On the original topic of transistor count- nobody knows.
On a number of occasions, the number 120 million was used multiple times in inteviews and informal information pimping the card, then recently this posturing was "100+ million" which some read to mean a reduction with viable hypothesis having to do with .13 yields or other factors.
We wont really know until the card does at least a paper launch, and from there we wont really know if the engineers quoted with the 120 million number were accurate from the get-go either. So it is a moot point.
You really can't instigate a factual design reduction until a paper launch, followed by a revised paper or shelf launch occurs.
But factual aside and working completely on assumption- it wouldn't be surprising if some original design decisions were deemed unnecessary or not to impact "bang for the buck" after tape out due to poor yields. After all, if you can shave 40% cost factor off a product for 8-12% performance loss, it would only make good business sense to reduce complexity and cut your losses. This allows a better, more timely delivery and better value for the end product.
On a number of occasions, the number 120 million was used multiple times in inteviews and informal information pimping the card, then recently this posturing was "100+ million" which some read to mean a reduction with viable hypothesis having to do with .13 yields or other factors.
We wont really know until the card does at least a paper launch, and from there we wont really know if the engineers quoted with the 120 million number were accurate from the get-go either. So it is a moot point.
You really can't instigate a factual design reduction until a paper launch, followed by a revised paper or shelf launch occurs.
But factual aside and working completely on assumption- it wouldn't be surprising if some original design decisions were deemed unnecessary or not to impact "bang for the buck" after tape out due to poor yields. After all, if you can shave 40% cost factor off a product for 8-12% performance loss, it would only make good business sense to reduce complexity and cut your losses. This allows a better, more timely delivery and better value for the end product.
