AMD Threadripper 12 and 16 cores

jmc

Active member
Finally some results out.

Handbrake x264 seems to only use 43% of a 12/24 Threadripper.

X264.jpg

From...
http://www.guru3d.com/articles_pages/amd_ryzen_threadripper_1950x_review,16.html

Looks to me like it is only really using the true 12 cores and not the Hyperthreads.
Would it use 16 true cores on a 1950x? I hope so.

This website also shows 1950x 16core at 51fps and 1800x 8core at 47fps. 8% faster.
That seems very iffy.
-------------------

https://www.pcper.com/reviews/Processors/AMD-Ryzen-Threadripper-1950X-and-1920X-Review/Media-Encoding-and-Rendering

This website shows 1950x 16core at 53fps and 1800x 8core at 33fps.
Around 60% faster on a 16 core Threadripper vs 8 core Ryzen.

Not the "double" I would wish for but seems much more reasonable.

My 3930 6 core intel basic dvd vrd x264 encode did 301 fps with a 42%
increase on an overclocked 1700 Ryzen @3900 for 428 fps.
(if) With another 60% boost from a 1950x 16/32 Threadripper...maybe 684 fps.

So if I double my fps from 300 to 600+ I'll be quite happy with that!

----------------------------
EDIT... for my non testing (keep the video) encoding I use the VRD Level 3 Preset
----------and only get 120-130 fps on a dvd level mpg. (5 year old Intel 3930 6 core)
----------------------------

Plus I won't have to worry what else I want to use my computer for while encoding
and not slow everything down.

Won't know till sometime in September when I hope to put together my new
main system 1950x 16/32 Threadripper.

The Noctua Threadripper (air) coolers won't be out till end of Aug.
Do not trust liquid cooling at all, have read they can dry out over time.

I have no idea where cpus are going from here.
I imagine I'll use my next system till it dies.

jmc
 
Last edited:

Dan203

Senior Developer
Staff member
Even the x264 encoder maxes out at 8 threads I believe, so I'm not sure how one of these would really help unless you're doing multiple encodes simultaneously.
 

jmc

Active member
Even the x264 encoder maxes out at 8 threads I believe, so I'm not sure how one of these would really help unless you're doing multiple encodes simultaneously.
Threads or true cores? I've found this to be confusing.

Are you perhaps thinking of H264? I find that very limited... below is real confusion with H264 testing.

X264...With my 6core/12thread cpu...
I see all 12 threads in taskmanager stay at 95+% with a single X264 encode
**on the second pass** -30Mbps HD file.

H264... Now that is very limited! On my 8/16thread Ryzen...
This will only use the 8 true cores and no other threads.
Only the first and every other column will show use in taskmanager.
At 25%...http://www.videoredo.net/msgBoard/attachment.php?attachmentid=2246&d=1497293656

BUT...
"Otter" in my FYI Ryzen thread has a image that shows all 16 columns being used with H264 (?)
at 76%...http://www.videoredo.net/msgBoard/attachment.php?attachmentid=2243&d=1497139864


http://www.videoredo.net/msgBoard/showthread.php?36244-FYI-8core-Ryzen-results-H264-vs-X264&p=125497#post125497


So confusing...Thoughts welcome!
Any tests you would like me to run to help sort this out?
--------------------------

I do want to be able to stack encodes if needed and still play games, stream movies,capture etc.
without being concerned...am I slowing everything down.
Don't want to have to worry about it anymore.

I'll find out next month when I get my 16 core Threadripper build done.
I'll be sweating when I flip that power switch for the first time!

Thanks,
jmc
 

Dan203

Senior Developer
Staff member
VideoReDo spawns threads for other things when encoding, like decoding, resizing, deinterlacing, etc... but the MC encoder is definitely capped at 8 threads. The x264 encoder may not be. I can't remember exactly on that one.
 

jmc

Active member
VideoReDo spawns threads for other things when encoding, like decoding, resizing, deinterlacing, etc... but the MC encoder is definitely capped at 8 threads. The x264 encoder may not be. I can't remember exactly on that one.
That is probably the answer to our taskmanager differences.

If I am just testing I always use the plain default VRD profile otherwise no one
can test and compare their results.

Thanks,
jmc
 
Top Bottom