Originally posted by: nitromullet
Originally posted by: apoppin
My point was replying to
AMD is about 2x better in this area. Unbelievable.
there is no "better" .. just different .. it costs Nvidia a lot of die size for CUDA
- since Jensen mentioned "15 years", i expect Nvidia is looking way beyond 3D gaming for their future
interesting indeed
Well, die size is somewhat proportional to power consumption, so I would consider it relevant to consumers. It's also no secret that getting more performance per nm^2 is "better".
Obviously, based on my rig in sig, this wasn't a deal breaker for me, but I did invest in a new PSU prior to going SLI with two GTX 280s.
not really
Nvidia has done some pretty good work in getting the power consumption of Tesla way down at idle compared to their previous architecture - for most of the time a consumer has their PC on.
Again ,, die size Depends on your goals
i am guessint Nvidia intends to be leading this kind of PC graphic market for 15 more years. That is what Jensen stated in his recent interview. Nvidia is concluding evidently that the market for PC gaming graphics is limited. i guess by then they expect the pro market and CUDA to largely replace the PC graphics gaming market for them. i think they can see the writing on the wall. They are getting solidly behind CUDA and have a new expanded vision to lead the new 3D market into many other consumer and pro applications besides gaming.
The GPU that most excels in performance per nm^2 is 4850. The 4850 is harder for Nvidia to counter because it also has such excellent performance/watt. If performance/mm^2 largely determines a company's gross margins and profits, then performance/watt determines a company's ability to put out a competitive product.
Nvidia does have the single best performing CPU, the GTX280 - i have one also. And i expect to try SLi with it eventually - after i explore PhysX.
However, a lot of that die size is reserved for *other* than gaming.
Nvidia's future