Benchmark Xeon E3-1245 v5 vs Xeon D-1557 on Plex video transcoding: Xeon D won't hit 100% CPU load

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Rand0mUser

New Member
Dec 14, 2019
10
4
3
Hello,

I'm trying to compare performances of 2 servers using very similar setups:

Server 1: Xeon E3-1245 v5, Asrock Rack C236WSI motherboard, 2 X 16 Go DDR4 2133 RAM, data pool with 4 X 3 To HDD in raidz, Freenas 11.2-U7

freenas-1.JPG

Server 2: Supermicro X10SDV-12C-TLN4F (Xeon D-1557 SoC), 2 X 32 Go DDR4 2133 RAM, data pool with 2 X 500 Go HDD in stripe, Freenas 11.2-U7

freenas-2.JPG

On both systems I have Plex installed as a jail (standard Freenas Plex plugin, same version).

I'm running a transcoding task on a 4K video, video is the same on both servers, and Plex transcoding settings are the same on both servers: make my CPU hurt, and transcode very slow (best quality).

Results are very close to each other, here are screenshots a few minutes after launching transcoding task, at the same time:

Server 1

xeon.JPG

Server 2

xeon-d.JPG

A few minutes later, some other screnshots, still at the same time:

Server 1

xeon-2.JPG

Server 2

xeon-d-2.JPG

And so on until transcoding is finished... I tried a few different files, and results are always similar. Server 1 is always around 5% to 10% faster than server 2 for the same task.

When looking at CPU usage on Freenas webUI, there is something different between the servers:

Server 1

cpu-usage-xeon-e3.JPG

Server 2

cpu-usage-xeon-d.JPG

On the Xeon E3-1245 v5, CPU usage is always around 100% while on the Xeon D-1557 it will vary between 50% and 75%.

I was wondering if this is a bug because Freenas won't report CPU usage the same way for the 4 cores 8 threads CPU and for the 12 cores 24 threads CPU, or if it may be a bottleneck because of some power / efficiency related settings in the BIOS?

From results I found on Geekbench, Xeon D-1557 should perform slightly better than Xeon E3-1245 v5 on multi cores tasks: Xeon D-1557 vs Xeon E3-1245 v5 while I can see just the opposite on my tests.

Load average is around 13 13 11 on Server 1:

htop-xeon.JPG

And around 12 12 11 on Server 2:

htop-xeon-d.JPG

Which seems a bit strange since the Xeon D-1557 has much more cores and threads.

BIOS settings on the Xeon D-1557 server are stock, here are CPU settings:

iKVM_capture1.jpg

iKVM_capture2.jpg

iKVM_capture3.jpg

Is there any way I can get this Xeon D-1557 hit 100% CPU usage on transcoding tasks, or is this normal behavior? What do you think?
 

i386

Well-Known Member
Mar 18, 2016
4,242
1,546
113
34
Germany
I remember a problem with x264 Not scaling Well beyond 20 or so threads.
Not Sure If IT IS the Same Problem here
 

Rand0mUser

New Member
Dec 14, 2019
10
4
3
Hello,

Thanks for this, I think you're right. I just tried a stress test on the Xeon D-1557 running Prime95 and CPU usage reached above 100%:

xeon-d-stress-test.JPG

Also, htop shows all 24 threads at 100% usage, while load average is around 25 which seems legit according to the number of threads for this CPU:

htop-prime95.JPG

System and CPU are working fine, so lowest CPU usage when using Plex transcoding is more likely transcoder related. I guess I'll have to play with Plex settings and tweaks if I want to improve that.

At the same time, while making stress test I also measured power draw from the X10SDV-12C-TLN4F server. Total system power draw whith all 24 threads at 100% on Prime95 stress test #2 and 3x SATA HDDs running was 87W only measured from the wall, which is really low power for such a system, especially compared to the Xeon E3-1245 v5.