FYI 8core Ryzen results H264 vs X264

jmc

Active member
Finally got my Ryzen 1700 8cores/16 threads installed and everything updated. (painful)

Was interesting to see H264 ONLY using the actual cpu cores,
Just every other column was used in taskmanager.
That does not happen on my 6/12 core. Stuff jumps around in it.

Now using a standard dvd 5 minute MPG segment and the Default .MP4 profile
I got the following...

EDIT... the % is cpu useage.

W10/64 H264------------------------------------------X264-----------------Ryzen 8/16 cores
-----191 to 195 fps 30+%------------------391fps 70+% (3.1Ghz)

W7/64 204 to 209 fps mid 20%---------391fps mid 60% (3.1Ghz)
-------------------------------------------------------------------------------------

------------------------------------------------------------------------------------- 6/12 core Sandybridge (2012)
My old 6/12 core beat the 8 cores with H264 because of the very
poor use of the 8cores. But then the 6core is overclocked to 4.3 Ghz

W7/64 -----H264----------------------------X264
---------225 fps upper 20%---------301fps mid 70% (4.3 Ghz)

The overclock won with H264 but the 8/16 cores won with X264.

So, looks like do not go for cores with H264 unless you are going to pile
on the encodings. like 4 or 5 VRDs running at once.

I removed an old Q94?? quad core and replaced it with the Ryzen so it was only $700
to test this out. ( I have been dying for a new computer for at least the last two years)

Of course as I put it together I get to read of all the new price cuts happening!
(I'm not going to look)

With the results here (intel vs Ryzen) and the wonderful price drops, I'm thinking
of going for a Ryzen 12 core threadripper whenever it gets released.
It HAS to be priced less then the Intel 12 cores. (hope,hope).

Be nice to be able to pile on several encodes and not have my computer get sluggish.

EDIT... I'm not overclocking till I get a feel for whats normal with it.
Once I do I'll report back here.

jmc
 
Last edited:

Dan203

Senior Developer
Staff member
Those results are similar to what my friend reported using Handbrake. He said he got better core usage running two enocdes simultaneously.
 

Otter

Member
I would question your results. My 1800x runs both Video Redo and x264 encodes on all 16 cylinders. Most encodes run in the 60%-90% load range on all 16 threads. Running 2 VRD or x264 instances at once will get the load to 100%. Even then, browsing, Office content and even doing a VRD edit session have no perceptible lag.

I'm using Windows 7 64bit and no overclocking of CPU or memory. Tried up to 4GHz on the Ryzen, but the encode times did not improve much - CPU just ran hotter and fans louder. Just leaving it on "Standard" clocking with auto core and clocking control.

Here's a screen shot of Video Redo running a single encode and nothing else. All cores and threads being used about equally.
VideoRedo_h264 encode.JPG

And here is CPU activity running a single encode with Handbrake v1.07 64bit
Handbrake x264 encode.JPG

Ryzen 1700 should give similar results, since it is basically the same chip, just didn't pass QC at the higher clocks.
 
Last edited:

jmc

Active member
@Otter


----------------------------------------------------------------------
EDIT Monday 12JUNE17.......
Boy, wonder sometimes about my memory.
I booted W10 Ryzen and the taskmanager I remember is not there.
I must have been in W7 to see what I remember seeing!

Gotta retest and write stuff down this time!


------------------------------------------------------------------------------

I would say that our test results differences are almost certainly due to using a different source video.

EDIT------------------------------------------------------------------------------------
Except the H264 encode using ALL taskmanager columns!

Our results there are really different. My H264 test only used core 0
and only every other column after for a total of 8 only.
That taskmanager result really stuck in my mind.

I'm going to have to crank up the Ryzen and try to figure this out.

I seem to remember even the VRD Devs saying the VRD's H264 encoder
was limited in core use.
But we got the results that we got. This is a weird one!

Oh, if you could list the frames per second for your test results that would be great.
--------------------------------------------------------------------------------------------------

My test used a 5minute SD standard-dvd 3-4 Mbps mpg file.

I just did a 16.13 Mbps(per VRD) HD.m2ts file with
"force recode" default X264.mp4 profile and my 6 core cpu stayed at 99-100%.

