Incidentally, the "1000x more" efficient" claim is a combination of speed and power improvements. So it may be 100x faster while using 1/10th the power, or vice versa. I haven't seen the original paper yet though, so I can't say which.
According to misterxmedia and company the Xbox one's SoC is stacked and the Xbox one was beta tested in the future, thus that means the Xbox one already has this before it was invented.
According to misterxmedia and company the Xbox one's SoC is stacked and the Xbox one was beta tested in the future, thus that means the Xbox one already has this before it was invented.
This just looks like even more parellelized computing than the current multi-core approach so barring some revolutionary breakthroughs in how to parellelise code this is far more likely to be useful for business and scientific computing than gaming. Though I could maybe see some use for physics effects if the professors are less isolated than those on say a graphics card.
Only realistic way to get such an increase in performance is to get memory latency down. (Which seems to be idea of this tech.)
L2 misses are the bane of performance.
Incidentally, the "1000x more" efficient" claim is a combination of speed and power improvements. So it may be 100x faster while using 1/10th the power, or vice versa. I haven't seen the original paper yet though, so I can't say which.
That's exactly what I noticed too. Hopefully it's not just a 5x performance bump and 200x less power draw. While that would be great for mobile, environment, etc, there are some great NP algorithms that could use a huge performance bump in order to be realistically solvable without resorting to a lot of heuristics.
That's exactly what I noticed too. Hopefully it's not just a 5x performance bump and 200x less power draw. While that would be great for mobile, environment, etc, there are some great NP algorithms that could use a huge performance bump in order to be realistically solvable without resorting to a lot of heuristics.