The interesting thing about the alleged flaw in Intel's solid state disc drives is not how they fix the flaw, but how they react to it.
Charlie Sporck, legendary CEO of National Semiconductor, and a close personal friend of the Intel co-founders, wrote in his 2001 book SPINOFF that Intel had a cultural flaw in that it could never admit to losing at anything.
"Intel, in my view, has in some of its management a character flaw that any loss can't be attributed to them, it must therefore be the result of some extraneous circumstances", writes Sporck, "to me, that's a character flaw, That's not intellectually honest. There is a feeling of arrogance to some extent at Intel. It was manifest in that minor Pentium bug."
With the Pentium flaw, of course, Intel initially arrogated to itself the decision about who should receive a replacement to a flawed Pentium, and who should not, depending on the nature of their of usage.
Now, with the SSD flaw, Intel is in denial. The flaw relates to an algorithm Intel uses to spread the data evenly across a flash memory array so as to evenly spread the wear on the memory. This spreading of data is alleged to cause excessive fragmentation which degrades the performance of the SSD.
In view of its past history, Intel might have done better to publicly thank the people who picked up the flaw at PCPerpsective and say they'd do everything they could to investigate it.
Instead Intel said it's own tests didn't show a problem but, Intel being Intel, had to add: "In our estimation, the synthetic workloads they (PCPerspective) use to stress the drive are not reflective of real world use. Similarly, the benchmarks they used to evaluate performance do not represent what a PC user experiences."
After the experience of the Pentium flaw: "We embarked on a whole new way of doing business,"wrote Intel co-founder Andy Grove in his book: Only the Paranoid Survive.
So the interesting question which the SSD flaw will reveals is: Has Intel reverted back to type?
Is Intel in its Pre-Pentium Flaw mode of arrogant denial? Or in its Post-Pentium Flaw mode of acknowledging error?
From what I understand, those Intel SSDs use a wear levelling algorithm that is very different from the other SSDs. Otherwise, free space consolidation at the OS level usually helps to restore the lost performance of SSDs and promotes sequential writes. This is what Diskeeper's Hyperfast for SSDs does.
Link at PCPerspective forums.
http://forums.pcper.com/showthread.php?p=4351675
Thanks Pringles, that's an interesting link
Let me explain how "honesty" works in the American Soviet Socialist Republic these days, Comrade.
You're honest.
You get sued. You get sued a LOT.
You find yourself in court with greedy, professionally lying lawyers, judges who couldn't judge between an ASCII code table from an Italian restaurant tablecloth, and, if its a jury trial, de facto jury tampering called "voir dire" (which, if I'm not mistaken, is Latin for 'jury tampering.' If you do end up with a jury of your peers, you will find they are not exactly swimming in the deep end of the technically-savvy gene pool.
So - you don't get to BE honest. Not in the Youse-Ess-Aye, folks.
And not in the EU, either.
Fix the legal system, and you'll get honesty. Until then, there is no incentive and quite a few disincentives.
For the record, I was on the team that found the Intel C compiler mistake that caused one of the SPEC benchmarks to be outrageously good (I was working for Motorola at the time, mid-90's). We politely told Intel. They investigated, and egads! We were right, thank you for pointing this out, we'll fix our compiler. No worries?
WRONG: Intel got sued in California by some wingnut who said he bought Intel PC's based on SPEC scores, and Intel lied, and he was screwed, blah blah. 12 jurors agreed, and Intel had to pay up. A lot.
We at Motorola were horrified, because there but for the Grace of God go any of us (compiler bugs being rather common back then on superscalar architectures). We played by the rules, and so did Intel, and Intel got screwed doing so.
By the way, the part of flash memory that deals with this is usually called the Flash Translation Layer (FTL), and the algorithm(s) are called the Wear Levelling algorithms. There is a lot of research going into this very interesting field, and we should encourage research and honest attempts at doing it better, since solid state drives are coming and have the potential for some interesting new products.
How do I know that? I did some work for (R.I.P.) Spansion here in Austin on NAND flash memory software.
That's fascinating stuff, Alan, with some excellent points:
On the Intel flash point I'd say why didn't Intel say publicly 'thank you we'll do what we can to fix it' instead of being publicly arsey about it?
On the Qualcomm point if they had such great inventions why did they lose practically every IP case they brought as well as having their lawyers reported for misconduct?
Ah, good points, Dave.
On Intel Flash .... I have no doubt that culturally it is difficult for Intel to admit a mistake - but my point was that there are huge disincentives to do so brought about by a broken American legal system, and the fact that when they have done that in the past, it has bit them in the ... arse, is it? Yes, arse.
Secondly, Qualcomm must have thought they had great inventions and must have wanted to try to protect them. They are, after all, a leading IP licensing/licencing company. The fact that the courts have not agreed with them is neither here nor there; as I said, court is where one litigates such things. I have no explanation for behaviour of any attorney!
I honestly do not think that Qualcomm has been trying to establish a "monopoly" except maybe in CDMA, which they did invent, but rather was seeking to rightly defend their turf.
Just my 0x02 cents worth.
Cheers,
Alan
Thanks Alan for a generous reply. I will adopt your view from now on that Qualcomm were not really intent on an Intel-style monopoly. I think your view is saner than mine.