His love of PCs and their components was born out of trying to squeeze every ounce of performance out of the family computer. Tinkering with his own build at age 10 turned into building PCs for friends and family, fostering a passion that would ultimately take shape as a career path.
Besides being the first call for tech support for those close to him, Ty is a computer science student, with his focus being cloud computing and networking. He also competed in semi-pro Counter-Strike for 8 years, making him intimately familiar with everything to do with peripherals.
Every time you spin up a virtual machine in Proxmox, the creation wizard makes a stop at the System tab that most people just click straight through. One of the fields there is Machine, and unless you change it manually, it defaults to i440fx, which tells QEMU to build your VM around emulating an Intel chipset from the 90s. Intel's 440FX chipset launched in 1996 for the Pentium Pro, and there is a much more modern option in the list, so why doesn't Proxmox use it?
It sounds like a problem, but it's not really a concern. The machine type is your VM’s virtual motherboard, and Proxmox defaults to i440FX because it’s the safe, compatible choice. Once you know what it actually controls, you’ll know when to leave it alone and when to pick something newer, which isn't all that often.
What machine type actually controls in Proxmox VMs
Proxmox's forgotten VM setting
The machine type dropdown selects the hardware layout of your VM's virtual motherboard. When a guest operating system boots, it finds a chipset, a set of buses, and a handful of built-in controllers, and the machine type dictates what all of those are.
It's easy to confuse with the settings you find around it, like the CPU type, which decides which processor features the guest can see, or the firmware setting, which chooses between SeaBIOS and OVMF. Your disk controller and network card are devices plugged into the virtual board, and the machine type is the board itself, so changing it is like keeping the same processor, memory, and drives but swapping the motherboard underneath them.
The 440FX that's the default inside Proxmox gives the guest a host bridge, an ISA bridge, an IDE controller, a USB controller, and an ACPI power-management device, all sitting on a conventional PCI bus. There’s no PCI Express anywhere on it, and you can confirm this yourself by running lspci in a Linux VM, where the first few entries list the chipset’s parts by their original Intel names.
Why a chipset from 1996 isn't making your VM slow
It doesn't have Pentium Pro-era performance
If your VM is built on a Pentium Pro-era board, you might expect it to perform like one, but the chipset has very little to do with performance in a VM.
Your guest’s code runs on your host’s processor through KVM, and the CPU type only controls which instruction-set features are exposed to the guest. The wizard’s default, x86-64-v2-AES, targets Intel CPUs from the Westmere generation onward, far newer than anything a real 440FX ever hosted. QEMU just reproduces the chipset’s layout, not its limitations, so there’s no 1996-era bus speed throttling your VM.
Storage and networking are where performance lives, and Proxmox recommends handling both with VirtIO. Rather than imitating a real disk controller or network card, a VirtIO device cooperates with the hypervisor directly, which is much faster than emulating legacy hardware. VirtIO SCSI single has been the default disk controller for new Linux VMs since Proxmox VE 7.3. These devices plug into the virtual PCI bus like any other card, so the age of the chipset around them barely registers.
Where the i440FX does hold you back
Handing physical devices from the host to the VM
The other option in the Machine dropdown is Q35, which emulates Intel’s Q35 Express chipset and ICH9 southbridge from 2007. That’s hardly cutting-edge, but it brings one thing i440FX can’t: a virtual PCI Express bus.
That matters most for passthrough, where you hand a physical device in your host, like a GPU, directly to a VM. Proxmox’s option to present a passed-through device on a PCIe port only works with Q35, and plain PCI isn't going to cut it here. Passthrough isn’t impossible on i440FX, but devices show up there as conventional PCI cards, which doesn’t work for every guest or driver, and if you're planning on passing through a GPU, Q35 is definitely the better starting point.
Q35 is also required for Intel’s flavor of emulated IOMMU. Additionally, it's what the Proxmox wizard selects automatically when you set the guest OS to Windows 11, along with UEFI firmware and a TPM. If you’ve built a Windows 11 VM, you may already be running Q35 without realizing it.
Why you shouldn't flip every VM to Q35
It's not always necessary
With all of that in mind, it's tempting to open every VM you run and switch it over now, but you shouldn't. At least, not without a plan as to why.
On Linux, the usual casualty is networking in an active VM. Most modern distributions name network interfaces based on where they sit on the bus, and that changes with the new layout. An interface called ens18 can come back as something like enp6s18, and if your network configuration still points at the old name, the VM boots without an IP address. Windows can be fussier still and may need repair work after the switch, and while neither is really hard to fix, both are annoying on a VM that was working fine beforehand.
That's probably why Proxmox hasn't changed the default, because a VM without an explicit type set just uses the default, so changing it would just swap every VM that was created without one.
Leave the machine type alone unless you need it
Seeing a 1996 chipset in your VM’s settings is a little jarring, but i440FX isn’t Proxmox neglecting to update a dropdown or a hefty performance sacrifice in the name of compatibility. It is for the sake of compatibility, but it works best out of the box for virtually all operating systems, and the parts that matter for performance have nothing to do with the VM's chipset.