Intel vt-d/AMD IOMMU support

[vizio]

I want to use the pci-e inside the guest os to accelerate the directx and opengl but virtualbox still not support it and I found that intel vt-d can use the host pci-e in parallel workstatio 4.0

Hope to support this in future version.

[Technologov]

Added myself to this request. +1.

-Technologov

[Sander van Leeuwen]

Let's include AMD's VT-d equivalent (IOMMU) as well.

[Technologov]

This wish looks related to wish #3854

-Technologov

[bpreston]

And I would like to add myself as well

-B

Replying to Technologov:

Added myself to this request. +1.

-Technologov

[zexelon]

I would very much like to see the AMD IOMMU or Intel VT-d feature implemented!

[Dusty]

+1 Would like to see this asap.

[zerix01]

I would like to see near host speed GPU acceleration within the VM. My understanding is IOMMU would make this possible.

[yesoos]

+1 From me, would be nice if we could "attach" a pci or pci express device to virtual machine, ie. GPU

[Jason]

+1 for me too.

[Bruce Edge]

+1 - PCI passthrough (which I think is what's intended here) is a must.

[Oberst]

+1 for me too...

[gu99roax]

+1 For me too, but I hope that the VT-d/IOMMU implementation not only implies support for PCI passthrough but also support for VGA passthrough and full support for PCI passthrough of multifunction devices which is what you can do at least in Xen 4.1 and later.

[esmail]

+1 me too please

[Aquous]

+1 for me too... (especially VGA passtrough, I'd love to be able to play my ancient DirectX 6 games again)

Also it looks like this support is already on the way for at least Linux hosts: ​http://www.virtualbox.org/svn/vbox/trunk/doc/manual/en_US/user_AdvancedTopics.xml "Experimental support for PCI passthrough was introduced with VirtualBox 4.1."

[Andreas Schöneck]

Any little update on this (regarding possibly ongoing work)?

[powerhouse]

+1 for me. I read it's partially supported now and wait for my new hardware to arrive to give it a test. Mainly need it for VGA passthrough to calibrate display from within Windows 7 VM. Hope this solves my having to dual boot.

What is the status for graphic adapter support? Since I am going to purchase an adapter soon, it would be good to know what's supported and what's not.

[Technologov]

AFAIK only network cards are supported ATM.

Not tested due to lack of hardware.

-Technologov

[powerhouse]

Thanks Technologov.

With regard to lack of hardware, I would be able to do some tests once I get the new rig by the end of the month: i7 3930K with C2 stepping for VT-d support; Asus Sabretooth X79 MB; NVidia GeForce GTX 550 Ti; lots of RAM; 120 Gig SSD drive; several TB HD drives.

I already have the NEC wide gamut screen with colorimeter (connected to USB port) for hardware calibration inside the screen. This is why I need PCI / VGA passthrough so that Win 7 and the calibration software can download the calibration profiles into the screen via the DVI connector on the graphics adapter.

May not be everyone's cup of coffee, but this is a basic setup for photo editing. Would be fantastic if I could get rid of a dual-boot setup.

[Alduin]

I hope we are talking about PCI/VGA passthrough for VirtualBox on Windows hosts here and not something else. It would be great if it also were possible to pass through entire host bus adapters such as LSI MegaRAID 1068e/2008e (e.g. Intel SASUC8I, LSI 3801e-r, Dell Perc 5i, etc) to virtual hosts.

I'm also wondering if FLR really must be supported by the PCI card for PCI/VGA passthrough to work or if it will work anyway. As I understand it, the hardware must be reset somehow before it gets passed through to the virtual host. I read on a pdf document a while ago on the VMWare website (​http://www.vmware.com/files/pdf/techpaper/vsp_4_vmdirectpath_host.pdf) that:

---

Reset Method

Possible values for the reset method include flr, d3d0, link, bridge, or default.

The default setting is described as follows. If a device supports function level reset (FLR), ESX always uses FLR. If the device does not support FLR, ESX next defaults to link reset and bus reset in that order. Link reset and bus reset might prevent some devices from being assigned to different virtual machines, or from being assigned between the VMkernel and virtual machines. In the absence of FLR, it is possible to use PCI Power Management capability (D3 to D0 transitions) to trigger a reset. Most of the Intel NICs and various other HBAs support this mode.

---

So there are other ways of resetting hardware to be used for passthrough than the FLR, namely d3d0, link and bridge. I think these features are based on the ACPI power management which allows for turning off individual pieces of hardware.

[Leho Kraav]

following