Intel halts production/shipment of 6-series chipsets with SATA flaw found
Subject: Chipsets | January 31, 2011 - 04:59 PM | Ryan Shrout
Happy Monday everyone; except those of you that might work at Intel. It looks like Intel has found (and is readily admitting to) a pretty substantial flaw in the current shipping Intel 6-series chipsets used by the Sandy Bridge processors that were released earlier this month. What could be so series that Intel is taking the time to publicly out itself and plan for a $700M budget hit for recall / replacement?
The error and source of degradation is in the silicon itself - not a driver or software or firmware error at all. In some cases Intel has seen silicon degradation in the section of the 6-series chipsets where SATA communication occurs. Apparently the error in the chipset was "obvious" to Intel engineers once they found the problem and the "circuit design oversight" can be corrected with a very simple, late in the production process, metal layer change to return it to a fully robust technology. Intel is apparently very confident in the solution and that they understand the issue affecting the 6-series products.
More specifics of the problem were discussed on the conference call. First, only the SATA 3.0 Gb/s ports are potentially affected, NOT the 6.0 Gb/s ports that are the latest addition to the platform. The "performance degradation" is actually a potentially increased bit error rate that could eventually result in a disconnected channel. You know that error you get on your system where a hard drive can't be found? Yeah, that.
Apparently Intel doesn't think that a single user has been affected by this problem as of today and that if they are that data loss isn't a problem. If you are using a retail motherboard you can just switch to another SATA channel but obviously on notebooks that isn't feasible. Intel estimates that the failure rate over a 3-year system life would be about 5% - that is a HUGE number of potential problems and thus the extreme caution with a halt and recall.
Anand Shimpi asked during the call when Intel discovered the problem, essentially asking them how long they have been sitting on it. Intel responded that it was a very recent discovery, as recent as last week, and that the engineering team basically found the error and solution around this weekend. The original source of the problem was an OEM that had tested their platforms with extreme heat and voltage environments and then turned the failure over to Intel to investigate.
An ASUS P67 chipset motherboard that will likely have to be returned now...
This is a pretty extreme error as 6-series chipsets are in every single Sandy Bridge-based platform sold or released including notebooks, desktop systems and even motherboards that our readers here may have purchased. It looks like there isn't a risk of data loss though, as Intel is comfortable with people continuing to use the product as it is today:
Intel is telling us they have completely stopped shipping the affected chipset and already begun production of the revised and fixed version of the 6-series chipsets. But shipping of those chipsets won't start until late February and full "inventory replacement" won't happen until sometime in April. Intel has committed to working with partners to help facilitate the recall of affected products once production is in full swing, but this isn't something easy to do. You can't just replace a hard drive or memory module here - full motherboard replacement is going to be required on desktop and notebook platforms - a costly error indeed. The company has declared a $700M write off estimate for this cause.
If anything, Intel can be ironically glad that adoption might have been slower than expected of the 2nd Generation Intel Core processor family. Intel is also lowering income expectations by another $300M for slower Sandy Bridge adoption.
Dual-core is going to have to wait
As you might expect, the pending dual-core versions of the Sandy Bridge processors that were set to launch in about two weeks, is going to be delayed another 3-4 weeks because of the chipset issue. I know that many notebook vendors that were eager to get their latest iterations of products out in the market will be forced to wait at least another month (and also replace any produced and stored parts).
Other chipsets coming down the line (like the rumored Z68) are apparently not affected, or rather, aren't affected in their roadmap as Intel will have time to correct any problems before launch.
We want to know what your thoughts are - join in the discussion in the forums!
Profits of $3.0 billion and then chipset errors that cost about $1 billion; quite a month for Intel.
- Newegg pulls Sandy Bridge CPUs and P67 MBs while NVIDIA snickers quietly
- Legacy technology is source for Intel's chipset woes and recall