jgreco
Resident Grinch
- Joined
- May 29, 2011
- Messages
- 18,681
Yeah. The important part to note there is where you can reach a solid conclusion that something isn't possible today and also won't be possible in the future. In some cases, it might be worth taking a look at potentially feasible technologies that are "way out" there and pouring money into them in the hopes that they mature. I think that's a better example of where ARM is at the moment, because it's theoretically feasible to make something better than what Intel offers, but that the world isn't quite ready for.
Some of us spent a lot of money on early SMP gear where the early naive kernel and software designs weren't able to take good advantage of the parallelism offered by two CPU's, so adding a second CPU maybe only got you to 1.5-1.6x the speed of a single CPU. Or the Opteron with its (at the time) unique architecture. A lot of these plays have since been shown to be the correct way to go, but even today we have software that isn't capable of taking advantage of the parallelism in multiple CPU cores.