Find the optimal route between devices?
Find the optimal route between devices?
So what's the issue? It seems like the newer generation is causing problems. I own a system with 2* e5-2680v2 and 256GB RAM, running Proxmox VE—free, open-source, Debian-based. It's quite stable and supports freenas, databases, media servers, and home assistants. I definitely recommend it.
It initially appeared one of the 10Gbit ports was nonfunctional. Swapping it meant covering the cost of the board for "Advanced Shipping," which I ended up paying twice. After they returned the old board, I received a refund. RAM proved to be the main issue—sticking modules together even with matching stickers caused compatibility problems. All components needed to match each other. The integrated X540 NIC overheats, and there’s no simple solution for cooling it when multiple cards are installed. It’s frustrating but functional once it works. I won’t consider purchasing Supermicro or NEMIX again. My current setup includes two E5-2698v3 servers with 0.5TB RAM, running Windows Server, alongside a PROXMOX VE VM server. I really enjoy it.
That's quite frustrating. Running at 10Gbps is demanding, especially with fiber. It seems the fiber isn't handling it well. You only have 1GbE NICs—really limited. Nice! Which GPU are you using? Are you asking about NVIDIA solutions for fixing code 43?
Other servers and desktops use fiber NICs and stay much cooler. They run AMD systems. A fix might seem like a change, but it’s not guaranteed. NVIDIA’s setup can detect virtual environments and reports an error when it thinks the system isn’t running in one. They push for their own GPU lines like Quadro or Tesla. Windows sometimes tricks it into thinking it’s not in a VM, which helps the NVIDIA driver but results aren’t consistent.
AMD seems indifferent, NVIDIA doesn't aim to damage their server/workstation lineup. Why let people bypass a $1000 card and force them to upgrade to a $5000 model with just a few extras? I'm wondering why Linus allows 1080Ti and 2080TI cards to run on UNRAID. It probably comes down to how the NVIDIA driver or hardware recognizes virtual environments—settings not all hypervisors support, so it sometimes works. Personally, I enjoy QEMU. I plan to create a guide for passing through QEMU GPUs in Debian soon.