
Furthermore, if that is your input format and SVRT compatible, SVRT would do it in a few minutes.

Quote I am going to do some playing around later on turning off SVRT is not yet an option i can do because i use xsplit for streaming and it utilizes almost a constant 80% with hyper threading on if i was to turn hyper threading off i would be able to stream at my 1920x1080 @ 60fps with Slow encoding preset.

SVRT utilized or not has no effect on xsplit, maybe confused SVRT with SMT?

Anyhow, I've suggested some options, none apparantly of interest, good luck.

Jeff


Filename	PD15_SVRT.png	Download
Description
Filesize	195 Kbytes
Downloaded:	5 time(s)


Filename	PD15_GTX1070.png	Download
Description
Filesize	752 Kbytes
Downloaded:	5 time(s)

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 01, 2018 21:52

Quote

That does not sound right, your video showed H.264 1920x1080 16Mbps encode, my GTX1070 would take under 20 minutes for that. GTX1060 would be about the same.
Furthermore, if that is your input format and SVRT compatible, SVRT would do it in a few minutes.

SVRT utilized or not has no effect on xsplit, maybe confused SVRT with SMT?

Anyhow, I've suggested some options, none apparantly of interest, good luck.

Jeff

Give me a few days I just had Surgery yesterday I promise I will look into this but we have tested an i7 8700 6 core 12 thread and with hardware acceleration disabled 1080p render it does it with all threads pinned.. I am not seeing this with Ryzen 7 1700 we are testing and comparing the platforms here.. I also just got a new Intel i7 8700k for my gaming pc and I will install PD15 on it to see what i get in performance comparasons to the Ryzen 7 1700.. The problem is I just had Shoulder Manupulation Surgery and require physical therapy every day until they say otherwise.. So I am a tad bit busy in health concerns..

I will try to get some thing up by next friday hopefully so sorry about this..

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 02, 2018 12:44

Here is a test with an Intel i7 8700K over clocked at 5.0 Ghz water cooled with Cinebench test and scores at the end using no Hardware Acceleration.

Not sure why a computer with less threads can out perform a Ryzen 7 1700 at 4.8 Ghz over clocked and cooled as well... Still trying to figure out why Power Director seems to underperform on Ryzen..

You would think that the Ryzen Box would have performed much better under load with 8 cores and 16 threads. Regardless of SMT settings.. Just my thoughts..

Jack Doon

Newbie Location: Rotterdam Netherlands Joined: Dec 14, 2017 22:07 Messages: 2 Offline

May 02, 2018 16:52

You should have bought an Intel cpu
Intel cpu's has Quicksync hardware encoders and are much faster then any amd solution.

This message was edited 2 times. Last update was at May 02. 2018 16:57

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 02, 2018 17:31

Quote You should have bought an Intel cpu
Intel cpu's has Quicksync hardware encoders and are much faster then any amd solution.

The point here is with out hardware Acceleration.. As far as gpu's the Ryzen 7 1700 has a Nvidia gtx 1060 and the Intel i7 8700k has Nvidia GTX 1070.. Using hardware encoding is actually slower then cpu encoding. From past experience..

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 09, 2018 12:42

No one has thoughts on Ryzen ?

pmikep

Senior Member Joined: Nov 26, 2016 22:51 Messages: 285 Offline

May 09, 2018 18:46

The guys in the Magix forum recommend buying no more than a hex core processor, saying something like more cores won't be used with MEP.

Of course, that's apples to oranges. I know that PD uses all my cores when I don't use HA on my old AMD quad core.

And for the videos I do (which is nothing crazy), HA is faster than CPU. (GTX 960.) But then, it's an old processor with an old instruction set.

This message was edited 1 time. Last update was at May 09. 2018 19:22

IgorTFerreira

Newbie Location: Rio de Janeiro - Brazil Joined: Jul 06, 2016 22:59 Messages: 11 Offline

May 12, 2018 19:20

Quote The guys in the Magix forum recommend buying no more than a hex core processor, saying something like more cores won't be used with MEP.

Of course, that's apples to oranges. I know that PD uses all my cores when I don't use HA on my old AMD quad core.

And for the videos I do (which is nothing crazy), HA is faster than CPU. (GTX 960.) But then, it's an old processor with an old instruction set.

Some weeks ago i opened a thread questioning witch processor is the best for powerdirector (Ryzen 2700x / intel 8700k or ThreadRipper 1950x), no one answerd me! Most answers linked to games benchmarks... I think there aren't huge number of powerdirector/ryzen (or threadripper) users. I'm really worried about the present thread, because i'm considering to upgrade from a 4790k with quicksync enable + R290x to a Ryzen 2700x, but i'm not sure if i'll get enough performance increase! I can't find benchmarks or user experiences about new AMD processors (even "old" ryzen 1) and powerdirector...
that's bad...
Does anyone have any ideia if upgrade the R290x to a Nvidia 1060 or even 1080 will make difference (to timeline editing)?