If my main 3930 6/6HT cpu is running an encode like that there
may be problems with gaming and **especially** capturing streaming video.

With more normal multitask usage my 6 core is great.

I have not tested my 8 core Ryzen like that yet but I doubt the 8 cores will be enough.
Would be great to find out that it was tho.

The 8 core was just a test build and I am hoping that doubling my cores with
a 12 core AMD Threadripper will finally give me smooth multitasking while
encoding is going on.

As H264 only uses 8 cores I'll have 4 more free for other uses.
I can not use VRD X264 for NTSC videos with "pull down" until they fix
that deinterlace bug.
Hope it is fixed in VRD6! -Fingers crossed!

Thanks for your thoughts,
jmc
 
Last edited:

jmc

Active member
I would question your results. My 1800x runs both Video Redo and x264 encodes on all 16 cylinders. Most encodes run in the 60%-90% load range on all 16 threads. Running 2 VRD or x264 instances at once will get the load to 100%. Even then, browsing, Office content and even doing a VRD edit session have no perceptible lag.

I'm using Windows 7 64bit and no overclocking of CPU or memory. Tried up to 4GHz on the Ryzen, but the encode times did not improve much - CPU just ran hotter and fans louder. Just leaving it on "Standard" clocking with auto core and clocking control.

Here's a screen shot of Video Redo running a single encode and nothing else. All cores and threads being used about equally.
View attachment 2243

And here is CPU activity running a single encode with Handbrake v1.07 64bit
View attachment 2244

Ryzen 1700 should give similar results, since it is basically the same chip, just didn't pass QC at the higher clocks.
I did get confused hopping between W10 and W7 on my dual boot Ryzen1700 so I
redid some tests today and here is my W7 taskmanager showing only the 8 cores being
used by H264 VRD.

No idea why our W7 results are so different.
SEE ATTACHED IMAGE-RyzenH264.png.... NOT THIS PNG.GIF JUST BELOW
http://www.videoredo.net/msgBoard/images/attach/png.gif

Now with W10 I noticed with CPUZ that only 12 threads were used with H264
the other 4 threads were listed as "parked". MPG to H264.mp4, 30%, 204fps

X264 use all the threads in both OSs.

jmc
 

Attachments

Last edited:

jmc

Active member
Overclocking Ryzen 1700 ...

Ok, used the BatchTool to run 5 identical 5 minute dvd mpg segments
through VRD 746 Pro's X264.mp4 Default profile.

EDIT... used AMD's "Ryzen Master" software for the overclocking.

The Batch seems to give real "stairstep" fps. Looks like it divides by whole second
intervals so out of 5 results I may get 2@ 375fps and 3@ 391fps.
Speed------fps--------------Voltage------temperature
3200 -------375
3300 -------375
3400 -------375
3500 -------2@375, 3@391
3600 -------391 ----------------1.2v.-------- 50c
3700 -------391 ----------------1.2v.-------- 51c
3800 Crash --------------------1.2v
3800 -------409 ---+.05v------1.25v --------54c
3900 Crash --------------------1.25v
3900 --------428 ---+.05v-----1.3v----------56c
4000 Crash --------+.05v-----1.35v (my limit)

AMD says past 1.35v to AMD's limit 1.45v the chip lifespan will drop.

With "H.264" not X264 the Ryzen had to reach 3900 to equal my 3930K 6 core @4300. Both 225 fps.

With X264 the 6 core got 301 fps compared to the Ryzen 428 fps-42% Faster! Nice!
The Ryzen overclocking gained roughly14% fps (roughly due to 1second limit in fps calculation)

Bit sad about the H264 core limits. 6 years later, +2 cores and overclocked to equal my old 6 core.
Guess I had it better then I thought.

Really surprised with the low temps. I only have a Cooler Master Hyper 212 EVO heatsink.
Midrange heatsink but well rated in reviews.

Oh, yes, if you have the AMD W10 power profile it will set the cpu MINIMUM setting @ 90% active.
Was not having any of that. Reduced it and my cpu is back to 1500 instead of maybe 3500 @ idle.

Now I just have try to get my highspeed ram up to it's actual highspeed of 3200 instead of it's now 2400 speed.
And that will probably means flashing the BIOS. ick,ick,ick.

jmc
 
Last edited:
Top Bottom