Sandy Bridge design flaw - Intel halted on NASDAQ - updated 2/8/11.

[spacejamz]

have u had problems yet?

I had this issue two weeks ago...not sure if it is related...the drive still doesn't show up in the boot section like it did when I put it together on Jan 9.

http://forums.anandtech.com/showthread.php?t=2137622

Seems to working fine right now... I made a copy of my 64GB SSD onto a hidden partition on my 2TB drive and the system can boot from it just in case something happens though...

 

[Dankk]

Newegg recently released a video explaining the problem with the Sandy Bridge chipset, and how to work around it:

http://www.youtube.com/watch?v=tJcE2alQPvY

 

[Dadofamunky]

My RAID boot drive is on SATA 2 as is my BD-ROM, no issues at all. When I install a new SSD boot drive near the end of this month that will be a SATA-3 SSD on a SATA-3 port and the HDDs will make the move over as well. The optical will stay where it is.

 

[maddie]

Don't know if anyone said this yet but Intel's explanation sounds fishy to me. According to Anand, "The solution Intel has devised is to simply remove voltage to the transistor. The chip is functionally no different, but by permanently disabling the transistor the problem will never arise."

If true, failure of the transistor in client's chipsets should accomplish the same result.

Something is not being said.

 

[Spike]

I'm running my SSD and one HDD off of the SATA6 ports and 2 more HDD's and my blu ray off of my SATA3 ports on my Asrock, no issues yet. I highly doubt anything will arise in the next few months before replacements are available.

 

[xeopherith]

that is an outrageous solution that would have done untold harm to their good name.
Replacing the defective chips with working ones is the correct solution.

I agree that the motherboards should be replaced at customer requests however I wouldn't be opposed to the offering of a SATA card in place of a motherboard replacement.

People like me don't have a free PCI slot to use and some of these boards don't even have any other PCI slots. That only leaves PCI Express and again those may also be used already (sound cards and video cards).

John's comments piss me off because it isn't as though a SATA card of any kind actually fixes the defect that the board came with.

 

[toolbag]

A bit off topic, but I've purchased at least four HDD's from Amazon and of the four that I can remember, three were packed in a "box in a box" - with the smaller box having the two molded plastic "cages" that slip over the ends of the drive and keep it suspended in the smaller box. I haven't bought a retail drive in YEARS, but the last one I did buy used the same style packaging.

The other one was wrapped up in bubble wrap, exactly like the optical drives I ordered from Newegg that were packed the same way.

There is some variety in packing methods though - I bought two of the exact same Samsung drive on two separate orders, one was bubble wrapped and the other was in the "cage".

Amazon knows Newegg is competition and they are usually competitive on price. They were more expensive on MB/CPU's when I build my rig last month, so those parts came from Newegg.

I used to be a packer at a computer warehouse. It was a distribution whs for SUN and a lot of other high end stuff in the 90's.

I can assure you I have seen memory that cost over 1000$ in litle plastic packages thrown into 12 cube boxes on the bottom with packing on top only.

Packing areas have boxes of foam, bubble wrap and paper they have sitting around to use. And the packer uses these like an artist. If he cares he packs well. if he is stupid he could care less if it gets there. And most of them are stupid because the packing area is the first place a noob is put usually. And when you train them you think they understand. Then when you stop back later you start to see their craft sucks sometimes. But it's too late hours later..

Everyone packs different. When I packed I wrapped it up good. But I seen some terrible packers do stupid things and a lot of it gets sent out since it can't all be monitored closely. If the supervisor is good he will catch it, but if not we see a bad review of packaging thrown at the whole company.

I know at a local Amazon warehouse you aren't even allowed to stand next to something and rest your arm on it because that's a possible accident waiting to happen. They are pretty strict at the Amazons.

 

[Diogenes2]

Don't know if anyone said this yet but Intel's explanation sounds fishy to me. According to Anand, "The solution Intel has devised is to simply remove voltage to the transistor. The chip is functionally no different, but by permanently disabling the transistor the problem will never arise."

If true, failure of the transistor in client's chipsets should accomplish the same result.

Something is not being said.

Actually, there is a difference between a leaking junction and one that is turned off.
That difference would seem to be a factor in this case..

 

[Idontcare]

Don't know if anyone said this yet but Intel's explanation sounds fishy to me. According to Anand, "The solution Intel has devised is to simply remove voltage to the transistor. The chip is functionally no different, but by permanently disabling the transistor the problem will never arise."

If true, failure of the transistor in client's chipsets should accomplish the same result.

