I just upgraded a system from a i7-7700 to a i7-10700. The hardware is nearly identical except for the CPU and motherboard. It's an asrock z490m.
The host operating system is vmware esxi 7.0b. (originally 7.0, but i tried upgrading)
When a 64bit guest is used, the system seems to work fine. A 32bit guest running MidnightBSD can cause a double fault when running certain applications. I've got a 100% reproducible case but I haven't been able to narrow it down to the exact command yet. It's a shell script (bsdstats) running several programs. I also have seen it when compiling part of QT5.
I'm not sure it's actually a bug with the CPU or UEFI, but the same code does work on a i7-7700.
Things I've tried:
* updating the bios to the latest.
* Turning SR-IOV on/off, and sxguard settings
* Slowing down the RAM to DDR4 2133 rather than 2933. They're corsair modules.
* Enabling a DEBUG kernel in the operating system. Seems to only capture the second fault as far as i can tell.
As it's a 10700 (not a k or f variant), it's not overclocked. It's using the stock cooler. It's a pretty boring setup. The onboard NIC wasn't supported in vmware so i put in an older intel pro gigabit nic. The memory modules are listed as supported but not overclockable on the asrock site. There are 4 8GB modules.
The guest system has 3GB of RAM allocated and 8 virtual cores.
Has anyone else experienced any issues with 32bit guests on these processors?
I've narrowed it down to /bin/mksh. Anytime that runs, it crashes. It does not happen on FreeBSD i386 10.4 running in a VM. I've also tried GCC instead of LLVM to build mksh and it still happens. I've tried a newer mksh and it happens.
I'm wondering if it's some kind of alignment issue and the newer CPU is just pickier about it.
I also tried an older MidnightBSD release (1.1) and it will launch mksh, but still crashes when mksh runs the bsdstats script in the same way.