Sandy Bridge E VT-d broken in launch C1 stepping but will remedied in C2 stepping

CAT-THE-FIFTH · 3 Oct 2011 at 09:58

Sandy Bridge E VT-d broken in launch C1 stepping but will be remedied in C2 stepping

More details here:

http://www.techpowerup.com/152978/S...ixed-In-C2-Stepping-Shortly-After-Launch.html

nightmare99 · 3 Oct 2011 at 11:56

I cannot see them releasing a version with non working VT-d.

N00bbuilder79 · 3 Oct 2011 at 12:03

Well it looks like my upgrade plans may be on hold for a while, might start looking at Sandy Bridge instead.

Tute · 3 Oct 2011 at 12:27

Surely this is hardly going to affect the average user?

stulid · 3 Oct 2011 at 12:32

Considering the chip isn't even out yet, I think they will wait till the fixed versions are ready or will have the fixed ones ready by the due date.

N00bbuilder79 · 3 Oct 2011 at 12:51

Tute said:
Surely this is hardly going to affect the average user?

The thing is why should a consumer buy a product if the producer knows that there is a fault or something not quite right, you wouldn't buy a car if you was told it is capable of 160mph but due to a manufacturing error, it will only do 70Mph, this would be fine for the average user. Would you still spend hard earned cash on a product that by admission is inferior to products released later for more than probably the same price.

joeyjojo · 3 Oct 2011 at 13:04

N00bbuilder79 said:
The thing is why should a consumer buy a product if the producer knows that there is a fault or something not quite right, you wouldn't buy a car if you was told it is capable of 160mph but due to a manufacturing error, it will only do 70Mph, this would be fine for the average user. Would you still spend hard earned cash on a product that by admission is inferior to products released later for more than probably the same price.

Not exactly. Virtualisation is used by a very small minority of users. It's more like having a car with a faulty coat hook. On one side.

drunkenmaster · 3 Oct 2011 at 13:18

Virtualisation used by home users, tiny tiny percentage, sub 1%, when you move to home extreme power users, or data centers, as this is one architecture coming for the high end home market and its the server chip, virtualisation usage is huge. Meaning this is a monumental **** up. Basically if this had happened to the 2600k, no one would bat an eyelid, happening to a server chip, its a HUGE problem.

Theres already been a story on how much is basically screwed up on Sandybridge -e, interconnects most likely causing pci-e 3 support to not make it, support for a lot more sata and various other bits and bobs all missing and mention of a potential new stepping to "fix" all that next year.

Why they would release it, I don't know, in terms of server right now, pre interlagos at least Intel is doing great, though there are plenty of area's opterons are very very competitive in price and power, but AMD are losing what small marketshare they have to Intel.

Will Interlagos change that, will Zambezi on desktop be faster than a 2600k and if so does Intel want a up to 30% faster server chip and desktop chip out asap to be able to still be the fastest out, possible.

AS I've been saying for a while, Sandybridge-e has as many, or more problems than Bulldozer and has been delayed longer, no one is having an easy time on ANY processes right now when it comes to bigger chips and more complex stuff.

aln · 3 Oct 2011 at 13:22

joeyjojo said:
Not exactly. Virtualisation is used by a very small minority of users. It's more like having a car with a faulty coat hook. On one side.

I gather those users tend to generally be the same users that extreme chips are aimed at, regardless of your perception of numbers.

Tute · 3 Oct 2011 at 13:54

N00bbuilder79 said:
The thing is why should a consumer buy a product if the producer knows that there is a fault or something not quite right, you wouldn't buy a car if you was told it is capable of 160mph but due to a manufacturing error, it will only do 70Mph, this would be fine for the average user. Would you still spend hard earned cash on a product that by admission is inferior to products released later for more than probably the same price.

Because

a) These will most likely be out first, so you can take it or leave it and wait for the new revision. To some, being the "first" to have new technology is the most important.

b) That's a very bad analogy.

eddiew · 3 Oct 2011 at 13:58

nightmare99 said:
I cannot see them releasing a version with non working VT-d.

Well one of my idential office machines has hardware virtualisation and one does not... Admittedly they are core2duos, but the point is that Intel have a history of the "same" CPU having different capabilities in this area without actually making it clear when you're buying...

N00bbuilder79 · 3 Oct 2011 at 14:03

Tute said:
Because

a) These will most likely be out first, so you can take it or leave it and wait for the new revision. To some, being the "first" to have new technology is the most important.

b) That's a very bad analogy.

