Message Boards Message Boards

0
|
4676 Views
|
4 Replies
|
0 Total Likes
View groups...
Share
Share this post:

How do platform, OS, language, and dollars affect computational speed?

Posted 10 years ago

I am writing in Mathematica on a laptop with an Intel i7 and 16 gigs of RAM. What happens (guessing in general) to the speed of my programs (genomics) if I:

1) Throw $20,000 at the problem and go to a Unix workstation?

2) Run in compiled c rather than Mathematica? (That would make me unhappy!)

For those of you who have been through this yourself, what did you do?

I was thinking that the system would need to be sold (used) about every 2 years with purchase of a new computer to stay at a "sweet spot" on price and performance. Are there companies that will lease with that in mind?

Thanks, Doug

POSTED BY: Douglas Youvan
4 Replies
Posted 10 years ago

Other Douglas, I haven't used those capabilities but no doubt it can be done.

http://www.wolfram.com/products/mathematica/analysis/content/GridComputing.html

If the grid is too much you could manually divide the work among inexpensive computers since it's naturally parallel work.

Have fun, Doug

POSTED BY: Douglas Kubler

My mother watched https://en.wikipedia.org/wiki/20,000_Leagues_Under_the_Sea_(1954_film) the night before I was born! (Kirk)

Some genomics programs written from scratch and running on a batch of $500 computers might make for a good SBIR, unless the license from Wolfram is huge.

Doug Kansas

POSTED BY: Douglas Youvan

Douglas,

I think you are correct to say to stay with what I have and use Parallelize[]. My DNA work linearly progresses through sequence, so it can be divided by ranges of sequence number.

With appropriate licensing, do you know if Parallelize[] will run over a LAN with a bank of several other inexpensive computers?

Thanks, Doug

POSTED BY: Douglas Youvan
Posted 10 years ago

What is the value of extra speed? I would profile the code and find the fraction that uses 90% of the cpu. Can your task be parallelized? It's instructive to Parallelize[] and see the elapsed time drop (and watch all 4 cores get busy). That's just with the Home Edition.

I'd throw money at improving what I had. Are you thinking of a fully supported system or a Linux offshoot? I see it as potentially taking more personal time than the processing time you want to save.

POSTED BY: Douglas Kubler
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract