How does Intel justify disabling VT-d with K parts?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
This is why I ended up selling my 3570K (managed to get a few bucks more than what I paid for it) and going with a 3770 non K model. I wanted to pass through my 7970 via VT-D to a Windows 8 VM using XEN as the hypervisor. And then have the ability to use the system for all operating systems mainly for training / testing purposes. This worked out well but I just wished my Asrock Z68 board supported more than 32GB of memory.

Also if you have an older motherboard like I did that supports VT-D, don't bother upgrading to Haswell. The 3770 can easily overclock 3-4 bins above stock which makes it faster than the latest 4770 (which can't overclock at all unless you mess with FSB which can be dangerous). I'm running at 4.2 Ghz on all cores at stock voltage. Haswell has some improved VM capabilities over Ivy Bridge but that does not really make up for the Mhz deficit as far as I can tell.

FYI if you're planning on passing through a video card you probably want to stick with AMD based cards, Nvidia cards can be troublesome unless you have a Quadro series.
 

ViRGE

Elite Member, Moderator Emeritus
Oct 9, 1999
31,516
167
106
Yes. I will probably turn this system into a hackintosh (or OSX could even be the guest OS). I plan to do mobile app development and some gaming. Ideally, gaming would be done in Windows (as the guest OS) whenever possible.

Do the latest iMac systems have VT-d disabled? I wouldn't think so. Virtualization with Macs has been huge since long before OSX...so I can't imagine that VT-d or any other virtualization tech would be disabled. I don't want to think that my hackintosh is missing anything that a full Mac would have.

I read an anecdote somewhere stating that, among Z87 boards, Asrock seems to have the best support for VT-d.
I could be wrong here, but it sounds like you're going to run a Type 2 Hypervisor, such as VMWare Fusion or Parallels? If so, then VT-d doesn't even matter for you. VT-d is only useful for Type 1 Hypervisor setups such as VMWare ESXi, where the host is the hypervisor and not an OS.
 

StrangerGuy

Diamond Member
May 9, 2004
8,443
124
106
Because like the haphazard AVX support that next to nobody cares about, Intel thinks it's still dictating the face of computing in their all controlling "I know what is best for you" former monopolistic glory.
 

ashetos

Senior member
Jul 23, 2013
254
14
76
I could be wrong here, but it sounds like you're going to run a Type 2 Hypervisor, such as VMWare Fusion or Parallels? If so, then VT-d doesn't even matter for you. VT-d is only useful for Type 1 Hypervisor setups such as VMWare ESXi, where the host is the hypervisor and not an OS.

With Linux and KVM you can have your pie and eat it too, because you have a fully functional host operating system and VT-d support.
 

chrisjames61

Senior member
Dec 31, 2013
721
446
136
Because like the haphazard AVX support that next to nobody cares about, Intel thinks it's still dictating the face of computing in their all controlling "I know what is best for you" former monopolistic glory.



I think Intel is in a world of trouble because of ARM. The mobile revolution caught the two giants- Intel and Microsoft- with their pants down. ARM and all its licencees and Google and Apple aren't the competition these two companies destroyed in the 1990 and 2000's. In ten years time these two companies will be shells of their former selves.
 
Nov 20, 2009
10,043
2,573
136
Yes. I will probably turn this system into a hackintosh (or OSX could even be the guest OS). I plan to do mobile app development and some gaming. Ideally, gaming would be done in Windows (as the guest OS) whenever possible.

Do the latest iMac systems have VT-d disabled? I wouldn't think so. Virtualization with Macs has been huge since long before OSX...so I can't imagine that VT-d or any other virtualization tech would be disabled. I don't want to think that my hackintosh is missing anything that a full Mac would have.

I read an anecdote somewhere stating that, among Z87 boards, Asrock seems to have the best support for VT-d.
GA-FYI, I am using a 4770K on a GA-Z87X-UD4H board and run OSX Mavericks and W7 on separate SSDs just fine. Neither are running as a VM, but they run just fine otherwise.
 

Kougar

Senior member
Apr 25, 2002
398
1
76
The only native Windows Virtual Machine software that can use VT-D is HyperV. Virtualbox and Windows-based VMware software doesn't support it.

SR_IOV does require VT-d capability, see http://technet.microsoft.com/en-US/library/hh831389 Because HyperV 2012 has SR-IOV capability it's therefore required to support VT-D. I'm not sure if HyperV requires a Gen2 VM to use it though, didn't spot direct settings regarding hardware visualization technology in Hyper V when I looked through it.
 
Last edited:

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
I should have done my research :(

My new build is going to be way more expensive than I had anticipated. On Sunday, I drove 1.5 hours away to Microcenter and bought an Intel Haswell Core i7 4770K from Microcenter with an Asrock Z87E-ITX board. Before this, multiple people told me "get the K" and "you want to get the K." As an on-again/off-again enthusiast, I was pretty sure it was the right choice. I opened the box and put the CPU in the socket...but I haven't technically "installed" it because I haven't attached a cooler of any kind (I haven't even ordered the cooling solution, PSU, or RAM).

VT-d matters *way* more to me than overclocking. I probably wouldn't even bother to try an overclock. Do you think Microcenter would let me exchange it for a non-K part? It would take me 1.5 hours of driving to get there and I'd have to wait a few more days due to this ice storm.

Anyway, it bugs me to think that Intel disabled VT-d for no real reason. Is there an actual reason for it?

Because thats how they get most money from the consumers.
Its actually pretty smart.
When you have rich consumers they can pay more for a small benefit, the poor ones will have little incentive to pay for it as the benefit is very small.

In India most taxis have no meter. Or it suddenly disappear when a rich tourist enter the car :). Now the taxidriver can charge the rich more.

The benefit for the poor is they will get it cheaper
The benefit for the rich is, there is a taxi for them
The benefit for the taxi driver on an average is, he earns just a little bit more

In this example the product is actually the exact same. But there is benefits for all.
Intel segmentation of custormers is a little different. There is benefit for us consumers, but most of the benefit obviously goes to the shareholders.

There have always been an extreme segmentation and relatively price difference for CPU. The more monopoly like the situation is, the more of those segmentation tricks.

The downside for the shareholders is that it is quite possible that the consumers simply move to other similar products like eg arm. So in the long run, those high prices will hit them. But most decisionmakers are pays for short term results. And most shareholders only think 4 year forward.

If we go to the mobile market, nobody (except nobodys like Asus or Lenovo) wants to use Intels product. No wonder. Its very simple, they know what happens. Buy cheap now, buy extremely expensive in 5 years. Samsung, Google and Apple is not going to do that. They are in a different weight class than us average Joes and have power to dictate. And Intel are in the process of learning that.
 
Last edited:

coolpurplefan

Golden Member
Mar 2, 2006
1,243
0
0
This may be slightly off-topic but does anyone know if it makes any difference if you want to use WinXP mode in Win7 with an Intel Core2Duo E6850 or E8500? The E8500 lists vt-d but the E6850 only lists only vt-x. However, I read once Microsoft no longer required CPU virtualization in order to use XP mode. Does this make sense?
 

Kougar

Senior member
Apr 25, 2002
398
1
76
Coolpurple, it wouldn't matter. VT-x is all that 'XP mode' would utilize and both chips support that.
 

daxzy

Senior member
Dec 22, 2013
393
77
101
VT-d is so nebulous in its implementation.

Technically, the first iterations of VT-d were available in the Core2 generation. That's probably the version most people are familiar with, the direct path-through. SR-IOV requires an enhanced VT-d implementation that wasn't around until Tylersburg B3 stepping. But there were many problems with that, stemming mostly around interrupt remapping problems.

The first platform to fully support VT-d was actually Romley (Sandybridge-EP). That's why if you enable SR-IOV for anything that's not Romley, you get a security warning.

From a feature perspective, I'm more annoyed Intel choose to disable ECC on anything above an i3.
 

sm625

Diamond Member
May 6, 2011
8,172
137
106
I'm sorry that a lifer here on these forums would be taken by surprise by this. I wonder how much grief and lost productivity these sorts of games by intel have costed people... and its all for nothing. It's not like intel is going to make more money by doing this.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
I'm sorry that a lifer here on these forums would be taken by surprise by this. I wonder how much grief and lost productivity these sorts of games by intel have costed people... and its all for nothing. It's not like intel is going to make more money by doing this.

Its funny enough always a onesided view. How about all the people that have saved money by not being forced to buy unneeded features.

Want everything, but not willing to pay. Classic.
 
Last edited:

zir_blazer

Golden Member
Jun 6, 2013
1,160
400
136
Its funny enough always a onesided view. How about all the people that have saved money by not being forced to buy unneeded features.

Want everything, but not willing to pay. Classic.
Actually, with a K series Processor compared to the equivalent regular Core i5/i7s, you're expending 20-30 U$D more to get an Unlocked Multiplier, yet losing VT-d and TSX (We can leave TXT and vPro out due Chipset requeriments). Paying more and losing features doesn't make sense, so its the other way around, because all mainstream Core i5 users get VT-d and TSX (Regardless that the vast majority aren't going to use it) but "enthusiasts" that spends more on it.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Actually, with a K series Processor compared to the equivalent regular Core i5/i7s, you're expending 20-30 U$D more to get an Unlocked Multiplier, yet losing VT-d and TSX (We can leave TXT and vPro out due Chipset requeriments). Paying more and losing features doesn't make sense, so its the other way around, because all mainstream Core i5 users get VT-d and TSX (Regardless that the vast majority aren't going to use it) but "enthusiasts" that spends more on it.

