I've only mentioned Muon1 in passing before, when it reached it's 20 Quadrillion particle-timestep milestone last year. however, I'm a strong supporter of Muon1, as I believe it's a massive step forward in distributing computing. Now you too can show your support by joining its facebook page.
On the page, there are weekly (or as near to weekly as I can get) updates of the progress of the various lattices (sub-projects) as well as news of updates to the project.
If you want a reason to join in, then think of it in terms of utility. For comparison, let's use distributed.net's RC5-72 project.
RC5-7 has a keyspace of around 4x1021 (4,722,366,482,869,645,213,696 to be exact) keys, and is going through them, according to their stats page, at a rate of "62 blocks/sec" (at the time of writing). A block is about 4.2 Billion keys, meaning they're doing 266 billion (or 2.6x1011) keys every second. They've been going for 8 years, and at this current rate, they still expect it to take 34,130 days (92 years) to do the whole space (although since they're searching for a single key, it won't NEED to do the whole space)
A typical Muon1 lattice by comparison is much larger. Often in the 1x10700 range (that's a 1 with SEVEN HUNDRED zeros after it. One lattice that has been running for the past year is Linac900Ext10d2. Over the past year, it has had 1.6Million of that 10700 simulations returned, and is doing fairly well, as can be seen from this results graph. In fact, it's getting close to completion (as can be observed from the previous lattices).
Clearly, a large difference in processing power, and ability, and this is the beauty of an evolutionary design method over a brute force one. This is how processing needs to be, using the results of previous tests to shape new ones, and reduce computational time. It's the direction that other projects need to move as well, like folding@home.
The only downside is, Muon1 is currently windows only.
 |
| My Stats |
No comments:
Post a Comment