I'd run Windows as a VM and then simply share your DrivePool to a Plex LXC as cifs. My plex lxc is entirely independent of where I store my media files.
Minisforum MS-01 PCIe Card and RAM Compatibility Thread
- Thread starter Patrick
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Odd question If you had to choose…..64Gb of DDR5-5600 or 96Gb of DDR5-4800 (if same price)? I9 13900.
Possible uses either as Windows workstation or as Proxmox +Truenas Scale and light VM’s?
I’m new to DDR5 and lot’s of opinions seem to say faster is marginal difference but more memory could help my use case?
Thanks!
Possible uses either as Windows workstation or as Proxmox +Truenas Scale and light VM’s?
I’m new to DDR5 and lot’s of opinions seem to say faster is marginal difference but more memory could help my use case?
Thanks!
There must be no problems especially if you disable in the bios everything that has the work "ASPM".
I dont use CEPH but i do replication only to another 3nd party brand machine with a 3nd machine as Quorum. No problems at all what so ever with HA and everything.
You must forget using imo the i226-LM because it does crazy things and keep it only for remote management and finally for the heat, dont expect to put 3 x NVMEs or 2 x NVMEs and one big U2 like i did and expect to be fine as matter the temps. If the machines are pushed and the NVMSs do work alot, expect temperatures above 70c. For this reason in my MS-01 i have extra cooling at the bottom to push air to the NVMEs and everything is fine after.
Its a solid machine but needs some care if you over work it as matter temps.
Start with 1. It has bad thermal management and it is not stable when all power save options are enabled. Disabling them increases heat..
Fans are rather noisy (compared to NUCs). But you get a lot of features in a small package. I would prefer a slightly larger size with better and less noisy cooling though. And a more efficient power supply.
I ordered 3x 12900H kits with 2x 16GB CRUCIAL CT16G56C46S5. I dont need 96GB for my setup but may upgrade later.
its the same ram just smaller size, I am getting my Ms01 tomorrow, I have ordered 2-48GB DDR5-5600 SODIMM 1.1V CL46 CT2K48G56C46S5 directly from Crucial, finger crossed
That's exactly what I tried to begin with. bond0: 2x2.5g ethernets, bond1: 2x10g ethernets. vmbr0 with bond0 and vmbr1 with bond1. Using LACP bonding for both, configuration on the physical switch too. The 10g seems to be working fine in proxmox and ceph. Noticed some real weirdness with the LACP bond for bond0. I really think it has something to do with the Intel AMT/vPro management on one of the 2.5g ethernets. Rather than diving deep into the rabbit hole of figuring it out, I just got rid of bond0 all together. Now vmbr0 in proxmox is just slaved to one of the 2.5g ethernets.
I have 5 of 13900H MS-01. I was real nervous after reading this entire thread and seeing all the weird stuff people were running into, but I haven't seen any of that weirdness at all in my setup so far. I'm using desktop grade NVME SSD for my Ceph cluster, since finding decent enterprise grade NVME was a real challenge.
Each node has 96GB Crucial RAM, 3x 2TB NVME. Ceph is running on the 10g ethernets and VM/mgmt running on the 2.5g ethernets. I'm about halfway through migrating less than 200 VMs over to this cluster (most stay powered off, used for testing).
Based on my experience so far, I can recommend the setup you're considering. When I went hunting for alternatives due to the extended wait for delivery, I really couldn't find anything as small, quiet, power efficient and network capable as the MS-01.
What weirdness are you seeing with the AMT vPro 2.5G port? Mine been LACP bonded for couple weeks now, I'm no seeing anything unusual. What switches and router are you using?
You got me paranoid. So I went to test my setup. Pulled network cable between each of two alternating 2.5G ports while ran ping test. Ping never got interrupted as long as at least one cable is still active. If both cables are pulled of course ping will stop. I did same test with the 10G ports, same result. Looks like my LACP bonds at solid so far.
Last edited:
Now you have me thinking I screwed something up when I created the bond and configured the switch. My AMT setup is dedicated IP, static IP address. The switches I'm using are cheap Sodola switches I got from Amazon. When I created the LAG group on the switch, I used MAC+IP hashing, when I created the bond in Proxmox, also used Layer2+3. Soon as I enabled the LAG group, the system became unreachable and the switch showed the LAG group was down. I didn't spend too much time trying to troubleshoot it, since I'd talked myself into not having a bond for the 2.5g is better for me. Maybe someday I'll visit it again.
What U.2 drives are people using? There are not a lot of sellers on eBay that I saw for the Samsung 3.84TB PM983a drive, and the one I did order from canceled because the drive was dead.
I'd be fine with a different make of drive, but would like to stay <300 USD.
I'd be fine with a different make of drive, but would like to stay <300 USD.
I am using the pm938a in 3 nodes running ceph, partly because I made a mistake and ordered the wrong thing, but it's working fine
I would avoid U.2 unless you need the larger capacity. They run hot and require more power draw. I opted for Samsung PM983a 1.88TB NVMe 22110 drives. Got them brand new on eBay for $75 each. Bought 6 (two for each nodes). They're running nice and cool. Both staying at ~52C with just onboard fan cooling them. I haven't gotten any overheating or thermal issues. All 3 of my nodes are inside 19" rack not the best airflows but they're fine.
To confirm are you on ZFS? I too have had no luck using these enterprise drives as boot. They, however, have problem with EXT.
Following up here, I added the MS-01 to my proxmox cluster and have a couple of light VMs running some docker containers (*arr stack and a small bit of HA stuff) and my media LXC (nothing streaming right now). Everything more or less idle, although it's replicating a decent number of VMs back and forth every 15 mins. Power consumption is bouncing around, but is averaging between 18W-21W when not doing anything particularly demanding. A couple watts more if something is downloading in the background. Now the real test will be once I migrate my pfsense VM over, as that is a more consistently constant load and will use both ports of the x710 NIC. But ~20W for this is right in line with my Lenovo Tiny and this has a ton more headroom, not to mention the additional capacity for expansion. Pretty pleased so far. Now, I just need to see if it's stable enough to act as the primary host for my firewall VM.
EDIT: I do still need to add two more NVME, so that will bump the power up by a watt or two.
Last edited: