x86: 0.989 architecture: 0.956 device: 0.827 graphic: 0.789 TCG: 0.757 performance: 0.697 risc-v: 0.667 ppc: 0.632 register: 0.608 network: 0.588 permissions: 0.575 semantic: 0.566 arm: 0.549 boot: 0.540 vnc: 0.527 socket: 0.461 mistranslation: 0.455 kernel: 0.451 files: 0.429 PID: 0.391 hypervisor: 0.379 i386: 0.345 debug: 0.336 virtual: 0.329 assembly: 0.328 user-level: 0.315 VMM: 0.302 KVM: 0.196 peripherals: 0.147 -------------------- TCG: 0.876 x86: 0.830 ppc: 0.653 architecture: 0.365 virtual: 0.205 arm: 0.194 hypervisor: 0.070 register: 0.041 performance: 0.023 KVM: 0.019 files: 0.010 user-level: 0.010 kernel: 0.009 PID: 0.009 i386: 0.007 network: 0.006 device: 0.005 semantic: 0.005 socket: 0.004 debug: 0.003 boot: 0.003 risc-v: 0.002 VMM: 0.002 peripherals: 0.001 permissions: 0.001 assembly: 0.001 graphic: 0.001 vnc: 0.001 mistranslation: 0.000 RFE: Please implement -cpu best or a CPU fallback option QEMU should implement a -cpu best option or some other way to make this work: qemu -M pc,accel=kvm:tcg -cpu best qemu -M pc,accel=kvm:tcg -cpu host:qemu64 See also: https://bugzilla.redhat.com/show_bug.cgi?id=1277744#c6 Instead of having a new CPU model called "best", we can simply make "-cpu host" available on TCG. I have submitted a patch for this yesterday: https://