I doubt the overall people wanting K models would be willing for example to pay 50$ more than the non K. Just to get TSX and VT-D as well. Assuming TSX and VT-D could be validated to OC modes. Elese people would scream when their data goes haywire.
 

zir_blazer

Golden Member
Jun 6, 2013
1,160
400
136
On LGA 2011 you also have models with both Unlocked Multiplier and VT-d support (Core i7 4820K), so that can't be the reason. I also doubt they need "validation for OC modes" - when overclocking, the end user is responsible for stability and data integrity, if they aren't willing to take risk, they shouldn't overclock at all.
Also, the money itself is relative. After all, all the features are already on the die itself, so there are no extra R&D or manufacturing cost associated with it, is simply Intel whom decides if they're willing to switch an internal flag to enable or disable a feature on some models. If they wanted to give you all them, they could, but seems that artificial market segmentation produces more cash. Problem is that some people are spending MORE money to get some feature, yet lose others in the process. Which seems rather dumb to me.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Also, the money itself is relative. After all, all the features are already on the die itself, so there are no extra R&D or manufacturing cost associated with it, is simply Intel whom decides if they're willing to switch an internal flag to enable or disable a feature on some models. If they wanted to give you all them, they could, but seems that artificial market segmentation produces more cash. Problem is that some people are spending MORE money to get some feature, yet lose others in the process. Which seems rather dumb to me.

There is cost associated with everything. Even if its there already, but not enabled. The same goes for speedbins and so on.

AMD does exactly the same. Takes a perfectly fine quadcore and then disables 1 or 2 cores, simply to make sure they can sell 2 or more products to consumers able and wanting to pay 2 different prices. Without sacrificing the value of the full product.

I can understand alot of people are unable to see these concepts. And just think its evil/greed/dumb whatever. But there is usually a logic behind it.

GPU makers do exactly the same as well. And you can continue on and on with essentially every manufactor.
 

zir_blazer

Golden Member
Jun 6, 2013
1,160
400
136
I can understand alot of people are unable to see these concepts. And just think its evil/greed/dumb whatever. But there is usually a logic behind it.
You totally aimed elsewhere with that comment.

I can understand market segmentation. It makes sense to charge more for Xeons E3 and their Chipsets when they're based on the same silicon than the Desktop counterparts because they got extra features like ECC Memory support, more features for remote management, maybe better validation, etc, and they will be used for serious production or enterprise use, and these guys got more money to burn. Sure, go ahead with adding a premium to them. Is the reason why they're different product lines aimed at totally different markets.
But in the K series case, you are paying for a product which is still supposed to belong to the Desktop series, yet misses several features than the conventional Desktop users DO get. An example is the guy that made this Thread. They take for granted that the 4770K is just a 4770 with Unlocked Multiplier, then later they realize that they're spending more on a product that got some other features removed. The extra 20 U$D for the Unlocked Multiplier should be based as a plus above the same feature set most other Desktop users get, not arbitrarily removing some of them.

And on AMD case, they play with base specifications (Take a bonus that during the 45nm era you were actually able to Unlock actual silicon), but in regard to feature support and instructions sets, they are usually quite straightforward. The only time I recall AMD did something stupid was with the early K8 based Semprons that didn't had AMD64 support. Semprons got AMD64 support by the time of 90nm Venice.
 

ashetos

Senior member
Jul 23, 2013
254
14
76
I have to agree that it is different to offer incremental features for a non-linear price increase than removing features arbitrarily.

I also find inexplicable the lack of transactional memory support in most Haswell models. Or AVX for that matter. Why should the x86 assembly just use the lowest common denominator that is x86_64?

Doesn't intel want to promote innovation, especially when intel is the one doing it? Why not have a new 64-bit instruction superset of x86_64 as a standard, so that software can finally compile in a straightforward manner and with straightforward assembly instructions.

Imagine a linux kernel arch that could be named something like x86_64v2, with AVX, AVX2, AES-NI, transactional memory etc. Now this would make it hard for both AMD and ARM to compete, and it's still backwards compatible with old programs.
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
People will always complain about it, if VT-D was only a feature to K models, people whould be complaining, funny enoght, if VT-D was only a Xeon feature no one whould complain about it, again, VT-D is an extra to Core I5 non K that they have no reason to give at all, and it may be related to no wanting to do that kind of support to OC models. Specially considering that no one whould buy a K model to pair it with a C and Q series chipset that are not meant to OC, Core I non K and C/Q chipsets is something that people do...

REMEMBER. VT-D is only 100% sure to be working on C and Q chipsets, Z/H/B are not meant to support it, its OEM choice.
Personally i see it as something good, because i dont have to buy a Xeon to have it.

So i can understand that one, TSX is the one i dont get it.
 
Last edited: