Intel 13 Gen or Ryzen 7 Gen for New Rig

trying to upgrade from current rig as below:
- Ryzen 5950x
- 128GB DDR4
- B550 motherboard
- RX6800 XT
- 850W(PSU)

to newer CPU mainly because of the single thread performance boost. But not sure what the considerations are here?
- future CPU upgrade (I mainly just need performance from CPU for virtual machines, not much gaming although I have RX6800XT and end up for fall guys…)
- reusing current parts

Main usage:
- running multi work virtual machines
- each VM runs a lot of applications for software development work (CPU and RAM intensive)

To clear the single thread performance vs multi VM confusion. The rig currently running 3 VMs, 2 are reasonably fast but 1 is quite slow. It is with enough cores allocation. On host side, 5950x usage is only around 30-40%. When peak is reached, I can see some logical cores are fully loaded while most are lightly loaded. Hence to me, only thing I can do is to get a new CPU to boost single thread performance and potentially buy CPU with less cores, as it seems I don’t need so many. I was using a 11800h intel laptop initially, but it just cannot handle the load hence I jumped straight to this rig. Not like some unreasonable comments here saying to burn money.

Budget can be put apart for now. It is for work to generate more revenue, so not first priority.

Thanks in advance.

Comments

  • Using this PC for gaming or productivity work (video editing)? Different recommendations based on your usage. A big overkill for VM? Not sure…

    Also assume you want to upgrade to the specs above, not from…? If actually from, what exactly is the problem you are having in VM apart from the single thread performance?

    • These are not normal hobit VMs, they are for software development work. Each VM is running a lot of applications. It is getting too slow for 1 virtual machines (RAM is fine, CPU is slow), other VMs are ok. Hence upgrading. Updated the post with main usage. Thanks for your time.

      • Perhaps worth first instrumenting that VM with perf or Intel VTune to see what the bottleneck is?

        If Linux, you could disable CPU vulnerability mitigations.

        You could also dedicate the fastest CCD and fastest cores to that VM and ensure other VMs and OS can't use those cores.

        • Thanks for your suggestions. All host and guests are win10. Is there any better way to find out the bottleneck? I have been using just the task managers on host and guest.
          When I stress CPU in all guests not host, I can see from host that the loads are spread nicely across 32 cores. So I think they are not completing with each other. There are still cores only lightly loaded with host workload.

          • @fan8956: You can perform an analysis at a system level and application level with: https://developer.amd.com/amd-uprof/#userguide

            Cores (and hyperthreads) compete with each other over shared resources at the core level and L3 (per CCD). Pinning the vCPUs to specific cores on a CCD may help depending on what the bottleneck is.

  • +2

    What a contradiction:
    Reason: "to newer CPU mainly because of the single thread performance boost"
    Usage: " running multi work virtual machines"

    Single thread speed is where you are not running a number of separate threads, like running multiple VM's or even one VM (One VM = two OS's as one is the host and one is the guest). VM != Container

    • If I tell you
      - the current CPU has mostly only around 30% load (shown from host)
      - 2 VMs are fast, but that last VM is quite slow (mostly due to the applications running)

      what would you do other than asking for higher single thread performance?

      • +2

        I don't think a VM is limited to a single thread. How many cores do you allocate to the slow VM, and how many CPU intensive processes are running there?

        • Tried 4, 8, 16 cores and no difference.

          • @fan8956: … no difference? thats where most of the performance comes from.

            How much ram have you allocated to each vm and how much do u have now?
            Since u said u run a lot of applications at the same time, how much storage have u allocated to each vm?
            How about video memory?

          • @fan8956: It could be that your application is a single threaded process and cannot make use of the extra cores.
            If that is the case, you'll need to find out if it's possible to divide you workload into multiple chunks and run them in parallel with multiple processes.
            Is your VM Windows or Linux? If Linux, you can use 'top' to check the CPU usage distribution among the cores. On Windows, Task Manager can give you the same info.

      • Think of a VM as a PC and therefore you need to use the same techniques you use to optimize a PC on the VM that is having performance issues.

        A VM you can limit to a single CPU core, but that's like running the OS on a single core CPU….. ouch.

        On the flip side if you allocate say 8 of the 16 cores to a single VM's and the main app in the VM only uses a max of 4 threads (not cores) then you will have overallocated the VM with too many cores.

        You need to optimize the app running on the VM that is slow, which could be due to I/O or GPU or RAM or something else.

        Be aware that the current Intel CPU's have two different types of cores and only the performance cores seem to be good for dev work.

        • You just answered your own argument of contradiction. VM is just like a PC. Loads of people here know better single thread performance means better for gaming, same for most productivity workload that cannot fully utilize a lot of cores.
          After I have verified that more cores with enough RAM, disk didn't help (the loaded cores reached nearly 100% while other cores were lightly loaded), I can assume better single thread performance will help here.

          • @fan8956: Good luck as you will need it in your single core expensive voyage.

            • @AndyC1: again depends on how you look at it. I see it more expensive with revenue lost due to the productivity lost, especially over a long time

  • +1

    Have fun with those Intel E-cores when playing with VMs:

    https://communities.vmware.com/t5/VMware-Workstation-Pro/Poo…

    • That is a very useful input, thanks a lot

      • Does this mean you'll go with the Ryzen 7?

        • Most likely rudeness 7 or 9

  • +1

    If you want to keep the DDR4 RAM, only option is Intel cos AMD's AM5 platform is DDR5 only. But since you are changing platforms regardless, AM5 would probs be better suited long-term, especially since you don't have to worry about P-cores vs E-cores from the 13900K for virtualisation. If you grab the 7950X and DDR5 + motherboard now, you have the option to upgrade down the line if needed. That's all assuming you aren't limited in budget, which seems unlikely from your post.

    • What gave it away? the 128gb ram? xD

  • Go for the more expensive option since you clearly just want to burn money anyway. Why would you care about single thread performance if you’re running multiple VM’s?

Login or Join to leave a comment