4

I have to calculate a huge differential equation. With my laptop, it's going to be computed for several days.

Is there a free (I need just for 3 days) fast server for scientific calculations? My calculations are not parallelized, and in fact, are computed on 1 core.

I tried Google free server, however, looks like it's not faster than my laptop.

Anton Menshov
  • 8,672
  • 7
  • 38
  • 94
Leeloo
  • 143
  • 2
  • 3
    Unless your laptop is really old, it would be difficult to find a free service that is faster. – nicoguaro Jun 10 '19 at 12:11
  • Which citizenship do you have? The US, for instance, has compute resources freely available for researchers. – Richard Jun 10 '19 at 15:09
  • If you provide more details it might be easier to help you. What equation are you solving? What method are you using? What does "huge" mean, how many degrees of freedom? – nicoguaro Jun 10 '19 at 17:21

1 Answers1

2

You will certainly benefit from parallelization of your code (or using/switching libraries that offer a native parallelization). So it might be worth looking into parallelization depending on your priorities in terms of time and possibility of new upcoming simulations.

In terms of external computing resources, I would say, that you are looking for a very rare unicorn here. It would be hard to satisfy all three criteria and be in the centre of the Venn diagram.

Here, by "Universal", I mean no limitation to one particular technology, no strict time limitations (trial period), and, the most important one, citizenship & affiliation of the user.

enter image description here

Example services:

  • Fast and universal: many now available cloud-based services: Amazon EC2, IBM cloud, and several others. The downside is that the cost is not very cheap and certainly not free.
  • Free and universal: Free tier of Amazon AWS, Free tier of IBM cloud, which obviously would have a lot of limitations on the available computational resources and might not be faster than your laptop.
  • Fast and free: XSEDE would provide US-based institutions with computational resources. Limitations: you have to have a collaborator who has access to those resources. Similar things would exist for other regions, but usually, they would be limited to researchers having affiliation with a limited set of institutions.
  • Free: (not necessarily fast or unlimited) free-trials to new services. Intel Xeon Phi Cluster from Fujitsu, limited to Xeon Phi's.
  • Fast: your own hardware. It may be the cheapest or the most expensive in terms of price, depending on the circumstances.
  • Fast, Free, and Universal: don't know anything that even comes close to be in this group according to my definitions.

Relevant posts:

Anton Menshov
  • 8,672
  • 7
  • 38
  • 94