[vfio-users] Physical vs virtual machine gaming performance comparison benchmarks

[Alex Williamson]

On Sun, Sep 6, 2015 at 8:55 AM, Jon Panozzo <jonp at lime-technology.com>
wrote:

> As for Alex's comment on hugepages, the challenge I've had in convincing
> our team to enable this is that we haven't been able to witness any sort of
> true gain from doing this.  I've turned it on before and from the desktop
> VM, I didn't really see any performance difference.  If there is a specific
> use case that someone could share or a tool/benchmark I could run to show a
> difference, we would gladly implement it and prioritize its inclusion.
>
Try this, run PassMark Performance Test, just the CPU Mark portion of it.
Set the test duration to Very Long, iterations to 5 and check Keep best
result (I'm too lazy to record the results and generate statistics myself
right now).  For me, the VM instance using 2M hugepages comes out over 10%
better.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20150907/c44084f1/attachment.htm>