2 hours later…
4:15 PM
5 hours later…
9:13 PM
@vzn Computers can be expensive. In small group, it may be few 10k only. But for those group with clusters of few hundreds computers, it can be of million dollars including the high speed connection. Not to mention the power consumption
And for scientific research, the computational power is not the most important thing. It is the physics behind it. If the problem slightly changes slightly with an effective model, the computational time can be reduce from hundreds year to a day.
Also, for most problem, there is a minimum problem size (or the requirements on the memory or computational power). As long as you have more than it, you are fine.
And it is usually expected that the physics should not be changed even with different problem size, otherwise, it usually mean something is wrong.
@ChrisWhite The key point is that it is a consensus protocol. A common consensus is the core value. In order to reach consensus, the rules must be fixed. So technically, you can add a rule to problem such as some special type of protein folding. However, you cannot change it from time to time because it will tear of the networks. It will allow someone to trick the network, involving billion of dollar.
2 hours later…
11:18 PM
re bitcoin, it seems there could be some ecurrency protocol that can calculate arbitrary problems as part of its "proof of work" certificates... something to muse on...
a question for you: what is the physics principle that keeps them magnetized for a long time as they cool? is that analyzed in papers?
that seems to have a connection to P/NP transition point... recently was poking around looking at that a bit...
as for programming languages, there is serious scientific computation being done in almost all of them surely, but yes they have their niches...
it does appear HPC languages/tech is diverging from more mainstream languages... eg client/server, web apps etc
« first day (1253 days earlier) ← previous day next day → last day (3677 days later) »