NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.

Forum Discussion

Sandshark's avatar
Sandshark
Sensei - Experienced User
Feb 27, 2023
Solved

RN3220 CPU upgrade -- Negligible effect??

I upgraded the processor of an RN3220 from an i3-3220 3.30GHz, which is dual-core 4-thread CPU, to a  Xeon E3-1225 V2 @ 3.20GHz, a quad-core 4-thread CPU as comes in an RN4220 and sysbench shows almost no gain.  Even more confusing is that the Xeon X3450 @ 2.67GHz in my RD5200 and RN4200V2 units gets a much better score.  Yes, it's a quad-core 8-thread CPU, but it's also a lot older and has a slower clock and memory.

 

After installing sysbench, I used this command:  sysbench --num-threads=8 --test=cpu --cpu-max-prime=100000 run.  This was with all the normal ReadyNAS OS processes also running.

 

And I got these results:

 

With the i3-3220 3.30GHz:

  • Test execution summary:
    total time: 67.1723s
    total number of events: 10000
    total time taken by event execution: 537.2301
    per-request statistics:
    min: 26.59ms
    avg: 53.72ms
    max: 101.90ms
    approx. 95 percentile: 56.86ms

 

With the  Xeon E3-1225:

  • Test execution summary:
    total time: 60.8870s
    total number of events: 10000
    total time taken by event execution: 486.8973
    per-request statistics:
    min: 24.32ms
    avg: 48.69ms
    max: 90.33ms
    approx. 95 percentile: 54.33ms

That seems to only be giving me the added speed of each core, not using two more.

 

And here are the results from one of the Xeon X3450 based units:

  • Test execution summary:
    total time: 42.6945s
    total number of events: 10000
    total time taken by event execution: 341.4370
    per-request statistics:
    min: 24.01ms
    avg: 34.14ms
    max: 83.87ms
    approx. 95 percentile: 34.20ms

I've never needed to go into the BIOS and make a change when upgrading a processor (which would be difficult to do on a 3220 w/o a display), but is that what I need to do?  Has Netgear done something to make a CPU upgrade not work correctly in the 3220 so people will buy a real 4220 instead of doing what I did?  Or is sysbench just not a good tool for this?  lscpu and cat /proc/cpuinfo both show the right CPU and architecture.

 

I have a real 4220 in storage and may pull it out to see what results I get with it, but maybe somebody can point me in the right direction and I won't have to.

 

 

 

  • OK, I'm not entirely sure what was going on.  I upgraded a second one and got way better results.  So I swapped the memory between the two and the first one now gave me the improved speed.  Just to check, I swapped the memory back, and now they both perform like this:

     

    Test execution summary:
    total time: 2.5567s
    total number of events: 10000
    total time taken by event execution: 20.4385
    per-request statistics:
    min: 1.02ms
    avg: 2.04ms
    max: 28.60ms
    approx. 95 percentile: 11.02ms

    Threads fairness:
    events (avg/stddev): 1250.0000/27.44
    execution time (avg/stddev): 2.5548/0.00

     

    So, the memory must not have been seated well, even though dmidecode showed both banks properly.  Since I swapped out the memory before the CPU, I suspect that the performance shown for the original processor is also bogus.

13 Replies

Replies have been turned off for this discussion
  • Sandshark's avatar
    Sandshark
    Sensei - Experienced User

    So I found a video card that would fit and booted to BIOS.  The NAS had already identified the CPU and, while you can limit the number of cores enabled, the default was "All".  Since I find it hard to believe Netgear would put in a more expensive CPU with such a little improvement in performance, and the specs show there should be significant improvement, I think that maybe sysbench just isn't a good tool to use here.  Then again, the Xeon does support VM and that Core i3 doesn't, so maybe that's the reason.

      • Laserbait's avatar
        Laserbait
        Luminary

        Ah, wait.  I see - this is a CPU perf question, not a disk IO question, sorry for misunderstanding.

         

        Did you upgrade the memory to 8GB to match the RN4220?  I have not been able to see a pic of the motherboard, but if the 3220 has only 1 4GB DIMM, and the 4220 has 2 4GB DIMMs, it could be using dual channel mode which greatly improves memory access throughput.

    • Rollerball2021's avatar
      Rollerball2021
      Aspirant

      What Video card did you use? , were all drive bays populated? I have one that seems to have been taken out by a power surge, any suggestions would help, Thanks

      • Sandshark's avatar
        Sandshark
        Sensei - Experienced User

        I used a Zotac GT710 Zone Edition PCIE=X1 (yes X1) card which is a low profile card.  It only requires a single slot unless you want the VGA connector.  The VGA uses a separate slot but is connected via a cable.  But they are hard to come by.  It's also available in a X16 form factor, so be sure you get the right one.  And be sure you get the low-profile bracket(s).

         

        Zotac also made a fewer older ones that are similar form factor and ought to work.  Several companies made GT210 based PCIE-X1 cards.  The problem with most is getting the low-profile bracket.  But if it's just for troubleshooting, you could go without a bracket.

         

        MSI makes a similar card in a X8 format that I suspect would also work, but I've not tried it.

NETGEAR Academy

Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology! 

Join Us!

ProSupport for Business

Comprehensive support plans for maximum network uptime and business peace of mind.

 

Learn More