I looked at desktop systems for some long while. None really do the needful very cheaply. I wanted to do PCI passthru at least and run a few more substantial vms. This meets the requirements and had pretty much everything

Looks like it was never powered on, not a spec of dust inside. Swapped out the supplied windoze disk with another SSD and so far, XCP likes it very much. I am very impressed with what I obtained for ~$1500 with tax.

@Andrew thanks, and I see the tx401 has a chipset AQC107, so I'll go with that if you think it's going to work out of the box, I'll probably put this project on hold for a moment, but as soon as I have a confirmation I'll update this post to let everyone know how it went, might take a few weeks but I will.

Thanks for the help ❤

Sometimes, I had to try 3 or 4 different sticks before one works. Those little things tend to die pretty quickly.

@stormi

I'll be testing the r8125 drivers.

Able to install r8125-module-9.012.03-1.xcpng8.2.x86_64 Able to reboot host and have Dom0 networking. Speed negotiated to 1000Mb/s. I don't have any 2.5G switches so can't test that. VM (Fedora Server 36) able to get an IP on untagged network VM able to ping Internet IP (1.1.1.1) VM able to ping internet DNS (xcp-ng.org) Changed to other network (VLAN 10 - my "Guest" Network) Able to get IP from DHCP Able to ping internet IP (1.1.1.1) Able to ping internet DNS (xcp-ng.org) Able to migrate VM to host

Thanks for the tips. All those items checked out. After a bit more troubleshooting involving a nearby switch and my VLAN routing, things are working a little more consistently, so I'm going to chalk this up to a routing issue elsewhere in my network. I'm just not used to a NIC turning off the link light if there's not a valid route.

@still_at_work Personally would say AMD, almost all XCP-ng servers I manage are AMD based and have been excellent so far. Usually a better value and better overall performance and their high core counts are great for virtualization unless you specifically need something that is faster single threaded.

I have 3 hosts in my lab, all Threadripper (so baby EPYC) and they've been rock solid as well.

It doesn't matter: if your XO dies, you can re-import a fresh one and reconnect to the backup repo, and restore metadata.

https://xcp-ng.org/docs/compute.html#vgpu
This is what you need, Pass through. Not necessarily VGPU ($400 per seat).

But! (Windows)
Follow these steps:

The link above. Do not connect the GPU to the VM you have. restart while holding shift so you can start with "safe mode + Networking" Before your VM gets to BIOS force Shut down. Attach RTX video Card Start "safe mode + Networking" Download and install NVIDIA drivers. Reboot

You will have to do this to every VM that you need that RTX card on. Hasta la victoria siempre

I see, so you might need the UEFI info from the host or something similar 🤔 I don't know if this exists, I have to admit I'm not using Windows in production myself.

You are welcome 🙂

I second @olivierlambert saying that you should consider EPYC, they are fantastic to work with, have moved my entire org to almost 100% EPYC systems and they have been nothing but rock solid reliable, fast, and efficient. You get more power for the money too and more expansion options.

And both Intel and AMD run just fine, there aren't really any issues or differences that matter for most workloads.

EPYC is mature at this point though, the "teething" issues from years ago are long gone, I have some first gen EPYC stuff running, some that have had uptime in the realm of a year without a single hiccup.

@olivierlambert I made a system update (yum update) and rebooted the server without interfering on grub menu... So it persisted on the same error... I nad to choose the second option again to do a good boot. I tried to find where those faulty grub log lines where saved, but journalctl just stored healthy boot.

Everything worked out of the box with 8.3 🙂 (including the 2.5G)

Didn't test anything GPU related.

Not sure to get it, so giving the question for @fohdeesha