Something is not being said.

Depends on whether the solution is to render the circuit hard-open or hard-short.

A blown transistor can do either, depends on the nature of the failure.

So it is possible, and plausible, that the xtor blows out in a way that renders it a short (substantial gate to drain leakage, an intermittent short that only occurs when voltage is applied to the gate, or source to gate leakage which would be persistent) and the metal fix is to remove voltage from being applied to the gate (permanent open, xtor never turns on again) or to remove the ground line from the drain (a floating ground, but open nonetheless if substrate leakage is minimal).

Likewise, it is possible and plausible that the xtor blows out in a way that renders it open (channel depletion and dopant deactivation...anything that raises the channel resistance or decreases carrier mobility) such that regardless of the voltage applied to the gate the xtor does not turn on because channel resistance is too high, but maybe the circuit needs the xtor to be "seen" as being turned on...so the metal wiring in this case would be to short the source to the drain so the circuit is always on.

 

[grimpr]

Hi IDC, in what timeframe you expect a new Sandy Bridge processor stepping for socket 1155, say Q3?

 

[Modelworks]

I got it right, it was the PLL circuitry.
The reason it is bad is for a PLL circuit to be anything other than near perfect is because that is the reference clock that everything the processor (not the boards cpu, the one inside the chipset), uses to time its actions.
Often they use multipliers and if you have something running at .8Mhz that is suppose to be 1MHz then multiply that times 20 the errors get larger and larger. Worse is when the PLL starts to drift randomly. One second it is .8 the next 1.3 . The code would execute speeding up, slowing down, speeding up, etc. Try aligning code to something like transferring data at a predetermined rate with a hard drive controller that expects everything to remain in sync and it would get slower and slower the worse the clock drifted.

The problem I have with the idea of removing the voltage to the transistor being a solution is that the transistor wasn't designed in that circuit just because they were bored, it had a purpose in the design. The only thing I can think of is that they designed the circuit with enough overhead that removing the bad transistor will not harm the performance. You can't just stop using a part in a circuit unless something else is going to take its place.

 

[wanderer27]

I got it right, it was the PLL circuitry.
The reason it is bad is for a PLL circuit to be anything other than near perfect is because that is the reference clock that everything the processor (not the boards cpu, the one inside the chipset), uses to time its actions.
Often they use multipliers and if you have something running at .8Mhz that is suppose to be 1MHz then multiply that times 20 the errors get larger and larger. Worse is when the PLL starts to drift randomly. One second it is .8 the next 1.3 . The code would execute speeding up, slowing down, speeding up, etc. Try aligning code to something like transferring data at a predetermined rate with a hard drive controller that expects everything to remain in sync and it would get slower and slower the worse the clock drifted.

The problem I have with the idea of removing the voltage to the transistor being a solution is that the transistor wasn't designed in that circuit just because they were bored, it had a purpose in the design. The only thing I can think of is that they designed the circuit with enough overhead that removing the bad transistor will not harm the performance. You can't just stop using a part in a circuit unless something else is going to take its place.

I'm glad I'm not the only one that was thinking this.

I was starting to worry

 

[Arkadrel]

I got it right, it was the PLL circuitry.
The reason it is bad is for a PLL circuit to be anything other than near perfect is because that is the reference clock that everything the processor (not the boards cpu, the one inside the chipset), uses to time its actions.
Often they use multipliers and if you have something running at .8Mhz that is suppose to be 1MHz then multiply that times 20 the errors get larger and larger. Worse is when the PLL starts to drift randomly. One second it is .8 the next 1.3 . The code would execute speeding up, slowing down, speeding up, etc. Try aligning code to something like transferring data at a predetermined rate with a hard drive controller that expects everything to remain in sync and it would get slower and slower the worse the clock drifted.

The problem I have with the idea of removing the voltage to the transistor being a solution is that the transistor wasn't designed in that circuit just because they were bored, it had a purpose in the design. The only thing I can think of is that they designed the circuit with enough overhead that removing the bad transistor will not harm the performance. You can't just stop using a part in a circuit unless something else is going to take its place.

Your not one of those people that always has spare parts unused, after fixing the car and such are you?

^-^ I see your point though... they probably had that component there for a reason, removeing it is bound to effect things in some manner, be that performance/stability or such to some degree.

However... leaveing it there when its makeing all this trouble, corrupting data and loseing performance, isnt gonna work either.

