Ivy Bridge Core i3s and ECC - Santa Clara, we have a problem...

[HoneyBadger]

But doesn't Intel's fine print basically say that nothing they've published can be relied upon?

Nothing on ARK for us filthy commoner consumers, no. But I'm willing to bet that the system vendors have included something a little more ironclad in their contracts that allows for repercussions in case of misrepresentation.

EU consumer protection law also famously doesn't care about whatever fine print is written in 6pt font, so that may be an option in applicable territories.

 

[Tywin]

I just thought about another method: radiation. We know cosmic background radiation can flip bits so increasing this by orders of magnitude with some radioactive source near the RAM sticks should produce some bit flip pretty frequently. Well, not for everyone but if someone has a source safe to handle by hand it can easily try this :)

We periodically have some beam time at TRIUMF, unfortunately the processor's going to be falling over every few seconds making it pretty hard to actually test :p

 

[Bidule0hm]

Well, there is a very small ( :D ) difference between handeld sources and a 500 MeV cyclotron... :p

 

[Whattteva]

I just checked the Intel ARK site.
It appears that "ECC Memory Supported" item on the i3-4160 still says "Yes". Does this mean I'm in the clear?

 

[Ericloewe]

As far as we can tell, Haswell is unaffected by this mess. All Haswell i3s, Pentiums and Celerons seem to support ECC, apart from Devil's Canyon.

 

[DrKK]

As far as we can tell, Haswell is unaffected by this mess. All Haswell i3s, Pentiums and Celerons seem to support ECC, apart from Devil's Canyon.

I'm afraid this really doesn't seem good enough.

I think Intel owes a very, very clear explanation and unambiguous declaration to everyone---but in particular communities like ours---convincing us that they have looked into this issue, and they are officially telling us that the following list X of non-Xeon processors support ECC fully, and the following list Y do not.

I'm hoping for something that sounds roughly like this:

"There has been some confusion about ECC support on our various SKU's, and we made some unfortunate mistakes in indicating on our ark site that certain SKU's supported ECC when they, in fact, did not. We'd like to set the record straight. In addition to our specific line of server-intended CPU's, i.e., Xeon et al, which of course support ECC, you can be assured that all Pentium, Celeron, and i3 CPU's in the 1150/Haswell design fully support ECC. blah blah blah for 1155 blah blah blah for 2011 etc. etc. etc."

That's what I'd like to see.

 

[Tywin]

I'm afraid this really doesn't seem good enough.

I think Intel owes a very, very clear explanation and unambiguous declaration to everyone---but in particular communities like ours---convincing us that they have looked into this issue, and they are officially telling us that the following list X of non-Xeon processors support ECC fully, and the following list Y do not.

I'm hoping for something that sounds roughly like this:

"There has been some confusion about ECC support on our various SKU's, and we made some unfortunate mistakes in indicating on our ark site that certain SKU's supported ECC when they, in fact, did not. We'd like to set the record straight. In addition to our specific line of server-intended CPU's, i.e., Xeon et al, which of course support ECC, you can be assured that all Pentium, Celeron, and i3 CPU's in the 1150/Haswell design fully support ECC. blah blah blah for 1155 blah blah blah for 2011 etc. etc. etc."

That's what I'd like to see.

Agreed. Unfortunately, I highly doubt you'll see anything close to that, if anything at all. It's frankly just too small of an issue for Intel to care about. The quantities manufacturers like Intel deal in are mind-boggling, and I suspect that in greater than 99% of cases ECC is a hedge and not a core requirement or design choice. You'd get a response if you were a customer like Apple promising ECC and discovering after the fact that it wasn't actually supported on their processors.

 

[DrKK]

Agreed. Unfortunately, I highly doubt you'll see anything close to that, if anything at all. It's frankly just too small of an issue for Intel to care about. The quantities manufacturers like Intel deal in are mind-boggling, and I suspect that in greater than 99% of cases ECC is a hedge and not a core requirement or design choice. You'd get a response if you were a customer like Apple promising ECC and discovering after the fact that it wasn't actually supported on their processors.

On the other side, this is a very small request. They can dispatch one or two nerds to check it out, and write something up.

 

[Tywin]

On the other side, this is a very small request. They can dispatch one or two nerds to check it out, and write something up.

The actually doing of it is small, and there may even be a couple nerds that have already looked into it and have written something up. That's a long way from a published press release that the company will stand behind, however. It doesn't just have to be factual, they have to convince the management that it would demonstrably increase Intel's bottom line to release it. I'm not saying it won't happen, I'm just trying to manage your expectations ;)

 

[Whattteva]

Is it just me or lately we have been getting a string of sketchy maneuvers from manufacturers?
It started with Kingston and their RAM modules, then their SSD's (PNY and several others did the same thing also), and now even Intel succumbs to this bait and switch game.

 

[avalon60]

I get this:
[root@freenas /mnt/Backup_Data]# python ecc_check.py
5004-5007h: 20 0 63 3
5008-500Bh: 20 0 63 3

Supermicro x9SCM-F
Intel(R) Core(TM) i3-3220 CPU @ 3.30GHz
x2 8GB Kingston value ram ECC DDR3

Like others, when I bought the cpu for the board last year, ECC was supported, but now it isn't?????
To avoid any problems in the future I have just bought a Xeon E3 1240v2 from fleabay

 

[cyberjock]

Like others, when I bought the cpu for the board last year, ECC was supported, but now it isn't?????

No, it means that when you bought it the ARK said it was supported. But now Intel has admitted it was in error and it should say it is not supported. Assuming anything else except what I just said is an indulgence.

 

[Jannis Jacobsen]

a real thread ressurection, but here goes.
Tried cyberjocks check_ecc.py on my freenas box, Intel S1200KPR with i3-3220.
Results:
5004-5007h: 20 0 62 3
5008-500Bh: 20 0 62 3

Seems like this does full ecc with i3.

-Jannis

 

[cyberjock]

We've already proven that you can't trust check_ecc.py when using i3s. ;)

It can be erroneous. :/

The reality is that if the Intel ARK says it is supported, that's about the best you can do. :(

 

[Robbks]

Apologies if i'm digging an old thread here,

But I'm planning a basic Media Storage build using an existing Board with i3 Processor.
reading through lots of information has shown me the light on using ECC RAM, but if it's not compatible with my hardware.

I'll interrogate the machine tonight and track down the models.
needing to change the processor will take the "budget" out of my plans and cause a re-think

EDIT:
Core i3-2100 Sandy Bridge
Gigabyte Ultra Durable 2, 1155 socket. can't find it's exact model# though.
Dual channel DDR3, dual bios.

 

[Stux]

Has anyone done the sticky tape test to get to the bottom of this?

 

[DrKK]

Apologies if i'm digging an old thread here,

But I'm planning a basic Media Storage build using an existing Board with i3 Processor.
reading through lots of information has shown me the light on using ECC RAM, but if it's not compatible with my hardware.

I'll interrogate the machine tonight and track down the models.
needing to change the processor will take the "budget" out of my plans and cause a re-think

EDIT:
Core i3-2100 Sandy Bridge
Gigabyte Ultra Durable 2, 1155 socket. can't find it's exact model# though.
Dual channel DDR3, dual bios.

The best that I can tell, you had absolutely no reason whatsoever to resuscitate this post, and bring it to everyone's "new post" list. I'm not even sure what your question, if any, is.

In the future, please make a new post, and refer/link back to this one if necessary.

 

[Ericloewe]

Has anyone done the sticky tape test to get to the bottom of this?

Not as far as I know, and software methods have eluded us, too.

At the very least, we know that Haswell i3s are indeed capable of ECC, since mine ended up correcting a whole bunch of errors earlier this year after a PSU decided to ragequit and bring down everything attached to the UPS it was running off of.

The best that I can tell, you had absolutely no reason whatsoever to resuscitate this post, and bring it to everyone's "new post" list. I'm not even sure what your question, if any, is.

In the future, please make a new post, and refer/link back to this one if necessary.

As @DrKK implies, there's little of consequence to add to this thread, unless some real information pops up.

 

[Robbks]

The best that I can tell, you had absolutely no reason whatsoever to resuscitate this post, and bring it to everyone's "new post" list. I'm not even sure what your question, if any, is.

In the future, please make a new post, and refer/link back to this one if necessary.

Apologies for that.

In all other Forums I'm a part of, it's good etiquette to search the forum for a recent, relevant thread and add your question there,
rather than have 50 threads all over the forum asking the same question, all containing little bits of information.

 

[Ericloewe]

Apologies for that.

In all other Forums I'm a part of, it's good etiquette to search the forum for a recent, relevant thread and add your question there,
rather than have 50 threads all over the forum asking the same question, all containing little bits of information.

It's mostly because your post doesn't add anything to the discussion, it focuses on a rather different issue. Please feel free to ask whatever questions you have (preferably sharing more specifics) in a new thread.