Maybe a bad analogy, however the point is still the same why should I or you for that matter buy a product that is not fully optimised just because the manufacturer has ballsed it up, this isn't acceptable regardless of whether the majority of users would use the part of the chip that isn't working or not. Intel should have the balls to delay release until its been fixed or give the consumer some kind of promise that the early adopters will be able to get their faulty chips replaced when the C2 stepping is released at no cost to them.
I appreciate that developing a new technology comes with headaches during the development process but surely they would be better addressing this before release as opposed to shipping them and rectifying it in the C2 version.

Subliminal Aura · 3 Oct 2011 at 14:19

Tute said:
Surely this is hardly going to affect the average user?

^ This

For those ranting ... why not go away and learn about VT-d and learn why it's broken by motherboard manufacturers even when the chipset and cpu clearly support it. It's been like that for almost 4 years :rolleyes:

As mentioned vt-x is used by a small minority of users and of those interested in vt-d like myself they are rarer than magic beans. I think there was '1' video on youtube dating back a couple of years showing a successful implementation last time I looked.

drunkenmaster · 3 Oct 2011 at 15:43

Subliminal Aura said:
^ This

For those ranting ... why not go away and learn about VT-d and learn why it's broken by motherboard manufacturers even when the chipset and cpu clearly support it. It's been like that for almost 4 years

As mentioned vt-x is used by a small minority of users and of those interested in vt-d like myself they are rarer than magic beans. I think there was '1' video on youtube dating back a couple of years showing a successful implementation last time I looked.

Again, these are SERVER CHIPS, not for average home use, the range will be $350-400 quad core that no one in their right mind would buy, or the "real" chips between $600-1000 hexcores, and the server equivilents that will go for thousands more. In the market this chips are absolutely designed for, virtualisation usage is high.

Subliminal Aura · 3 Oct 2011 at 16:11

drunkenmaster said:
Again, these are SERVER CHIPS, not for average home use, the range will be $350-400 quad core that no one in their right mind would buy, or the "real" chips between $600-1000 hexcores, and the server equivilents that will go for thousands more. In the market this chips are absolutely designed for, virtualisation usage is high.

Not for vt-d it isn't

Please specify an application that uses vt-d in a rackmounted server farm - educate me

stulid · 3 Oct 2011 at 16:16

Something that isn't released yet is broken. Why is this news.

If (and thats a big IF) they do release a first batch of cpu's with a bug, as long as it doesn't impact performance for the average home user then I couldn't care less, its certainly no Phenom TLB bug.

eddiew · 3 Oct 2011 at 16:56

stulid said:
If (and thats a big IF) they do release a first batch of cpu's with a bug, as long as it doesn't impact performance for the average home user then I couldn't care less, its certainly no Phenom TLB bug.

The implication seems to be that they are already manufactured... But I agree that with it being such a small thing and still pre-release, they could just NOT state that it has the feature, rather than stating that it doesn't have the feature. Most won't notice (myself included)

Martini1991 · 3 Oct 2011 at 16:58

Similar to the 2500k, they could list it without the feature, whereas the 2500 has certain features.

Besty · 3 Oct 2011 at 18:43

Subliminal Aura said:
Not for vt-d it isn't

Please specify an application that uses vt-d in a rackmounted server farm - educate me

Some Enterprise-Level virtualisation hypervisors (VMWare vSphere) will not install without hardware acceleration enabled.

Even non-bare metal hypervisors (Parallels, VMWare Fusion, Workstation) will suffer a real performance penalty, if they install at all.

Details are a bit sketchy in the article, I would be surprised if Intel launched a chip with errata that could not be microcode patched after release.

Subliminal Aura · 3 Oct 2011 at 18:52

Besty said:
Some Enterprise-Level virtualisation hypervisors (VMWare vSphere) will not install without hardware acceleration enabled.

Even non-bare metal hypervisors (Parallels, VMWare Fusion, Workstation) will suffer a real performance penalty, if they install at all.

Details are a bit sketchy in the article, I would be surprised if Intel launched a chip with errata that could not be microcode patched after release.

(we're talking about vt-d not vt-x)

I give up with you lot, if anything the lack of knowledge with regards to vt-d goes to show just how little vt-d is actually being used and goes to what we said earlier - namely it's insignificant