Sandy Bridge Boards -- Which ones have vt-d?

/dev/null

/dev/null

[H]F Junkie
Joined
Mar 31, 2001
Messages
15,182
Anyone know? I think traditionally only the Q series have had vt-d support (my DQ45CB) has it, but I don't think my P series has it.

I would assume Sandy Bridge procs support vt-d no?

Edit: Mods, move to motherboard section please...
 
P67 supports it, but motherboard makers may not care to support it. Asus doesn't "support" Linux. If you don't have a NB configuration option to enable/disable VT-d in CMOS setup, that's a bad sign. Even if it's there, it could have broken support.

The k processors don't support VT-d.
 
My thoughts exactly. Either I get something that can't overclock worth a damn (for no good reason) or can't do VGA passthrough. I can't have both, apparently. Oh, wait, there's 890FX...

Unless Intel drops a K part that has VT-d enabled in the next month (literally), AMD is getting my money. I hate having to reset my machine just to play any game more GPU-intensive than Jagged Alliance 2 or Minecraft.
 
Um, I'm building my system specifically for this. Sandy Bridge may be the return of Jesus, the hidden imam and the release of Duke Nukem Forever all in one, but if it doesn't have VT-d enabled on the performance-oriented parts then I. Don't. Care. I've been waiting for Xen VGA passthrough to mature enough that it's worth building a system around for a few years now, and now that it's here (and I have the money) I'm more interested in that than 8GHz on air with the stock cooler or whatever.

I'm coming from a single-core Pentium M and a mobile Radeon 9600. I doubt I'll notice, honestly. ^_^

ETA: Interesting. Apparently the i5-2400 does beat the X6 1090T at quite a few things (even though the margin isn't as extreme as a 5.7GHz 2600K). It *might* be worth it, even with crippled overclock capabilities. Especially if there's an mATX P67 board with known VT-d support - a huge weakness for AMD if that's the case, because I'd strongly prefer mATX if I can get it.
 
Last edited:
So I guess 2400 would be what ill go for. What about the boards? Q only?
 
Robstar said:
Anyone know? I think traditionally only the Q series have had vt-d support (my DQ45CB) has it, but I don't think my P series has it.

I would assume Sandy Bridge procs support vt-d no?

Edit: Mods, move to motherboard section please...
Click to expand...

My MSI H55-E33 board does support VT-d with an i5-650. There is even a BIOS option to enable it. So it is not limited to Q series chipsets.

EDIT: The board is actually called H55M-E33, not H55-E33.
 
Last edited:
omniscence said:
My MSI H55-E33 board does support VT-d with an i5-650. There is even a BIOS option to enable it. So it is not limited to Q series chipsets.
Click to expand...

Thanks! That info is golden!!! That is why I come to [H] :)

You are sure it's "Vt-d" not "vt" right?
 
Robstar said:
Thanks! That info is golden!!! That is why I come to [H] :)

You are sure it's "Vt-d" not "vt" right?
Click to expand...

Yes, I'm sure. But please note that this is a H55 board for Lynnfield/Clarkdale. Your best bet is to download the board manual of your desired board and look for a VT-d option in the BIOS description section. This may be different between manufacturers. My Gigabyte P55A-UD7 does not have a VT-d BIOS option if I remember correctly and I never checked if the functionality is actually there. But I guess not.

For example the manual for the mentioned MSI board describes on page 3-19:

Intel VT-d Tech
This item is used to enable/disable the Intel Virtualization Technology for Directed
I/O (VT-d). For further information please refer to Intel’s official website.
 
Last edited:
Hi, my apology for bringing old thread back. I am borrowing again your thread to link to usage case.

I have been reading data storage forum for sometimes, this is relevancy to that topic per VT-D

The discussion only meaningful for folks looking to extract everything from their equipment. It is of no concern to generic usage.

1. For some with very critical demand on their storage, Windows-based software-raid/fake-raid is not enough for them per their ideal due to concern on RAID software sophistication.

2. For another group, there is a feeling they like other file systems more than NTFS for running certain type of VM workloads. (I may have slight tendency here, I do not have enterprise need, just preference after so many years of using BSD/Linux/misc)

3. This is the final item that ties everything together, that some acknowledge the above 2 points, but need to run Windows as Host.

So, if the combination is true, one way of doing this

1. Windows as host. Activate CPU/Motherboard VT-D support
2. Install virtualization software that supports VT-D
3. Have additional storage HBA with drives attached directly.
4. Configure VM, pass-through HBA to VM with the aid of VT-D.
5. Then one can configure the storage per their ideal and running any disk layouts/filesystems directly, without the need to deal with virtual disk images on top of NTFS filesystem.
6. Since the additional storage HBA is in a way directly controlled by the VM, you also eliminate worry about dependence on Host system driver support or behavior. For some, they want the absolutely minimal distraction from host per the VM operation. With raw access to everything, the VM is as close as possible to running on bare-metal.

Obviously you can run entire scenario without Windows. But for some they have one desktop with Windows for daily needs, but also want to try other custom systems without too much hassle. VT-D potentially can help here.

Finally, per SB, many common models now support VT-D by default, so you have more choices. the P6x chipset series should have good combination. But my understanding is H6x chipset series now common as well. Best is to download manual to read b4 purchasing. (the issue is whether MB makers make effort to include in BIOS, which you can read on BIOS manual)

Cheers

Edit : Jan 13, 2011 : My mistake. The scenario is not wrong, but unfortunately no current Windows VM software supports the feature. It however is a feature frequently requested in VirtualBox case.

If you are running Linux host with Xen dom0, then it is possibility because many refer to Xen documentations for such special configuration. Linux KVM is also reported to be able to do this currently. ESXi is best example for storage HBA pass through but then you cannot use ESXi simultaneously as your desktop, which defeats the purpose of having only one desktop to do everything.

My apology again because I only read the Xen documentation and assume Windows VM software should have no problem doing this. It is a serious lack of concentration and verification when writing to audience of highly technical oriented crowd. Points deducted. :) :)
 
Last edited:
Hi, one final post to correct my mistake.

1. I already updated my previous post regarding wrong information. So it is concluded there.

2. The final piece is implication per consumer buying SB.

2a. Since it looks like for sometime Windows-based VM software options running in generic environment do not support PCI pass through with VT-D, if you are Windows Host person looking at 2500 or 2600, you might as well go 2500K or 2600K and enjoy overclocking if you have such performance intention.

2b. If you are Linux Host type of person with special VT-D requirement plus maximum overclocking potential, then non-K and hope for Z68 to deliver as expected if want everything. Windows-person can go this route if they have intention of later turning their box into test VM lab running other environments.

2b-1. Even that, since generic consumer boards do not support ECC, those do not meet production deployment requirement, and Xeon/VT-d setup is still recommended, as Intel intended.

2b-2 Therefore SB-1155 consumer VT-D basically for home users willing to go Linux running at home environment.

2b-3 As claimed by others 2500/2600 should be useful for OEM workstation with ECC-chipset support combination.
 
Last edited:
Any update on this? The only board I found was the Intel DP67BG, which isn't available yet aparently.
 
I can't seem to find anything, but is there a utility that can report if your system actually has support for VT-d? I don't expect it to report that it actually works, but that nothing is turned off? Everest/AIDA reporting is too generic (it has a VT entry) and I'm not sure it's doing anything other than testing flags.

And it's not like it's trivial to set up the hypervisor and VM to check it.
 
I naively bought a H67 board, which supposedly wont work...so I probably wont even try. Even Intel reports only one of their P67 boards supporting vt-d.

sigh...
 
The best test would be a booting a Linux kernel with enabled PCI passthrough and DMAR and looking into the kernel log. If you are neither familiar with Linux nor ESXi, VT-d is not for you as there is no solution supporting that under Windows as far as I know.

Looking for entries in those benchmark/test tools is not the best option, as at least for the Nehalem generation boards there have been some with bugged implementations.
 
Last edited:
How much faster is a Virtualized OS with vt-d anyway or is it just more secure? I don't get it. :confused:
 
teletran8 said:
How much faster is a Virtualized OS with vt-d anyway or is it just more secure? I don't get it. :confused:
Click to expand...

VT-d lets you virtualize an entire device, such as a PCI-E RAID controller. You could have a file server VM with direct access to said RAID controller.
 
A single review mentions this board has VT-d capability, but I haven't found anything to substantiate that:

http://www.newegg.com/Product/Product.aspx?Item=N82E16813138319

Otherwise, an Asus P8B WS would do (sans overclock-ability), as it definitely has it:

http://www.newegg.com/Product/Product.aspx?Item=N82E16813131725

As PGHammer often states in similar threads, VT-d is often unnecessary for most folks (usually meaning you end up "feature" shopping). There are cases where it's necessary, but modern virtualization methods don't require it (whether using a hypervisor or software such as VMware/VirtualBox/etc).
 
VT-d or Virtualization with Directed I/O is a technology that allows you to assign hardware to VM's. An example might be I am running 5 VM's and I want to ensure that the 2nd VM gets NIC port 2 and then the other VMs can fight it out over the other NIC ports. VT-x is our normal Virtualization technology and can work fine in creating and managing most VMs that you would run on a desktop environment. VT-x is supported on all of our current 2nd generation Intel® Core™ processors. VT-d is supported on our 2nd generation Intel Core i5-2300, i5-2400, i5-2500 and the Intel Core i7-2600. Also in order to tap into VT-d you would have to a board that supports it and there are very few of them in a desktop environment but they would all be using a Q67 chipset. The C206 that is used on the Asus P8B WS is designed for the Intel Xeon® E3-1200’s and doesn’t support the 2nd generation Intel Core processors.
 
Yep, P8B WS takes both SB server/workstation and desktop chips. Having such a long official CPU Support List is a perk of the board imo.
 
Check here of users posting VT-d support on their exsi installs.

I been researching parts to upgrade my VM and one of my requirements is VT-d. Planning to get a ASRock Z68 Extreme3 Gen3 with i5 2400 as my lowest cost option. My other options that I'm considering are SUPERMICRO MBD-X9SCM-F-O with Intel Xeon E3-1220 or 1240 (ECC RAM needed) or a GIGABYTE GA-990FXA-UD5 with AMD Phenom II X6 1090T.

I would like to avoid the ECC RAM requirement if possible since I already have 2x4GB RAM and ECC RAM is more expensive.

Intel Qxx and C2xx chipsets support VT-d.
 
Just to add in case somebody else searches for this: I'm using VT-d on an Asus Z68 Extreme4 (not Gen3) with i5-2500 So not limited to Q chipsets (which was valid for Core2Duo as I've seen VT-d cited for Q35). From what I read about VT-d, I get the impression that the chipset has absolutely no responsability for VT-d since the memory controller was moved to the CPU (1st and 2nd Core i CPUs). As a proof, I would say that the chipset has no relation to the onboard PCIe lanes comming from CPUs (which are obviously availabel for VT-d).
So the only support from MB makers is related to BIOS (all hypervisors query some memory ranges from BIOS).

So regarding my setup:
- Virtualization: Xen 4.1.2
- Host OS: debian testing
- Guest (the one with VT-d): Windows 7 Home Premium (both 32 and 64-bit work)
- Primary GPU: Intel i5-2500 IGP (for host OS)
- 2nd GPU: AMD HD5850, drivers loaded at host OS boot to spin-down fan, but unloaded quickly (so not really used by host)
- Other devices: 2nd USB2.0 controller (for direct use of KB, mouse, wheel, USB sound card)
- Games: Deus Ex HR, Skyrim, Dirt3, iRacing, SWTOR

To-do list:
- GTX260 - I tested yesterday, but does not seem to like the same setup as 5850. Will try some different things over the next weekend (note: nVidia seems to advertise virtualization on their Quadro products, so this may be a SW limitation rather than HW)
- 7950 - looking to purchase one this year. Hopefully, AMD has kept the same compatibility.

PS: I don't have any self-compiled packages. All my host OS is provided debian testing binaries.
PS2: regarding my X-Fi, I have given up on using it though VT-d as I could only use it with Win 7 HDA drivers (I was very surprised when I saw them) and not the original Creative (VM would lock).
 
You must log in or register to reply here.
Back
Top