3 ms·
> Moore's law is gone For CPUs, yes. For GPUs, not as far as I am aware. And rendering is the traditional GPU task.
by majewsky 9y ago
> Moore's law is gone
For CPUs, yes. For GPUs, not as far as I am aware. And rendering is the traditional GPU task.
- ekianjo 9y agono he is right. Moore's Law in its original meaning is gone, and with GPU we are just shifting goal posts to pretend it still applies.
- kbutler 9y agoMoore's law in its original 1965 meaning was an observation that transistor counts doubled every year and a projection that this would continue for the next decade. Moore revised the projection downward in 1975 to every two years, which remarkably has held into the 2010s. But it is transistor count, not necessarily performance. A coworker calculated that performance was doubling in 18 months. https://www.kth.se/social/upload/507d1d3af276540519000002/Moore's%20law.pdf https://www.kth.se/social/upload/507d1d3af276540519000002/Mo... Transistor counts continue an exponential growth curve, particularly in GPUs. It may be time for another downward revision on the rate (2.5 years? 3 years?), but we aren't getting as much single-threaded performance out of the increased transistor count. This provides increasing performance difference between the highly parallel GPU architecture and the more limited parallelism of traditional CPUs.
- ekianjo 9y agoWasn't the Moore law about transistor count PER surface unit? If that's the case, then even GPU are way below the predictions now, because chips have certainly more transistors, but their surface is also increasing.
- kbutler 9y agoComponents per integrated circuit. The link above quotes Moore's 1965 statement.