0
$\begingroup$

Parallel computing allow us to compute tasks in parallel. And if there was a system like torrent that unlike sharing files, it share computing power? For example I have a gpu and I use only 25% of its power, so the remaing 75% is used for compute tasks of others computers.

Sorry for my poor english

Improve this question
$\endgroup$
1
  • 2
    $\begingroup$ Welcome to SciComp.SE! It seems you are interested in distributed (or grid) computing such as Folding@Home (which does exactly what you are asking about; see also en.wikipedia.org/wiki/List_of_distributed_computing_projects). However, this site works best for focused question with a single detailed answer that actually helps you (see scicomp.stackexchange.com/help/dont-ask), so you should make your question much more specific (what kind of problems are you trying to solve? what kind of resources do you have available?) $\endgroup$
    – Christian Clason
    Commented Mar 24, 2015 at 15:46

1 Answer 1

Reset to default
1
$\begingroup$

The BOINC project does what you describe. Some of the most famous projects using this concept are Folding@home and SETI@home.

Improve this answer
$\endgroup$
4
  • $\begingroup$ yes, but noone use this concept to render a game for example $\endgroup$
    – Luca Punes
    Commented Mar 24, 2015 at 11:08
  • 2
    $\begingroup$ The latency would be far too high for such a critically real-time application. $\endgroup$
    – cfh
    Commented Mar 24, 2015 at 11:24
  • $\begingroup$ let's say one node would run a game, then the rasterization process (for example) would be shared by all other nodes. If there are 1000 computers that use few cpu, then I think both the time to send/recive the task and the time to compute the task would be little $\endgroup$
    – Luca Punes
    Commented Mar 24, 2015 at 13:32
  • 2
    $\begingroup$ We are talking about workloads of 1/60 seconds here if we assume that the game should run at 60 frames per second. The overhead of sending out all the geometry information over the network, rasterizing remotely, then sending the results back in, waiting until all nodes have synced up, etc, would be quite high. $\endgroup$
    – cfh
    Commented Mar 24, 2015 at 14:16

Not the answer you're looking for? Browse other questions tagged or ask your own question.