Best Value PCIe 4.0 NVMe M.2 SSD for Home Server

I built a home server in January, and the SSD I installed is already halfway through it's TBW limit (currently using a MX500 500GB with 180TBW to host proxmox + LXC + VM's).

I'm getting fairly regular bottlenecks on the SSD when conducting large system operations, so I figured I probably need to replace the SSD (unless high RAM utilisation could be the bottleneck?).

I'd like to get either 500GB or 1TB PCIE 4.0 NVME drive with a balance of high IOPS & TBW with relatively low cost ($100-$150).

Lexar NM790 and Samsung 980 Pro both seem like popular options, but I'm open to suggestions?

Comments

  • You want a NAS specific SSD - otherwise you are basically setting the cheaper consumer SSDs on fire.

    Look at WD SN700 and similar.

    • Just to clarify, this SSD isn't being used for NAS purposes in any way. It's only hosting Proxmox + LXC + VM files. All NAS data and cache is on separate drives.

      Or are you suggesting a NAS SSD is a good idea for this purpose as well?

      • +2

        MX500 500GB $55 for 180TBW = 3.6 TBW/$

        SN700 500GB $97 for 1000TBW = 10.3 TBW/$

        Your usage does involve frequent drive use so while its not NAS you are going to want something above consumer grade.

        • +1

          Yep, with hindsight the MX500 wasn't a great choice. Will definitely be including TBW and MTBF into the purchase decision. It looks like the Samsung 980 Pro has the same TBW but double the IOPS so might prioritise that over the WD.

  • I bought a Teamgroup MP34 for my NAS a few months ago and im pretty happy with it. Its probably not as good as the 980 Pro but performs well for its price.

  • do you have enough ram ? do your vm's have swap enabled ? what do you mean by large system operations ?

    • RAM is only 32GB, so that is another area for potential upgrades. TrueNAS has a dedicated NVME cache + 16GB RAM, however I haven't checked how swap works for TrueNAS so it could be relevant. The only other always on VM is Home Assistant, which seems to only use 1-2GB on average.

      I am fairly certain it isn't the RAM though, as the system bottlenecks with 1-2GB of RAM still available and unused. I've confirmed the IO is at 100% using iostat and iotop.

      I'm not sure exactly what operation's are causing the problem, but doing something as simple as pulling and recreating 5-10 docker containers is enough to push it over the edge. It actually seems to be a disk-read issue, with the system hitting 400MB/s of reads consistently while everything's bottlenecked.

      • im not a docker expert but they are basically vms. if truenas is already using 50% ram, then you might have only 6gb for your docker stuff before proxmox/lxc starts swapping …

Login or Join to leave a comment