 

[spacejamz]

Just got my email from Microcenter:

eNews Update: Important Information
Micro Center: computers & electronics

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

An Important Announcement regarding Intel(r) Series 6 Chipsets

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Intel has recently identified an issue with their 6 series chipsets,
which are used with all their 2nd generation Core processors
(code-named Sandy Bridge). This is a potentially serious issue, but it
should not affect your data, just your system's performance. Intel believes
that consumers can continue to use their systems with confidence, while
working with their computer manufacturer for a permanent solution.

However, some users may see degradation in the performance of SATA
devices attached to the system, whether internal or external (such as hard
drives and DVD drives). Intel is not aware of any end-user who has seen this
issue yet, but they expect it to affect a significant percentage of users
eventually, and to worsen over a three year period.

Please be assured that Micro Center will stand behind every customer who
purchased a system or a motherboard from us that features this chipset.
Intel has already made the necessary change in the manufacturing process
to correct the error, and properly functioning replacements will be available
in approximately 8 weeks.

To minimize the disruption to you, we suggest that you continue to use your
system until replacement parts are available. At that time, we will contact
you with instructions regarding how to get your motherboard replaced or your
system repaired.

PLEASE NOTE: There is no problem with the Intel 2nd Generation Core Processors themselves.

Please accept our apologies for any inconvenience caused by this issue,
and our assurance that we will keep you informed of any further developments.

 

[Idontcare]

The problem I have with the idea of removing the voltage to the transistor being a solution is that the transistor wasn't designed in that circuit just because they were bored, it had a purpose in the design. The only thing I can think of is that they designed the circuit with enough overhead that removing the bad transistor will not harm the performance. You can't just stop using a part in a circuit unless something else is going to take its place.

It might be a circuit for a feature that impacts performance slightly if negated. Maybe latency goes up a skosh, or realized bandwidth drops by 1%, or CPU overhead to do something has now gone up 0.1%, or power-consumption has gone up a bit, etc.

They haven't told us the side-effects of the fix yet, just that it accomplishes the primary goal.

Remember AMD's TLB fix had side-effects, but the bug itself was definitely mitigated.

The way these projects go, you can bet every chipset project manager is taking advantage of this opportunity to preach to upper management the needs for larger budget and more resources so their existing chipsets can be all the more rigorously tested and validated.

Kinda like when the bridge collapsed in Minneapolis, suddenly the Feds had 50 states all screaming for infrastructure handouts. Good time to be a project manager provided you weren't THE project manager of this specific chipset.

 

[Diogenes2]

According to Anand, it was part of a circuit held over from a previous fab, that is no longer needed. The indication is that this happens all the time.

 

[Arkadrel]

yeah... someone must have overlooked something, that someone... well... 1bn+ is a big pill to swallow, so that person probably getting fired.

Plus it kinda hurts brand when this stuff happends, and sending hardware back and forth, not something costumers love either (I know I hate RMA issues Ive had (I know this is a differnt sort of thing)).

Then theres the fact that they just fought a law suit against nvidia makeing chipsets for motherboards, *IF* nvidia had a working chip without this error, they could have still made processor sales, I believe that was a estimated 300+ million lost due to the 8+ weeks minimum they wouldnt have hardware to sell.

So not only did nvidia walk way from the bank with 1bn+ from intel, but it ends up costing them 300+ mil extra in sales they could have had (intel). Plus it looks bad, they wont allow nvidia to make them chips, and then they make a chip that has a flaw.


yeeep... that person responsible... lost their job.

 

[videoclone]

For those that have a P67 Board and require extra SATA2 ports its probably best to use one of these .< http://www.dealextreme.com/search.dx/search.PCI-E&#37;202-Port SATA > until you can replace your board.

Lots of PCI-e Sata cards to pick from starting from $10 to 17 with FREE shipping

Cant go wrong reviews on each card are also below each product!

 

[Martimus]

It is not a good idea to trust what Intel is saying the consequences of removing the switch is. I have a large amount of experience in auditing technical processes and proposed changes, and the company that is making the changes will sugar coat it as much as they can, and try to make it seem as believable as possible. chances are that there is a negative consequence, just like there was for the TLB fix. We just may never know what that consequence is.

 

[Diogenes2]

If we don't know what the consequence is, how can it be a problem?

 

[ahurtt]

From the article here on AT:

"The source of the problem is actually not even a key part of the 6-series chipset design, it&#8217;s remnant of an earlier design that&#8217;s no longer needed."

Hence, it was left in there out of laziness or an oversight...a vestigial remnant. Nobody thought leaving it in there would make any difference...except they neglected to be neat and tidy and shut it off so now it leaks. And that leak is a problem.

