radar1200gs said:
*bangs head against brick wall*
I could be coy and say that this kind of behavior may explain some of your opinions, but I won't do it. I might recommend using a pillow the next time you feel the urge coming on, though...
NV30 was released. People own them. Whether or not it got to retail is irrelevant (I think a few 5800 non ultras did find their way into some stores btw).
NV30 was cancelled, not released. And it would not be until late August of '03 that nVidia would claim that the chip it did release through its OEMs into the retail channels, nV35, would begin hitting its yield targets. nV38 came still later in the year. Neither nV35 or nV38 was cancelled as was nV30. Here's how it worked:
(1) nVidia announces nV30 in Late '02.
(2) nVidia OEMs subsequently announce products of their own based on the nV30 reference design supplied by nVidia, and place their advance orders for the gpus (and other things) with nVidia.
(3) Prior to receiving enough nV30 gpus from nVidia to take their nV30 products to retail distribution, nVidia's AIB partners at the time began the process known as "pre-orders," which allows them to take advance orders for an announced product they will ship later, which sometimes involves taking money in advance from their customers and sometimes does not, depending on the OEM and its policies in that regard.
(4) Prior to meeting the orders nVidia's AIB partners had placed with it for nV30 gpus in quantities sufficient to ship their nV30 products into the retail channel, nVidia informs its AIB partners that nV30 is a dead duck and they will not be filling those orders for nV30 after all--with one single exception:
nVidia agreed in the cases of some of its AIB partners, like BFG, to ship them enough nV30 product to cover the pre-orders they'd received advance payment for. And that is all that ever happened, IIRC.
I can't fathom why you'd say that "Whether or not it got to retail is irrelevant," since clearly nVidia would be out of business in short order if all of its gpus followed the nV30 pattern in that respect...
The idea behind launching a gpu is to be able to sell it in quantities sufficient to meet the demand for it in the retail channels served by your AIB partners (otherwise, there's no point in making it, since your partners won't buy it if they cannot sell enough of it to turn a profit.)
Yes, Huang did say NV30 was a failure,
Which presumably is why he cancelled its production, right?
and a joint conference between nVidia and TSMC was planned to let people know why it was a failure, except that TSMC never showed for that conference...
The problem is that although JHH stated that nV30 had failed (which we both agree he did), he did not state that it was a failure because of TSMC at that time. Instead, the entire "blame the FAB game" was conducted by way of a PR campaign which caused people to *infer* that "nV30 was TSMC's fault." JHH never stated that it was. People will infer what they may, but in this case it seemed to me at the time highly unlikely that TSMC was responsible for nV30's failure, and I said as much then. I think that assumption has been proved correct, and nVidia still uses TSMC for a good portion of its business, accordingly.
As to the hypothetical "conference" you allege, from which you infer TSMC was going to stand up and say: "We suck and that's why nV30 sucks," I think TSMC would have to be insane to make such a statement, most especially if TSMC knew perfectly well it had nothing to do with nV30's failure.
The only reason NV30 was released at all was because they couldn't get NV35 into production quickly enough at IBM to replace NV30 (it taped out very soon after NV30's official launch). nVidia decided they couldn't wait any longer, whether that was a wise decision or not is a matter of opinion.
I hope you can see what a ridiculous statement this is...
What you are saying is that nVidia never planned to ship nV30, and that the entire nV30 announcement was a deliberate sham; and that the entire time nVidia was working with TSMC on nV30, through one delay after another, that its use of TSMC was also a sham, since nVidia planned all along not to use TSMC to manufacture nV30--but the *real and secret plan* nVidia had was to actually make nV35 later on at IBM instead...
Heh...
(As some have pointed out, though, it seems like it was TSMC, not IBM, which fabbed nV35, which completely erodes your notion here.)
Radar, I think you are registering one too many blips for your own good...
If nVidia had made a habit out of doing as you suggest, the company would have expired long ago from general incompetence laced with a healthy dose of insanity...
