HP WS460C HowTo enable MultiGPU for XenServer 6.x
XenServer 6.x makes it possible of passthrough of multiple GPU’s.
HP blade workstation WS460C is an enterprise solution for this area that fits perfect with Citrix XenServer 6.x multiple GPU passthrough. You can attach an Graphic Expansion to the WS460C so its possible to integrate two graphic cards like (Nvidia FX3800, FX4800, FX5800 or the new fermi cards 4000, 5000, 6000)
The target here is to support high-end 3D professional graphics applications with multiple GPU passthrough. Applications like Autodesk/Dassualt Systems and many more CAD/CAM systems are often using OpenGL and or Direct-X and Citrix provides the full user experience with HDX3D to deliver this everywhere. HDX3D is available both for XenDesktop 4/5.x> and XenApp 5/6.
This article descripes HowTo enable MultiGPU with HP blade workstation WS460C with HP Graphic Expansion blade and Citrix XenServer 6.x
Bios Settings for ws460cG6 with Graphics Expansion Blade
Option “HP Graphics Expander x16”
Help (F1 pressed) associated to this option
“When enabled will allow x16 PCIe connection on HP Graphics Expander slot 1. Slot 2 will NOT be functional”
HowTo enable MultiGPU in HP WS460c with Graphic Expansion blade
Choose this BIOS setting for “HP Graphics Expander x16” for enabling both graphic cards. When the BIOS setting is disabled the Citrix XenServer 6.x is able to see both Graphic cards and you can then passthrough the GPU’s to the Virtual machines.
HP Graphics Expander x16: DISABLED
- HP Graphics Expander slot 1 : x8
- HP Graphics Expander slot 2 : x8
Enabling “VT for direct I/O” in the BIOS

HowTo identify the Multiple GPU’s in Citrix XenServer 6.x
Note XenServer 6.x XenCenter Console only shows 1 GPU if you have multiple GPU’s *this is default design belowed explains why:
If you have multiple GPUs in different PCI slots then Xenserver forms the groups of identical GPUs. For example if you have 2 – Quadro 4000 GPUs then there will be one Quadro 4000 GPU group identified by Xenserver.
Now when you try to assign the GPU to a VM from Xencenter it shows the GPU group and not the individual GPU. In your case one GPU group has 2 – Quadro 4000 cards, so you can assign this GPU group to 2 VMs wherein each GPU is attached to an individual VM

If you want to identify that XenServer 6.x is showing multiple GPU’s access local command shell
Execute #lspci to view multiple GPU’s
In belowed picture you can see i have identified 2 GPU’s with the command #lspci




Thomas Poppelgaard
| #
Hi Steve normal I install the GRID driver to device and when its workin, I disable the default VGA device.
Have you applied latest hotfix to the XS6.1 there are some crucial hotfixes to the GPU pass-through, that could crash the GRID’s, when you power on a VM with GPU pass-through.
Steve
| #
Just a quick question on your setup. We just got a system similar to yours (R720, 2xGrid K2 cards, XenServer 6.1, XenApp 6.5). The GRID K2 driver installs (320.00), but the device shows up as stopped in device manager. Did you use any special options when installing the nVidia driver? Or somehow remove the default VGA device?
Dan
| #
Hi Thomas,
Some features in your screenshots (eg create appointments and contacts) seem to be missing from the iOS version of @WorkMail that got released in April, do you know if these features are still coming in a future release? The Android client is far more functional by comparison.
Regards
Dan
Christian Eilskov
| #
You can see the DHCP options here:
http://www.wyse.com/kb and search for 21501
You can transfer a image using Wyse Device Manager(WDM), the same goes for smaller updates like new ICA client and so.
Thomas Poppelgaard
| #
Thank you Barry =)
The deep compression codec for Citrix XenDesktop HDX 3D Pro will be intergrated for Citrix XenApp in Excalibur so there is a big difference with bandwidth consumption. This means that XenApp in Excalibur will be the best platform for user density and works great over WAN with high latency as HDX 3D Pro have been known to deliver for several years. Yes i know of cases with WAN optimization, I will gather these and share them.
Best regards
Thomas