Geekbench on TrueNAS through Docker weird results

Communities
1

After following @wendell tutorial on getting VM for docker working on TrueNAS scale, I did some GeekBench results to see what effect running under VM vs native would be.

Geekbench Compose

version: "2.1"
services:
  geekbench:
    image: davidsarkany/geekbench:latest
    container_name: geekbench
    environment:
      - TZ=Asia/Bangkok

Now none of this is scientific, and yes, I could have screwed stuff up, but meh.

This is the results of it running on the VM: QEMU Standard PC (i440FX + PIIX, 1996) - Geekbench Browser
1 Processor, 16 cores
Single Core: 503
MultiCore: 5416

I thought ok… should be a lot faster on TrueNAS native docker (this is while still running Kubernetes and using the dopcker-compose app to run portainer).

2 Processor, 16cores, 32 threads
Single Core: 542
MultiCore: 1232

This made zero sense to me and kind of explained why the VM dockers seemed to perform way better than TrueNAS apps, also gave me vindication for my very real dislike of the system they use for docker.

Since then, I have completely disabled Kubernetes in TrueNAS and am just using Pure Docker. I found some useful scripts to do this. Everything felt snappier tbh, and apps didn’t go “missing” after reboots or “hang” for no apparent reason. Been happy so far and system seems a lot more solid. On a whim I decided to run the geekbench again. I have even more dockers running in the background as well, which might affect things.

2 Processor, 16cores, 32 threads
Single Core: 610
MultiCore: 9263

What the actual hell. Is the kubernetes implementation that TrueNAS use a pile of poop, or does it just artificially limit stuff for the fun of it? This is just weird.

1 Like