Date: 11/19/2001 9:48:48 PM (GMT-7)
IP Address: 188.8.131.52
In Reply to: Re: CPU lifetimes/MTBF?? posted by
I don't have the kinds of connections or the knowledge profile that MS possesses, so please do NOT quote me as an authority
Suffice to say I have placed my grubby little paws on about 25 CPUs, including older Intel 486/Pentium/PII parts, and a heavy quantity of AMD equipment including nearly a dozen K6-series parts and three Athlon units.
I have never had a CPU failure, in spite of my best efforts to the contrary -- such as a K6-2/400 that nearly lost its life in an extreme overclocking experiment.
In reference to the "mom-and-pop" computer store owner, you made the following statement:
Although not his main focus, at one point he pointedly said, "Do not buy an AMD CPU". The reason, he claimed, was because AMD chips go bad just beyond the warranty period. He said he has repeatedly observed this, so he prefers not to sell AMD systems anymore.
Allow me to pick that apart:
1. Though not extremely likely, it is possible that he, or the equipment he uses, may somehow have introduced a systematic error that resulted in CPU failures on his AMD-based systems.
2. He may be suffering from selective memory. The idea that he would "repeatedly observe" a CPU failing "just after the warranty period" seems to confirm this. CPUs do not "just fail" shortly after the warranty unless some third factor makes it occur that way. More likely, he is pre-disposed to look at things a particular way and this impacts his observation of the results -- to some extent we are all wired that way, we tend to see that for which we search.
3. His typical AMD customer profile may be playing with his theory in a rogue way. The Athlon has been popular among gamers and hardware enthusiasts, precisely the kind of people who might attempt to overclock the system or modify its hardware configuration (either of which could lead to CPU damage).
I am not rejecting his observations, I simply do not have enough data to vote one way or the other. But IMHO there are far too many "free variables" to take a statement like that at even a fraction of its face value.
Unless a market-research company has performed a CPU MTBF study (and if so, they probably want money for the privilege of your examination therof), I highly doubt you will find one. Your best bet is to talk to real-world users, where you can (a) eliminate systematic error due to the wide variety of supporting hardware in use, (b) evaluate a user's general perspective for evidence of personal preference or bias that might skew (deliberately or otherwise) their perception, and (c) evaluate the person's experiences for evidence that they were the caues of a hardware failure, and not the hardware itself.
In other words, exactly what you are doing right now