Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Hmm, to a degree Nvidia are actually doing some very interesting things with Tegra, but ultimately when you are late with everything, using too much power and providing more power than any application needs in the target segment...... you're destined to fail. If they did a bog standard dual core A15 and were first out, it would have cost significantly less in R&D, lower cost chips to sell in higher volume and worth more money.
Charlie has often praised Nvidia engineers and mercilessly mocked the management, much the same with Intel. I think he has said, and I've always thought Intel could without question make fantastic drivers for their gpu's, they can go to Nvidia, or AMD and buy the entire driver team with higher wages, they just have no interest. Atom could have been on 32nm years earlier at far lower power and higher performance but they just decided not to, making Atom literally useless in the market they wanted it in.
Low margins doesn't mean no profits, and as you say they see discrete GPU all but disappearing for them which is very likely in the not too distant future, low end market, by FAR the biggest is shrinking stupidly quickly and with Intel actually putting in marginally more driver effort, things like quicksync and Intel/AMD moving into the future with far more intergrated IGP acceleration of basic apps, well that shrinking of the low end discrete will accelerate, couple that with losing out in all the consoles and Nvidia is screwed.
Essentially low margin anything is better than no margin nothing
Nvidia really need to stop being stupid about it though, a 5 core 40nm chip months before everyone else gets their a15 28nm chips ready that will spank them in every useful metric in the targetted markets is just utter utter madness. They also delayed their 28nm chips to "force" this behemoth stupid chip into the market, when the exact opposite was required, dump the quad core, bring the dual core 28nm forward as much as possible.
Charlie has often praised Nvidia engineers and mercilessly mocked the management, much the same with Intel. I think he has said, and I've always thought Intel could without question make fantastic drivers for their gpu's, they can go to Nvidia, or AMD and buy the entire driver team with higher wages, they just have no interest. Atom could have been on 32nm years earlier at far lower power and higher performance but they just decided not to, making Atom literally useless in the market they wanted it in.
Low margins doesn't mean no profits, and as you say they see discrete GPU all but disappearing for them which is very likely in the not too distant future, low end market, by FAR the biggest is shrinking stupidly quickly and with Intel actually putting in marginally more driver effort, things like quicksync and Intel/AMD moving into the future with far more intergrated IGP acceleration of basic apps, well that shrinking of the low end discrete will accelerate, couple that with losing out in all the consoles and Nvidia is screwed.
Essentially low margin anything is better than no margin nothing

Nvidia really need to stop being stupid about it though, a 5 core 40nm chip months before everyone else gets their a15 28nm chips ready that will spank them in every useful metric in the targetted markets is just utter utter madness. They also delayed their 28nm chips to "force" this behemoth stupid chip into the market, when the exact opposite was required, dump the quad core, bring the dual core 28nm forward as much as possible.