Perhaps it is just me, but something about Intel's Sandy Bridge chipset recall just does not add up. Nvidia may have a few words to add to the story.
I don't want to dive into conspiracy theories and assume the worst, but I just don't think we have heard everything from Intel yet and I doubt that we will get to hear much more than the company revealed yesterday. The bottom line: 5-15% of Cougar Point chipsets, which are tied to Sandy Bridge chipsets can ultimately fail to recognize connected SATA devices -- under extreme conditions and down the road, in about 2 to 3 years.
I had to scratch my head when Intel announced the issue yesterday. Intel claims it shipped roughly 8 million chipsets when it discovered the issue and that a new revision of the chipset that fixes the issue is already in the works. But then you look at the expected $700 million charge, which could turn out to be twice the (absolute) cost of the famous FDIV bug in 1993. When you break it down, this number values each chip and processor at $80 and change, which would include the cost to find the boards that carry the chip, ship them to Intel and provide replacements and labor reimbursements as well as some extra cash for damage control and lost revenues.
Is it just me or is that high for a $5-or-so chipset?
If you think about it, it may actually be a realistic number that is most likely, too low. And I would argue that the reason for the replacement is not because Intel wants to be nice to you, the consumer, but the stock market and our creative lawyers.
Almost three years ago, Nvidia revealed that it has found material defects in some of its GPUs (it turned out that up to 75 million GPUs were affected) and decided to replace the GPUs, supported by a $196 million fund. However, the company spent (FY2010 PDF file) $189.3 million as part of the warranty program in FY2009, 95.8 million in FY2010 and recorded an extra charge of about $164.4 million in "related" cost. The cost so far? $449.5 million. And Nvidia isn't done yet. Most notably, there are three related punitive class-action suits pending against Nvidia at this time. This could get rather expensive. Lesson learned. Take class-action suit lawyers seriously. A little birdie told us that the usual suspects are already in talks with the usual tech consultants to evaluate how a class action lawsuit could work against Intel and break down Sandy Bridge. Expect a suit to be pitched within two or three weeks.
So, in all common sense, Intel's Sandy Bridge chipset issue could be a minor issue as the company claims and common sense suggests that Intel should have just placed the issue under an errata, put a warranty program into place for those rare cases when the chipsets in fact fail. Even if the possible replacement of at least 400,000 boards could have turned into a logistic nightmare. Like it or not, that's just business. However, Intel would be extremely vulnerable from a legal perspective in such a scenario and taking care of the problem right now is painful PR, but it defends the company from greedy lawyers. The coming lawsuit could end up in court, but I guess it will be settled out of court so both the lawyers and Intel are happy and close the matter. But is that cost is already included in the $700 million estimate?
I spent some time with Intel on the phone yesterday and noticed that the company is focusing very much on the fact that it really isn't a big deal and that it has caught the issue early on and is taking care of the issue right now. However, the answers get somewhat thin when the company is asked about the validation process, which involves more than 3000 Intel engineers globally.
They take new silicon through a grueling testing process for several months and into environments the typical user will never see. Kinda like a new convertible that is stress-tested in the very north of Finland. If it is a minor issue, and if it is a failure of a trivial component like an SATA port, which is somewhat a carryover from a previous-generation device, why didn't Intel's validation team catch this problem? I was told by Intel that the issue was discovered by partners, when many more platforms are put under stress tests. That may be true, but it may also point, if true, to a problem with Intel's validation process. Imagine the same problem in cell phones or smart TVs. If you think $700 million is a lot, think again.
Intel may not be revealing everything it knows about this issue. Either that, or Intel is conservative in its estimates.
The good news, of course, is that it seems that Intel is taking care of the issue thoroughly and that ends up to be a good thing for the consumer. It is tough finding any Sandy Bridge boards on the market one day after the recall was announced. For example, NewEgg has pulled all boards and even Sandy Bridge CPUs from its website already.
Editor's note and correction: A previous version of this article incorrectly stated that the replacement cost of each affected chip was $880. We apologize that the extra zero accidentally slipped in.
It is because of different revisions. On A Revision the faulty transistor isn't present. However, it is present on the B revision which is the one that has the issues. This is why. That information is widely available all over the web.
Yes, I've noticed that too.
I think 700M now is way cheaper than a big lawsuit from a lot of people sown the road. Way cheaper.
Cheers!
"The scenario where Intel got 492 specific, repeatable, and very emblematic problem reports on the same day late last week that allowed the engineers to pinpoint the error in record time seems impossibly silly. Intel has some of the best silicon engineers in the world, bar none, backed by the best tools in the world, but this is too much of a stretch..."
I'm not supporting Intel here, and the article raises some questions, but:
1. Any sufficiently advanced technology is indistinguishable from magic. If Intel's engineers and their tools are that good (I have no reason to believe otherwise), then I can accept that they accomplished discovering and implementing a fix in record time. Consider the impact of satellite technology on the first phase of the Gulf War; Saddam had no clue then, and those outside Intel may not either now.
2. Who knew what, and when, may be of considerable interest, and more details may come out that point to dishonesty or other wrongdoing. As the article points out, the costs to Intel of this kind of nonsense would be huge; hopefully knowing that, they had enough sense not to have pulled any of that sort of nonsense.
We'll just have to see what happens!
It's all good though, just use your SATA 6Gbps instead of the 3 and wait to hear something from Intel regarding an RMA process.
If you read in here - http://www.anandtech.com/show/4143/the-source-of-intels-cougar-point-sata-bug on Anand's analysis (The "It’s Notta Recall" paragraph), you'll see that the cost Intel estimates comes down to more or less full motherboards - $87.50 per chipset, which kind of makes sense, even if they don't have to make a whole new motherboard, it still costs to desolder the old chipset, labour, and testing, packaging, etc.
The thing is, all the stores and the distributors also have a stock of Core i5's an i7's they can't sell. Well, they can, but nobody will buy a CPU without being able to buy a Motherboard until March or April.
So, the stores will want to receive Nehalem based CPU's for free and only pay for them once they can restart selling SB CPU's. Or they will want to return the SB CPU's and get Nehalems in exchange. Either way, Intel must account for the CPU cost part of the equation, which I think they haven't (at least publicly).
It's unclear how they are going to handle the situation, but this is not going to be easy for them in the next two months.
And even consumers are left in an awkward situation. They can't return the motherboard now, because there is no fixed motherboard available to exchange with, unless they can also return the CPU (but what if they bought it from a different store ? - how is Intel going to handle this ?), and chose to get the money back or exchange with a Nehalem based system instead.
And what about consumers that use more than 2 Sata ports ? It's actually quite easy to happen: 1 main HDD + 1 DVD RW/Blu-Ray already fill both Sata 3 ports. With more and more people buying a second HDD to store videos and photos (or a low capacity SSD to boot from, putting the HDD as the "documents disk"), you're walking into trouble already.
How does the store solve this problem, especially if you bought the CPU and motherboard from different shops ? Does the motherboard seller give you a PCIe sata add-on card for free so you can solve the problem ?
This is a PR nightmare.
Protip: Proof read next time.
1. If Sandy Bridge really wasn't ready, can AMD's lawyers convince a [non-techie] court that Intel released it early anyway just so people wouldn't buy AMD?
2. If Bulldozer is "decent," how many people left with a sour taste in their mouth by this mess might decide to give AMD a shot?
The plot thickens.
Exactly. I just swapped my HDD and DVD to the 6G ports... problem solved.