Vince said:
All things are never equal though Vince. In a competetive environment you've also got factors such as cost and time to market to consider - no-one in is disputing that 130nm will ulitmately be better than 150nm, but that doesn't necessarily mean its the case that its going to be the second its available - arguably, even now, 130nm is more expensive in a similarly performing part to one designed on 150nm so is it actually currently better for the consumer?
You didn't answer my previous question though - was it not a good thing that ATI brought an excellent perfoming DX9 part to the market in 2002? Did that not advance gamers experiences and move along the delopment of the industry? If you say no to that then you're living in cloud cookoo land, if you say yes to then then there is a clearcut case where 150nm was better than 130nm.
Vince, you are ignoring that there are cases where similar thing could have been exploited in the past. The case of NV20 is pertinent - it was late from when NVIDIA had initially wanted it to be to the market, and a similar situation could have occured but none of the competition was in a position to do it. Half of a large chunk of your precident is set in a competetive vacuum.
The other issue is that this apparent precident is set on relatively low complexity part, processes and shorter cycles. We are now moving into a different era, of more complicated processes and longer cycles - can you say for certain that a new precident hasn't been set? No you can't but I'm not actually saying this is a precident. Regardless of whether you think this is a one off or not, the fact of the matter is that it has occured and you can't ignore it.
What, the move from integer to full float pipes? The 3 fold increase in vertex processing, the two fold increase in pixel shading (at higher float precisions), the larger on-chip HierZ, the move to MSAA, the higher Z compression ratio, colour compression, DX9 shader functionality... need I go on?
It was the fault of the manufacturing processes of that lithography not being ready at a time NVIDIA pushed to use it. The fab warned them of this - the simple fact was that lithography was not ready at the time for that application; similar things are still happening with things such as low-k (and at other fabs as well).
Vince, you're the only one saying this - we all know what smaller proces sizes bring, and frankly your constant pigheadedness and ability of putting things in other peoples mouths is pissing me off. What you don't seem to be appreciating is that other factors also go in there as well - timing, need and familiarity (hence what cen be achieved) are also factors that need to be tipped in and swilled about.
a.) When is that likely to be, b.) how do you know ATI won't be using a similar process at the same time, c.) will they have an architecural advantage to ofset process differences?
I don't know the answer to those questions, and I'm not suggesting any of those factors will occur, but they are possabilities, just as there is a possability that NVIDIA may hook themselves into one of these process choices again and the process is late to the market again.
Don't you dare...
Again, its not always about "pushing the envolope" just in termd of new processes - ATI pushed pushed the envolope with 150, and they succeeded (up until now). As said, there is also the other factors that need to be considered as well and they are equally important. You also then have to sell that product to people, and in the PC space that isn't like the console market - it has top be produced at prices that are palletable for people to buy.
Their lithography policy does not preculed their latest high end designs utilising the current latest process if they have trialed that process elsewhere. Between now and R420 look to the products they release in that period to gain an understand of what R420 is likely to utlise.
