Originally posted by Hellbinder
Guys.. I dont work at nvidia, and I dont have *That* kind of information. And anything I tell you is never ging to get validated in any kind of an official way. I simply had a conversation with someone who works at a another company whos identity will remain unrevealed. It was basically a long exchange regarding the true nature of the delays behind the GFFX based on obvious connections all IHV's have with TSMC. The gist of it is that in the .13u process was *NOT* all Botched up at TSMC, there was defenitly problems getting the low-K Dialectrics working towards the end of the process.
[story summation]
The Nv30 origionally taped-out in its origional form back in feb of last year. Which is why you'all around here kept getting in flame wars over the summer after every single Stockholder address. Its taped out, its not taped out, its taped out etc etc etc.. There were even *very* early revisions of the Nv30 in peoples hands before the R9700 debuted at E3 in that doom-III system.
The problem was, that for whatever reason the Nv30 was having serious Signal bleeds and other anomalies which caused various low level revamps of the design. Yes there were a few issues with the .13u process where the Nv30 was concerned but other products had already been completely finished at .13u all through last year. However the entire nature of the issue changed once the true nature of the R300 was revealed. Then the entire focus changed on getting the core speed up. Origionally the Nv30 was designed to be released at 350 and 400mhz. which was clearly not going to cut it against the R300 from a technology and performance leadership position. Back to the drawing board for a face lift, and Final revisions etc were finished around august, and the chip was finally ready for a Final Tape-out in early september. However this is where all the MAJOR issues with .13u comes into play. They hit serioius issues at that point with the implimentation of low-k. Which is what they were completely banking on for getting lower power and heat, in order to get the process running at 500mhz, the new target Clock Speed for the ULTRA. the whole Train wrecked for about 14 days. Nvidia at that point said.. screw this low-k, we can make it without it. They got the chip back in its final form witout low-k. It *almost* worked but there was some issue that required a final Metal layer respin. Which had them back and with a final working prototype board with initial drivers just before their big launch. However, the card was still only running at 400mhz at that point. It would run at 500mhz.. but it was very unstable. Thus they at this point (I believe) did one more revision to the chip they have now. They already new that they would need a giant cooling system as soon as low-k fell through. What they did not count on was the necessity of a 12 layer PCB for complete stability due to power requirements and other issues until pretty late in the game. Thus the big announcement a few weeks ago.
Which all brings us to where we are today. And why the Nv30 is in the current condition that it is. Ultimately it boils down to an Nv30 that was origionally targeted at 350-40mhz, with about 2.5x the performance of the GF4 in the *best* sinario. Which of course would have been a great product if the R300 was targeted at the GF4.
In this sense Type os perfectly correct in his statement about the early overhype. It put Nvidia in a position where they HAD to deliver or it was going to be a complete disaster, they did the only thing they could.
[/story]
Please dont flame me to hell if you think this is bunk or whatever. I understand if everyone wants to reject it. However, i was asked to elaborate on my little comment. So i did. I believe its a pretty accurate summery, even if some of teh specifics are a little off.