I will keep my 4790k for some more time...

Thanks for reading!

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 13, 2018 16:09

Quote

Some weeks ago i opened a thread questioning witch processor is the best for powerdirector (Ryzen 2700x / intel 8700k or ThreadRipper 1950x), no one answerd me! Most answers linked to games benchmarks... I think there aren't huge number of powerdirector/ryzen (or threadripper) users. I'm really worried about the present thread, because i'm considering to upgrade from a 4790k with quicksync enable + R290x to a Ryzen 2700x, but i'm not sure if i'll get enough performance increase! I can't find benchmarks or user experiences about new AMD processors (even "old" ryzen 1) and powerdirector...
that's bad...
Does anyone have any ideia if upgrade the R290x to a Nvidia 1060 or even 1080 will make difference (to timeline editing)?

I will keep my 4790k for some more time...

Thanks for reading!

Yeah I would hold off on buying anything ryzen if your wishing to use Power Director.. It seems that they have a serious issue witht he code that wont actually let it use the cores.. I however do have an i7 8700k for a gaming computer so i just use the gaming computer to do my renders really quickly.. I am hoping that future graphics card updates actually make my GTX 1070 faster at rendering video footage.. Maybe we will see..

IgorTFerreira

Newbie Location: Rio de Janeiro - Brazil Joined: Jul 06, 2016 22:59 Messages: 11 Offline

May 13, 2018 16:21

Quote

Yeah I would hold off on buying anything ryzen if your wishing to use Power Director.. It seems that they have a serious issue witht he code that wont actually let it use the cores.. I however do have an i7 8700k for a gaming computer so i just use the gaming computer to do my renders really quickly.. I am hoping that future graphics card updates actually make my GTX 1070 faster at rendering video footage.. Maybe we will see..

HI lorentedford
Thanks for the reply and the suggestion!
I7 8700k with quicksync enable should get VERY fast rendering time (for x264 format, and even HVEC - i think 8700k GPU suport it) on powerdirector...

Quote

Yeah I would hold off on buying anything ryzen if your wishing to use Power Director.. It seems that they have a serious issue witht he code that wont actually let it use the cores.. I however do have an i7 8700k for a gaming computer so i just use the gaming computer to do my renders really quickly.. I am hoping that future graphics card updates actually make my GTX 1070 faster at rendering video footage.. Maybe we will see..

HI lorentedford
Thanks for the reply and the suggestion!
I7 8700k with quicksync enable should get VERY fast rendering time (for x264 format, and even HVEC - i think 8700k GPU suport it) on powerdirector...
On my particular system (i7 4790k, water cooled, 32gb ddr2 ram at 1800mhz and AMD R290x, during edition, the processor usage stays between 60 to 70%. Can you edit 4k footage without any throttle on the timeline with 8700k?

lorentedford

Newbie Location: Olney, IL USA Joined: Feb 16, 2018 12:56 Messages: 15 Offline

May 13, 2018 16:42

Quote

HI lorentedford
Thanks for the reply and the suggestion!
I7 8700k with quicksync enable should get VERY fast rendering time (for x264 format, and even HVEC - i think 8700k GPU suport it) on powerdirector...

HI lorentedford
Thanks for the reply and the suggestion!
I7 8700k with quicksync enable should get VERY fast rendering time (for x264 format, and even HVEC - i think 8700k GPU suport it) on powerdirector...
On my particular system (i7 4790k, water cooled, 32gb ddr2 ram at 1800mhz and AMD R290x, during edition, the processor usage stays between 60 to 70%. Can you edit 4k footage without any throttle on the timeline with 8700k?

Its ok but the 8700k isn't like gold when it comes to rendering.. Although I strive for the best etc..

IgorTFerreira

Newbie Location: Rio de Janeiro - Brazil Joined: Jul 06, 2016 22:59 Messages: 11 Offline

May 13, 2018 17:32

Quote

Its ok but the 8700k isn't like gold when it comes to rendering.. Although I strive for the best etc..

Thanks again for the reply...
I read that quicksynk from 8700k is better for rendering than Hardware Acceleration of 1070 in powerdirector... use it can get you extra speed for rendering...
(you just need to plug in a HDMI cable at the motherboard output and connect it to any monitor/tv input (and enable on windows desktop manager)... I can use quicksync + r290x and i get awesome rendering speeds... my only complain is with timeline lag on some projects...