During the AMD Evergreen launch event in Alameda, CA – nVidia hosted a traditional annual series of meetings among its executives. "Everybody who’s anybody" flew in to Santa Clara and it was quite interesting to see a lot of European and Asian employees coming into town.
The topic of those two weeks was the reform of the company and newly-found focus on bringing a unified computing platform to every business unit. Long story short, the Fermi architecture is the bid for the future of the company and during 2010 and 2011; Fermi-based silicon will be coming to Tesla, GeForce, Quadro and Tegra business units. Before nVidia PR’s department gets another heart-attack, we have to tell them to get a life, since we disclosed those details many moons ago. Just like was the case of GT300 specifications. Pardon, according to information we had on the NV70 architecture diagram. Fermi? Furbie? Die hards use only two letters and two numbers.
Getting back on topic, nVidia senior management came to conclusion that the company is going again through the NV30 phase. Now, even though NV30 silicon is the topic that PR folk don’t like to discuss, the real old-school engineers will tell you that the behavior exhibited by Jen-Hsun and senior management during NV30 fiasco was the cornerstone for latter success. Same people told us they feel the same thing happening right now and think that’s good for the future of the company.
According to nVidia employees, this dollar was rated well in 2002 and now is back…
So, what happened back in 2002? NV30 was hot and late [GT300 is just late], and nV management did the same – instead of firing people, they tightened the belt and with $1/year salary, rode the storm. At the same time, engineers paid back the trust by evolving the under-powered NV30 architecture into a good NV40  and NV45  and NV47 [GeForce 7900 and RSX, PlayStation 3 GPU], thus leaving enough time for creation of NV50 .
Jen-Hsun and most of members from this page are now taking a $1/year salary, just like Steve Jobs does to this very day. This is also the reason why nVidia has such fanatical employees – instead of massive firings – the company execs are going to tight the belt and ride the storm. There is no doubt in our mind that "Furbie" i.e. Fermi architecture is much more stable and should have enough horsepower, with 512 cores, 128 TMUs, 48 ROPs and the like [did we just disclose additional, GeForce/Quadro details? Yep, we did].
Unlike Hector Ruiz and AMD management, who took more than a quarter of a billion US dollars in bonuses, resulting in AMD being forced to fire around 2,000 employees during 2008 alone. The consequences of last year’s actions are still felt inside the company [PR doesn’t have enough people, OEMs are complaining about zero-level support etc.], but the pre-cull era at AMD resulted in creation of today’s winning Evergreen architecture.
The biggest winner in this situation of course, could be AMD – the company has products out the door and two of their competitors are in doldrums – with Larrabee and Fermi sharing the similar fate of being late [at the moment]. Then again, if nVidia repeats NV30 situation with GT300-based follow-ups such as NV40, NV45 and NV4 were – both of their competitors will have a complicated situation, since Fermi is going to end up everywhere.