Hi Team, I’m working on my 2nd server build, first one in my signature was 9 years ago with the x10sl7. It’s served me well but getting a bit aged so I’m looking at this build. To be honest, I’ve looked at many systems and am starting to second guess myself in terms of future proofing, functionality and speed/energy usage balance. Any help you can provide would be great.
Around 10 containers running things like Plex and Home Assistant via jlmkr
1 to 2 VMs (Windows or Linux)
That seems reasonably affordable for the next 9 years of my life. It would offer plenty of Sata ports, two m.2 slots, I could use the PCIe 4x slots for m.2 adapters or an sfp+ card in the future. The only downside I could find is the x16 slot does not bifurcate. The processor is a significant upgrade with idle draw likely less than today. While compute would be more than I need now, it leaves room for growth.
Your VMs (block storage) and apps would do better on a mirror than on raidz1.
And you may want a GPU for the Windows VM. Maybe go for the previous generation, and cheaper DDR4?
Understood on the mirror, thank you. Based on your GPU comment, when I dug back in I realized all of the xeon E-2400 skus are missing an iGPU. Does that complicate IPMI iKVM or does the BCM handle that video? If I went with an older supermicro board and an E-2378G with an iGPU would that mean I wouldn’t need a discrete GPU to handle the Windows VM?
fyi, all the supermicro x13 boards supporting E-2400 have broken BIOS with regard to PCIe ASPM. There are options for it in the BIOS, but all devices no matter the settings, even with pcie_aspm=force kernel command line parameter result in ASPM disabled
this is true at least for the X13SCL-iF and X13SCH-F/LN4F boards with the latest BIOS 1.4.
i am running OPNsense (FreeBSD 14.1 based) on the X13SCL-iF and TrueNAS SCALE 24.10-RC.2 on the X13SCH-LN4F and cannot get ASPM to work after many many iterations of settings, etc. Results in about a 10w+ higher idle when i compared it to similar boards both these setups replaced, X11SCL-iF and X11SCH-F.
On the contrary, E-2400 means you’d need a dGPU while a E-2300G would let your dedicate the iGPU to a Windows VM. The BMC has its own little video controller for IPMI.
Great stuff, then the E-2300 line it is. Last question @etorix I looked all up and down the manual for the x13sch-f and it never mentioned anything related to bifurcation for the x16 slot. The block diagram isn’t something I’m comfortable reading yet today but didn’t see a mention of splitting the lanes, how did you know it would do x8x4x4?
i have not seen any bifurcation options in X11 or X13 boards for the E series Xeons. i had an X12 board for a Xeon 6326 scalable and there was clear bifurcation options with multiple ways to bifurcate each slot.
I don’t know for sure, but that’s what Intel Core and Xeon E are capable of. x4x4x4x4 is not supported.
You’ll face the same limitation with Xeon E-2300 and X12STH boards, or the equivalent AsRockRack E3C256D4U(-2T/2L2T).
Ryzen CPUs (not APUs) can do x4x4x4x4, and most support ECC, so you could also go with a Gigabye MC12-LE0 or MC13-LE0, AsRockRack B550D4U or B650D4U and a suitable Ryzen or EPYC 4004 CPU.
I had to actually talk myself off the ledge of spending more time looking at the X13SAE-F. It was sexy and nice in all the right places but simply for the fact that it’s a Workstation board I had to stop myself. I’m sure it would be fine but over the years I’ve learned to heed the advice from this forum and stick with server grade hardware no matter how hot the other motherboards are. The X13SAE-F was like that girl that was so much fun going out (and taking home) but you weren’t quite sure what she’d say in front of your parents at the dinner table.