Which is why when I saw item #10 in the errata section of this document:

http://www.intel.com/Assets/PDF/specupdate/324646.pdf

It made me scratch my head and go ..."hmmmmmm"

But anyway, you can all stop speculating about what the effect of shutting off the leaky transistor will be. It's already been stated in the article. Reading comprehension FTW!

 

[Martimus]

From the article here on AT:

"The source of the problem is actually not even a key part of the 6-series chipset design, it’s remnant of an earlier design that’s no longer needed."

Hence, it was left in there out of laziness or an oversight...a vestigial remnant. Nobody thought leaving it in there would make any difference...except they neglected to be neat and tidy and shut it off so now it leaks. And that leak is a problem.

Which is why when I saw item #10 in the errata section of this document:

http://www.intel.com/Assets/PDF/specupdate/324646.pdf

It made me scratch my head and go ..."hmmmmmm"

But anyway, you can all stop speculating about what the effect of shutting off the leaky transistor will be. It's already been stated in the article. Reading comprehension FTW!

Again, I would believe what was stated in that article about as far as I could throw the person who wrote it. I wonder what it really was for and what affect removing it will do. It won't really matter since it is what it is, but it is still a curiosity.

That article didn't give nearly enough detail to tell us anything about what that transistor was for, why it was felt to be needed in the past, why it is said to be unneeded now, and why if it is unneeded now it will negatively affect the SATA bus if it doesn't work properly. Suffice it to say, I would go back to the contractor and get more information or tell them it is not an acceptable change if they were asking me for a deviation based on the information given.

 

[Axon]

Just got my email from Microcenter:

eNews Update: Important Information
Micro Center: computers & electronics

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

An Important Announcement regarding Intel(r) Series 6 Chipsets

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Intel has recently identified an issue with their 6 series chipsets,
which are used with all their 2nd generation Core processors
(code-named Sandy Bridge). This is a potentially serious issue, but it
should not affect your data, just your system's performance. Intel believes
that consumers can continue to use their systems with confidence, while
working with their computer manufacturer for a permanent solution.

However, some users may see degradation in the performance of SATA
devices attached to the system, whether internal or external (such as hard
drives and DVD drives). Intel is not aware of any end-user who has seen this
issue yet, but they expect it to affect a significant percentage of users
eventually, and to worsen over a three year period.

Please be assured that Micro Center will stand behind every customer who
purchased a system or a motherboard from us that features this chipset.
Intel has already made the necessary change in the manufacturing process
to correct the error, and properly functioning replacements will be available
in approximately 8 weeks.

To minimize the disruption to you, we suggest that you continue to use your
system until replacement parts are available. At that time, we will contact
you with instructions regarding how to get your motherboard replaced or your
system repaired.

PLEASE NOTE: There is no problem with the Intel 2nd Generation Core Processors themselves.

Please accept our apologies for any inconvenience caused by this issue,
and our assurance that we will keep you informed of any further developments.

E.g., no refunds.

 

[smakme7757]

E.g., no refunds.

I suppose as a consumer this might annoy some people, but getting a direct swap to a brand new motherbaord (I can't see stores soldering new chips onto the motherboard) is a pretty decent deal. No downtime apart from pulling out and putting in the new mobo.

At the end of the day why get a refund? Why would anyone want to do that? The only logical explanation would be, Principle.

 

[Axon]

You make a good point, and it is a solid deal, even if GB did say it would issue refunds at one point. But here are my points:

Initially, I have a problem with the death grip stores keep on our money these days, and this is just another example of that behavior. It's becoming so ridiculous that I will debate for days or weeks before I give anyone a nickel, because I know I will be locked into that purchase forever. You might feel some impulse to argue against this notion, but by doing so, you're arguing against consumer rights. Is it wrong to ask for a refund of a defective product? At the least, give out store credit. In this situation, the CPU and Mobo are intrinsically tied together, so it should be issued for both. If they want their ridiculous "restocking" fee (e.g. our profit margin when we resell as open box), fine. At least give me the option; the product is defective at its base design. Complacency leads to acceptance, and we're starting to become very accepting of these practices. If I had purchased from a local shop, I'd be more accepting of this kind of policy.

Secondly, newegg is issuing no questions asked refunds for motherboards. I would expect MC to match; if not, they're just a second-tier vendor and I'll consider them as such going forward.

Finally, a refund just gives me more options. Rather than have to wait and go through the later inconvenience, I'd go X58 now - which I just switched from, so I know it does more than what I want it to. Or I could switch up to BD if it comes out and is appealing in April.