Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In Register

Quick Links


Categories

In this Discussion

› › Very poor geekbench for dedicated server
New on LowEndTalk? Please Register and read our Community Rules.

Very poor geekbench for dedicated server

RazzaRazza Member
edited January 2017 in Help

I was helping a friend setup a new server with Cloudlinux, when i was setting it up the cpu performance seemed bad so i run Geekbench http://browser.primatelabs.com/geekbench3/8227629 the single core is that bad ever my cheap E5 vps's can beat it.

Is the poor cpu performance due to the usage of a older kernel 2.6.32.

Ryan

Comments

  • yomeroyomero Member

    Probably your CPU is being used for something else? Take a look at your processes

    Also, check /proc/cpuinfo to check the frequency of the processors, and take a look at the cpu governor being used (probably the conservative one).

    Finally, take a look at the temperature sensors, probably the cpu is being throttled because is overheating.

  • AnthonySmithAnthonySmith Member, Patron Provider

    A few possible things come to mind.

    Your friends server is running mdadm/linux raid and the sync is in progress which will impact cpu.

    It is not a dedicated server at all it is a 'smart server' being sold as a dedicated through shitty marketing techniques.

    Something else is going on at the same time which is hurting performance.

    Some sort of CL limits kicking in preventing you from eating up the threads.

    Maybe boot in to the standard kernel and check or boot in to a live/rescue and run it again to rule out any OS operations?

  • MagicalTrainMagicalTrain Member
    edited January 2017

    2.6.32 reminds me first of openvz. :D

  • RazzaRazza Member

    yomero said: Probably your CPU is being used for something else? Take a look at your processes

    The server is not in use yet so the load is 0

    AnthonySmith said: Your friends server is running mdadm/linux raid and the sync is in progress which will impact cpu.

    The server is using Mdadm but its not syncing or anything

    AnthonySmith said: It is not a dedicated server at all it is a 'smart server' being sold as a dedicated through shitty marketing techniques.

    It's a DEDIBOX® MD 2016 from online.net so am sure it's a real dedi.

    AnthonySmith said: Some sort of CL limits kicking in preventing you from eating up the threads.

    That what i was thinking so i run it as root user as i don't think CL limit's that user.

  • AnthonySmithAnthonySmith Member, Patron Provider

    Razza said: It's a DEDIBOX® MD 2016 from online.net so am sure it's a real dedi.

    Or maybe you found the reason they can sell them so cheap, they are actually big 256GB dual 10 core/20 thread E5's running KVM with fake CPU flags etc :p

    /jk

    Thanked by 1doughmanes
  • RazzaRazza Member
    edited January 2017

    yomero said: Finally, take a look at the temperature sensors, probably the cpu is being throttled because is overheating.

    That's not the issue the cpu temp is 29.0°C.

  • FlapadarFlapadar Member

    What CPU governor are you using? If its not performance, try that.

  • williewillie Member

    I've heard a lot of complaints of this type about these servers. I'll try on my own soon.

  • RazzaRazza Member
    edited January 2017

    Flapadar said: What CPU governor are you using? If its not performance, try that.

    Cant seem to find what it's set to but i think it must be set to performance as the cpu clock speed is set to 3504.072 on all core

  • myhkenmyhken Member

    I just tested my Online E3-1230v3 server and got a result of 13199
    https://browser.geekbench.com/v4/cpu/1684565

    And I tested my own E3-1245v5 server at home and got a result of 11942
    https://browser.geekbench.com/v4/cpu/1684614

    And I tested my Hetzner E3-1245v2 server and got a result of 11728
    https://browser.geekbench.com/v4/cpu/1684631

    So clearly something wrong with your server.

    Thanked by 3ehab Razza windytime90
  • RazzaRazza Member

    Must be due to Cloudlinux or old kernel verion 2.6, run geekbench in rescue mode and got a good result https://browser.geekbench.com/v4/cpu/1684821

  • FredQcFredQc Member
    edited January 2017

    I would start by checking in the server BIOS if those things are enabled in the CPU menu

    Cores Enabled All
Hyper-treading Enabled
Intel AES-NI Enabled
Intel Virtualization Technology Enabled

Also check in CPU Power Management Configuration 

Power Technology set to Enabled
Energy/Performance set to Performance
Turbo mode set to Enabled/All
  • RazzaRazza Member

    FredQc said: I would start by checking in the server BIOS if those things are enabled in the CPU menu

    Thanks for the advice, it find the possible cause mostly the medieval 2.6 kernel as it worked fine on rescue mode that run's alot newer kernel.

  • doghouchdoghouch Member

    @yomero said:
    Probably your CPU is being used for something else? Take a look at your processes

    Also, check /proc/cpuinfo to check the frequency of the processors, and take a look at the cpu governor being used (probably the conservative one).

    Finally, take a look at the temperature sensors, probably the cpu is being throttled because is overheating.

    Basement DC

  • FredQcFredQc Member

    @Razza said:

    Thanks for the advice, it find the possible cause mostly the medieval 2.6 kernel as it worked fine on rescue mode that run's alot newer kernel.

    Glad that you found the issue. On my part, I found that 50% of time I setup a new dedicated server, many features that effect directly and negatively the CPU performance are disabled in the BIOS. I wonder why ¯_(ツ)_/¯

  • YuraYura Member

    @FredQc said:

    Conspiracy theory time. Maybe providers choose settings which would save electricity and tear&wear of the hardware?

  • hzrhzr Member

    Yura said: Conspiracy theory time. Maybe providers choose settings which would save electricity and tear&wear of the hardware?

    Realist theory time. Maybe providers are lazy and leave it default.

  • YuraYura Member

    @hzr said:
    Realist theory time.

    Worst opener ever.

    Thanked by 1WSS
  • WSSWSS Member

    @Yura said:

    @hzr said:
    Realist theory time.

    Worst opener ever.

    @Nekki theory time. The less resources available, the less heat. Less heat means less scars.

  • ClouviderClouvider Member, Patron Provider

    @hzr said:

    Yura said: Conspiracy theory time. Maybe providers choose settings which would save electricity and tear&wear of the hardware?

    Realist theory time. Maybe providers are lazy and leave it default.

    Default, you mean the config marked by the motherboard manufacture as optimal and recommended as a baseline to use?

    Yeah, most do. That's because this config is supposed to work for normal use, and if you need custom, well, you customise it.

    Thanked by 1Junkless
  • FredQcFredQc Member

    Clouvider said: Default, you mean the config marked by the motherboard manufacture as optimal and recommended as a baseline to use?

    Like leaving Turbo Boost and AES-NI to Disabled and CPU to "Power saving mode" ...

    Yeah.

  • windytime90windytime90 Member
    edited January 2017

    @myhken said:
    I just tested my Online E3-1230v3 server and got a result of 13199
    https://browser.geekbench.com/v4/cpu/1684565

    And I tested my own E3-1245v5 server at home and got a result of 11942
    https://browser.geekbench.com/v4/cpu/1684614

    And I tested my Hetzner E3-1245v2 server and got a result of 11728
    https://browser.geekbench.com/v4/cpu/1684631

    So clearly something wrong with your server.

    So the E3-1230v3 is generally stronger than E3-1245? I think there must be something wrong here.

  • myhkenmyhken Member

    windytime90 said: So the E3-1230v3 is generally stronger than E3-1245? I think there must be something wrong here.

    I did see the same. No clue why, or my online server do not do anything. The two other servers I tested run several VMs each, but the CPU usage when tested was almost at 0. So maybe something there?

Sign In or Register to comment.