Thanks much for the reply and for the attachment.
Given that iGVT-g can't yet be used with a Windows host, I certainly
agree that most OEMs would be more receptive to increasing the
aperture size for their Linux servers/workstations, and less receptive
to increasing the aperture size for their Windows laptops. There just
aren't that many of us (yet) who run Linux on laptops.
But I can tell you (and you probably know) that many of us who run
Linux on our laptops are very excited about iGVT-g. We'll finally be
able to run VMs with acceptable graphics performance!
As you continue to reduce the minimum aperture size needed for iGVT-g,
the situation is getting better for Linux laptop users. But I'd still
like to run more concurrent VMs on my laptop than I currently can,
owing to its small (non-configurable) aperture size.
I predict that, once we Linux users demonstrate the incredible
usefulness of iGVT-g on a laptop, Microsoft and others will provide a
means for using iGVT-g on Windows laptops. But I think that day will
likely come sooner if the default aperture sizes on laptops were
Even if OEMs can't be convinced to include a configurable aperture
size (and a config menu for it in the UEFI), I still hope that more
OEMs will increase the default (non-configurable) aperture size for
laptops to at least 512M.
So I still continue to hope Intel will encourage the major laptop
manufactures (and UEFI developers) to use larger default apertures (at
least 512M) moving forward.
Thanks again for helping to bring iGVT-g to many eager Linux users!
On Mon, Dec 5, 2016 at 9:29 PM, Wang, Hongbo <hongbo.wang(a)intel.com> wrote:
Thanks for your feedback.
1) The reference UEFI BIOS provided by Intel has a configurable aperture size and a
menu for it.
2) The default aperture size of Intel client machines and Intel E3 server is usually set
to 1G and
be configurable. If you're looking for an off-shelf machine for your evaluation,
check out the attachment.
3) We also work with OEM to push larger aperture size. My feeling is the client machines
like desktop, laptop are in slow pace to accept this idea, because not all vendors
the usage benefit of running VMs on client machine. On the other hand, the server
is flexible and willing to set a large size.
Tel: +86-21-6116 7445
MP: +86-1364 1793 689
> -----Original Message-----
> From: iGVT-g [mailto:firstname.lastname@example.org] On Behalf Of Gizmo
> Sent: Tuesday, December 6, 2016 6:00 AM
> To: Zhang, Pei <pei.zhang(a)intel.com>
> Cc: igvt-g(a)lists.01.org
> Subject: Re: [iGVT-g] Minimum GPU aperture size per VM?
> Hi Pei,
> Thanks much for the reply. That's great news!
> I agree that a 256M aperture is the default and only choice on most laptops.
> That severely limits what can be done on otherwise suitable laptops. I
> don't know what Intel can do to help remedy this situation, but I hope that
> maybe Intel will consider sending out notices to the major laptop
> manufactures (and UEFI developers) encouraging the use of larger default
> apertures (at least 512M), or better still, allowing for user selection of
> aperture size, moving forward.
> Thanks again for all the hard work!
> Best regards,
> On Mon, Dec 5, 2016 at 2:32 AM, Zhang, Pei <pei.zhang(a)intel.com> wrote:
> > The minimum aperture size is 256M in fact. In current design, a guest needs
> at least 128M (Windows guest) or 64M (Linux) aperture memory to support
> internal resource requirement. And host also reserves 96M by default. So
> normally 256M aperture memory is the minimum size a user could choose in
> the BIOS setting to run and test our KVMGT. In most laptop, 256M aperture is
> the default and only choice.
> > We are working on a vgpu scalability feature, which requires smaller
> aperture(64M) for Windows 8.1 and later guest OS. So more KVMGT guests
> could be created with same aperture size, if the guest is running Windows
> 8.1 and later OS. This feature is planned to be included in next release.
> >> -----Original Message-----
> >> From: iGVT-g [mailto:email@example.com] On Behalf Of Gizmo
> >> Chicken
> >> Sent: Friday, December 2, 2016 16:53
> >> To: igvt-g(a)lists.01.org
> >> Subject: [iGVT-g] Minimum GPU aperture size per VM?
> >> If I recall correctly, at one time, 512M was the minimum GPU aperture
> >> size required for running a single VM using KVMGT.
> >> Has the the minimum GPU aperture size changed? If so, to what?
> >> And Is the minimum GPU aperture size expected to decrease in the
> >> Also, does the minimum GPU aperture size differ depending on whether
> >> running a Windows VM or Linux VM?
> >> And finally, is there a Linux command for determine GPU aperture size?
> >> (GPU aperture size isn't adjustable in my UEFI settings.)
> >> Thanks,
> >> GizmoChicken
> >> _______________________________________________
> >> iGVT-g mailing list
> >> iGVT-g(a)lists.01.org
> >> https://lists.01.org/mailman/listinfo/igvt-g
> iGVT-g mailing list