@MattRC Thanks for your reply. I am using RTX 4060 and have tried to install new drivers as well. But no matter what I do, I am having problems. And yes I have passed both the pci ids to the vm. For Linux Guests: Remove Nvidia Drivers -> Guest is able to see the Card in lspci Install nvidia-drivers-575/570 or *-open or -server-open, then the VM cannot detect the card. For Windows Guests: I get Code 43 even after installing the latest driver. I have done the passthrough as per the documentation and @lawrencesystems video. One thing I noticed is that when I attach the PCI to the VM the Console Display in XO takes longer to load (don't know if this is related) PS: even I have the same use-case of passing the PCI SSD directly to the VM for gaming but that doesn't seem to work either. Can you confirm if you used passthrough in the Advanced->GPU Section and set it as passthough or used the 'Attach PCI' section below? or did you enable at both places ?

@andSmv Thanks for the good info! Sure if you guys provide the beta I would be more than happy to test. Unfortunately, I only have access to this hardware for another week before I need to ship it out. If you can get me the beta today or tomorrow I can get it in place and start exercising it over the weekend. Thanks!

Ahh excellent news! Thanks for keeping us posted

@dmz0001 In theory, XCP-ng 8.3 is mostly on par with XenServer 8.4, it's only a naming thing that they did as their previous naming was confusing some of their customers. Long story short, XCP-ng ~= XenServer 8.4. So what you see in the compatibility list of XenServer 8.4 is the right reference for XCP-ng 8.3. Unfortunately we do not have much feedback on Zen 5 at this time on our side.

@TeddyAstie Thanks for the input. This is what I was afraid of somehow In the meantime (a few months ago) I have done some tests with the machine. Installed opnsense as described in the opening post and directly to the device without xcp-ng just to see if there is any issue. The result is the same (i have used in both cases the same settings and same spoofed mac) the spoofing did work without xcp-ng, this is why I was hoping that something can be done via xcp. @olivierlambert said in NIC Passthrough with mac spoofing: Out of curiosity, what about SRIOV in that case? Regarding SRIOV right now its not possible for me to check if thats enabled/supported in the bios but its a good point, thanks Will take about 3-4 weeks to check.

@XCP-ng-JustGreat I only use the passthrough part of disks, not controllers. VUSB part in xo i don't use. I think that's a completely different approach. Maybe you could check your drivers and manager. There is a cleaner somewhere, don't know the link, to remove the old drivers in windows with powershell and after that install the xen drivers XenServer VM Tools for Windows 9.4.1. I got it from here: https://www.xenserver.com/downloads https://github.com/xcp-ng/win-pv-drivers/releases/download/v9.0.9065/XenClean.zip docs: https://docs.xcp-ng.org/vms/#fully-removing-xen-pv-drivers-with-xenclean

No, it's not possible to choose explicitely/easily them because they are all the same from Xen perspective. I wonder if you can "cheat" by doing vCPU pinning though

@MasterOSkillio Typically, C-states need to be changed in the BIOS. In some cases, it can be very helpful. I wrote a few blogs including this topic entitled "A Tale of Two Servers" but cannot readily find them on-line at the moment. Alas, Citrix has purged a lot of still relevant older content over the years.

Reping @dthenot

@GTwannabe I assume he means using PCI passthrough for the USB port/controller so HA has direct access to that level of the hardware to circle around some USB passthrough problems. In my case this did not helped with the ConBee, but you might try it anyway and it is no problem to reboot the host which is needed to activate the PCI passthrough Another pro for switching to an IP based controller is that you can backup and migrate the VM without any problem.

@sotero could you post the output of: # Collect xen info xl info # Collect C-States info xenpm start 3 ?

@Nevering We used to offer a “hosted Vates Stack” service, but right now we’re rethinking that and don’t have anything "official" to share just yet (though you might wanna keep an eye on our newsletter ). In the meantime, here are two easy options: Head over to our Partner Locator and use the filters to find a CSP (Cloud Service Provider) — some MSPs also offer hosted setups. Shoot us a message at partners@vates.tech or via our Contact Form, and we’ll do our best to point you in the right direction.

@Riven, What I am not sure is how Xen arrange the CPU, is it core first followed by the SMT/HyperThread Core? or is it alternating ie RealCore, HyperThread Core.

@2planks said in NIC not working on new HPE DL360 Gen11: Has this issue happened before? I don't remember such an issue on this hardware. Can you share the outputs of lspci -v, lsmod, dmesg, and dmidecode?

@efny I thought it wouldn't but in fact since I have it (almost a year) I had 0 bug. It's a Core i7-1255U.

@olivierlambert Looks like it would arrive April 14th: https://www.dell.com/fr-fr/shop/intel-flex-140-pcie-75-w-12go-passif-unique-wide-profil-bas-gpu-installation-par-le-client/apd/490-biio/cartes-graphiques

@MJGZ51 Hi! Since 8.3 PCI passthrough is part of the XAPI and come with dedicated xe commands: xe pci-list to have the list of PCI devices and then: xe pci-disable-dom0-access uuid=<pci uuid> to hide th device from dom0 xe pci-enable-dom0-access uuid=<pci uuid> to unhide the from dom0 These are the same API calls made by XO or any XAPI clients No need to edit manually the cmdline anymore! Regards

@yzgulec as people have rightly said above, no hard and fast rule. But when it comes to sizing compute for mixed workloads (or unknown workloads) I have used a 1:4 (pCPU:vCPU) contention ratio. I also try and size 2GB of memory per vCPU (so 8GB RAM per pCPU), and with this 'rough guide' I find that my CPU and memory consumption scale nicely. Obviously it gets difficult when someone has a workload that needs very low CPU but high memory (e.g. I've seen a 2-core VM with 64GB memory) and conversely high CPU and low relative memory (e.g. 8-core VM with 4GB RAM)... but these are typically exceptions. Please share your decisions and build configs here, I'd be interested to see how it worked out in terms of compute:memory, costs, etc.

@manilx It is not cheaper, but it draws about half the power when idle. When doing work the power still gets up there, they are 60-90 watt computers, which is still less than the 300 watt capable HP servers at full go. If you really stress your lab a lot, real servers are a win as they will draw less power for more performance, but if yours is like mine and it idles a lot, cutting the power is nice. I'm about 200 watts at idle with the above, and that's two labs worth. I was 400 with the single old lab system. Mounts are cheap book ends, drilled and screwed to that specific shelf, and drilled for short m4 thumbscrews to get into the VESA mount on the t740. If anyone reading this wants to duplicate, and they see HP T755 cheap enough, I recommend these because it has 6 cores / 12 threads over the t740 with 4 cores / 8 threads. Both are DDR4 sodimm and top out at 64GB. Both have a PCI 3.0 x8 (x16 slot) that takes half height cards. And if the BIOS is locked, you can write over them with a programmer and enter the data needed to function (see badcaps forum for my guide).