CyberLink Community Forum
where the experts meet
| Advanced Search >
Speed Test after PC upgrade
Reply to this topic
tomasc [Avatar]
Senior Contributor Private Message Joined: Aug 25, 2011 12:33 Messages: 5532 Offline
[Post New]
It does seem that optodata smoked the tests with his modern pc. He couldn’t use PD14 with the older Nvidia drivers for that RTX 2070. Out of curiosity I will be running the test too.
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 6212 Offline
[Post New]
Quote
The old formats clearly stated for the old tests:
Output format:
Standard 2D
H.265 HEVC
MKV
HEVC 4K, 4096x2160/30p, (37 Mbps)

My mistake in using MP4 instead of MKV for the 4K HEVC test. The table below shows the HEVC 4K 4096 x2160/30p (37Mbps) MKV results:

Profile Encoding Time File size
H.265 MKV NVENC 1:16 421MB
CPU 2:54 427MB

With the Matrovska container, NVENC takes more than 2x as long to produce than MP4 and is also clearly slower than OP's system. MKV HEVC Video Encode usage was at 100% with CPU @ 24%, while with an MP4 container NVENC was 96% and CPU was at 4%.

With PD14, my CPU-produced times were identical to PD365.

This message was edited 2 times. Last update was at Jan 13. 2021 23:20

Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4692 Offline
[Post New]
Quote It does seem that optodata smoked the tests with his modern pc. He couldn’t use PD14 with the older Nvidia drivers for that RTX 2070. Out of curiosity I will be running the test too.

"smoked", not exactly, for H.264 producing I've been noticing that PD has kind of plateaued on modern high end pc's for CPU encoding. Probably just internal PD code inefficiencies, especially when PD14 performance is the same as PD19, i.e. stagnant CPU encoding development 5+yrs.

My 7 yr old colossal pc with RTX2070
H.264 test, 1:48 CPU, 0:27 NVENC
vs
optodata, 1:52 CPU, 0:23 NVENC

Basically the same results modern pc vs colossal, no significant H.264 CPU encoding benefit.

Jeff
Reply
MojoBari [Avatar]
Newbie Private Message Joined: Jun 08, 2020 18:28 Messages: 11 Offline
[Post New]
I reran my tests. I did not use MKV either but it did not change my results much.

All were 4K 4096x2160/30p, PD 19

Profile Encoding Time File [MB]
H.264 MP4 CPU only 1:20 594
H.264 MP4 Fast Vid Rend Tech 0:52 598
H.265 MKV CPU only 2:19 427
H.265 MKV Fast Vid Rend Tech 0:52 432


AMD Ryzen 9 3950X 3.5GHz 16-Core Processor
AMD Radeon RX 5600XT - XFX Thicc II Pro
32GB RAM
Sabrent Rocket 1TB NVMe Gen4 M.2 Solid State Drive

This message was edited 1 time. Last update was at Jan 14. 2021 11:17

Reply
[Post New]
Quote I just upgraded my PC. So far I am a little disappointed at the performance increase. I guess I was expecting it to be like 10X faster instead of 2-3 times faster. I was hoping that in Edit mode in PD that I would be able to play the project I was working on without having to render it. It is better though. On my old system my test project would play for 13 secs before the video would stop/stutter while the audio kept going. On the new system it gets up to 1:09 before it stops. It does not crash, it just stutters. Here is my data. The last two lines are render speed times.

All the source video was shot by 16 musicians in their homes using smart phones. So they vary in quality but none of them are higher res than 1080p.


I am more amazed you actually got your hands on that CPU and GPU!!!!! The are in such short supply!

Hoping to get a similar set up.

Thanks for sharing this! Looking forward to the follow up!
Reply
[Post New]
Quote

"smoked", not exactly, for H.264 producing I've been noticing that PD has kind of plateaued on modern high end pc's for CPU encoding. Probably just internal PD code inefficiencies, especially when PD14 performance is the same as PD19, i.e. stagnant CPU encoding development 5+yrs.

