Originally posted by: Some1ne
So is there a form to sign up for these replacements, or are they all made up? My chip seems to be fine, though if given a chance to swap it out for one that might overclock even better, free of charge, I certainly would. Especially if I could keep the old one too...why would Intel want it back anyways, "defective" as it is? Bet ebay wouldn't complain about taking it, though.
Intel should be able to give you a CPU with a different stepping, assuming you provide some sort-of proof that you're being affected by one of the flaws documented in the Errata. (Meaning, if you had a B1, you could probably get a B3 assuming that addressed the issue you were encountering).
But don't hold me to that statement! I'm going purely off of my experiences with owning Pentium CPUs in the days of the FDIV bug (which Intel honoured replacements for) and the 0xF00F bug (which I believe Intel would replace assuming you asked, but by that time most operating systems had implemented patched).
That said, I'll remind people that since the Pentium 2 (possibly earlier), the CPUs are basically just "emulators" running microcode that translates x86 instructions into whatever the manufacturer decides is correct.
You can actually update the microcode inside of the CPU. Yes, you read that right -- you can actually change the functionality/behaviour of the opcodes in the CPU by updating the code that's in it. Here's valid proof of this, running under Linux:
http://www.urbanmyth.org/microcode/
Now, that said, there's something I need to state about this:
microcode updates do not get permanently stored on the CPU. There is no "flash" chip or EEPROM or anything like that -- the updates disappear once the CPU is reset in some way (power-cycling the system is guaranteed to handle it ). You have to load them every time you want a newer revision. If you want something that's permanent, you have to get an updated processor from the manufacturer.
That said: a friend of mine took the time to read the thread here and point out one shortcoming of my statements:
Intel claims that this bug can be worked around with a BIOS upgrade. I initially stated I didn't see how this was possible. I was downright wrong (and I've no shame admitting that). My friend reminded me that CPU microcode updates can actually be done via the BIOS. Your machine will boot the BIOS, and during the POST phase, will actually spend some time applying microcode changes to work around such issues.
So, if this is indeed a serious problem people are encountering, BIOS manufacturers can provide a BIOS update that can provide a fix. That should be a relief for anyone who's worrying about it (including me). Heck, maybe this is what some motherboard manufacturers have already done since the Core 2 Duo release. Pure speculation on my part -- who knows.
So, as for my own concern -- am I still going to refuse to get a Core 2 Duo until this is fixed?
NOPE! I'm going to purchase one once the 6600s are in stock at the sites I purchase hardware from. There have been numerous people in this thread who have posted 100% success with the Core 2 Duos under heavy load when using threaded applications, and I would expect this problem to surface almost immediately under such conditions -- but it doesn't, so it must indeed be a rare sort-of thing.
Some other flaws in the Errata do also appear major (someone mentioned buggy REP STOSB/STOSW support when fast strings are enabled, for example)... but like some of the processor fanboys here on the forum mentioned, AMD has its share of issues too. I don't want to sound like I'm being anti-Intel or even pro-AMD. I'm fair about it. (Heck, I'm a lot more concerned with NB and SB chipset flaws than I am with CPU bugs).
All in all, this shouldn't stop anyone from buying a Core 2 Duo or Core 2 Extreme. But it is something to keep a close eye on... and I really do hope someone (Anand? ) can get Intel to give a statement about it, or at least get the Errata updated with some clarification.
Remember, it's a good thing that Intel is at least aware of these issues -- companies which aren't aware of problems usually deny they exist... then some smart-ass mathematician comes along and says "What's with this broken rounding going on?!", and before you know it, you're having to re-release a product you thought was "rock solid". ;-)