@ JoannaK
Your first example involves an unusable computer.
Your second example doesn't specify what the resulting problem would be for the end user, but actually putting the *wrong* part on a piece of equipment does suggest that it will go very much up the creek at one point or another.
And PS: I said "in my experience". I didn't say I had the entire history of motherboard recalls memorised.
How can my "AFAIK" be be "bogus"? Bogus implies I'm trying mislead people.
And even then.. Eyetech has not shown responsibility HW manufacturer (as they claim to be) should. They have been sweeping these issues under the rug over two years now.
Eyetech isn't a motherboard manufacturer. If anything, MAI should have admitted the problem, and you're right, it should have been admitted sooner. AFAIK about the problem, one of the standard tests I run on a new motherboard would have picked that up, a large network transfer to an IDE disk. It's pretty poor that it wasn't picked up. But if my understanding of the problem is accurate, Eyetech should have picked up the problem when they were putting the board through its paces for the first time before making a volume purchase.
And btw, you seem to be trying to attack *me* here. I'm not defending Eyetech's behaviour, I'm just clueing people in on how companies in the real world typically work, and no, it's not with the highest of moral standards.
An example of what is apparently considered pretty standard behaviour - a few years ago, Intel used to sell (and carried on selling until the motherboard was superseded), the Intel D845(can't remember which type, might have been G). Now this model of motherboard had a similar fault, a glaring one that should have been picked up in testing - If the PCI bandwidth usage (including AGP) went over 90MB/sec, the machine crashed. Considering a Windows machine sitting idle will manage to take quite a bit of that through driving the graphics of a high-res display, that's not very difficult. In fact, the same test, a large network transfer over the network, showed up the issue. Intel had *actually* known about the issue for at least a year, it had been confirmed way before I noticed it, but that didn't stop them just (and only) filing a reference to it in the erratum for the motherboard in the developer docs of their website. Customer-friendly, huh?
At the end of the day, the manufacturer decides *could* the product do what it is supposed to be able to, by the average customer's usage of it? Like if the secondary IDE channel can't do DMA to save its life, the average customer can still burn a CD in PIO. It totally sucks performance-wise, but they still can. That's the sort of day-to-day screwing that goes on.