My 7 yr old colossal pc with RTX2070
H.264 test, 1:48 CPU, 0:27 NVENC
vs
optodata, 1:52 CPU, 0:23 NVENC

Basically the same results modern pc vs colossal, no significant H.264 CPU encoding benefit.

Jeff

My experience too, my GTX1080 gets almost identical results with yours. For fun, I have looked on the CPU and GPU usage and the GTX1080 encoding was used at 50%, CPU usage was 18% (I have 2x8=16 real cores plus 16 virtual on my dual Xeon E5-2667 v2).
The limitations are somwhere else.
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4692 Offline
[Post New]
Quote My experience too, my GTX1080 gets almost identical results with yours. For fun, I have looked on the CPU and GPU usage and the GTX1080 encoding was used at 50%, CPU usage was 18% (I have 2x8=16 real cores plus 16 virtual on my dual Xeon E5-2667 v2).
The limitations are somwhere else.

Yes, that similar performance posted several times prior, like here for another profile, https://forum.cyberlink.com/forum/posts/list/0/81097.page#333777

Keep in mind, PD never updated to the recent SDK so one does not get the speed/quality bump for Turing GPUs, HEVC B-frame support,...so on. So yes, no RTX2xxxx gain over GTX1xxx. RTX3xxxx same encoder as RTX2xxxx so nothing in the pipe either. Would be nice if CL finally did a PD NVENC true overhaul and supported other improvements as well.

Jeff
Reply
tomasc [Avatar]
Senior Contributor Private Message Joined: Aug 25, 2011 12:33 Messages: 5532 Offline
[Post New]
My test on PD17 with the docs to fill in that spreadsheet are as follow:
H265 hve mkv 1:15 431 MB
H.265 cpu mkv 4:51 437 MB
H.264 hve mp4 0:24 579 MB
H.264 cpu mp4 1:32 609 MB

GPU or CPU loading can be 100% as shown in the screenshots.
[Thumb - Perf comp hve 4k h.265 mkv.jpg]
 Filename
Perf comp hve 4k h.265 mkv.jpg
[Disk]
 Description
h.265 hve gpu load
 Filesize
418 Kbytes
 Downloaded:
5 time(s)
[Thumb - Perf comp cpu 4k h.264 mp4.jpg]
 Filename
Perf comp cpu 4k h.264 mp4.jpg
[Disk]
 Description
h.264 cpu load
 Filesize
417 Kbytes
 Downloaded:
4 time(s)
Reply
[Post New]
Quote My test on PD17 with the docs to fill in that spreadsheet are as follow:
H265 hve mkv 1:15 431 MB
H.265 cpu mkv 4:51 437 MB
H.264 hve mp4 0:24 579 MB
H.264 cpu mp4 1:32 609 MB

GPU or CPU loading can be 100% as shown in the screenshots.


Please have patience and help out a noob.

The H265 codec used up 100% gpu (1 core)
The H264 codec used up 100% cpu (1 core)

Is that correct?

I am assuming PD is not optimized for hardware set ups? More generic.

Other assumption:

CPU will likely be used more for encoding, gpu would be use for displaying in real time?

Thanks,

VeepO
Reply
tomasc [Avatar]
Senior Contributor Private Message Joined: Aug 25, 2011 12:33 Messages: 5532 Offline
[Post New]
Quote Other assumption:

CPU will likely be used more for encoding, gpu would be use for displaying in real time?

You could be right on a cpu that has 1 core and any gpu as you have stated. Most cpu today are multicore. If all the cores are physical then the 100% seen is possible. If half the cores were logical then 50% is more likely seen in tests.

Look at the tests in that spreadsheet linked earlier to get an understanding of the results seen.
Reply
Reply to this topic
Powered by JForum 2.1.8 © JForum Team