optimal software

announcements (new versions, changes, bugs, installation problems...) related to program MARK

optimal software

Postby andy » Wed Sep 29, 2004 5:28 am

I have a large small mammal data set to analyse using MARK and am finding some of the models take up to 15+hours to resolve. I am running a pc with 256 ram and 1.5 ghtz processor.

Can anyone tell me whether MARK is able to run faster with an upgraded RAM, to, say, 1024? or a faster processor?
andy
 
Posts: 2
Joined: Wed Sep 29, 2004 5:23 am
Location: UK

Re: optimal software

Postby cooch » Wed Sep 29, 2004 7:37 am

andy wrote:I have a large small mammal data set to analyse using MARK and am finding some of the models take up to 15+hours to resolve. I am running a pc with 256 ram and 1.5 ghtz processor.

Can anyone tell me whether MARK is able to run faster with an upgraded RAM, to, say, 1024? or a faster processor?



Many MARK models, especially those involving a large number of occasions, and multiple parameters (e.g., multi-state models) can take a long time to run - this can also be exacerbated if you're running with individual covariates. This is especially true on older, slower hardware. And, the system you describe fits into that category, both with respect to RAM, and CPU speed.

Personally, I generally suggest nothing less than a 2.6 GHz machine, and 1 GB RAM. Because the latest Intel chipset is 3.6 GHz, the basic rule of thumb is they virtually start giving things away for older chips that are <=0.75 GHz away from their fastest. Moreoever, RAM is (currently cheap). You could pick up the minmal machine (sans monitor and other such things) for $<650 US. You can go really high end for ~$1200 for a 3.6 GHz machine with several GB of RAM.

But, be advised, with complex jobs, lots of parameters, some model fitting can take a long time even on super hot-rod machines (I have some runs that take many hours each on a very hot machine - days to complete the entire model set).

If the length of time it takes to run an analysis is prohibitive, it gets worse - wait until everyone goes Bayesian, and we use MCMC to get all our answers. :wink:

So, minimally, I'd say you need to get yourself a new machine. I realize that this is not always practical. Dropping a new CPU into an existing machine is often limited by the type of mother board in the system. Adding more RAM will give you only marginal gains for MARK.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: optimal software

Postby andy » Wed Sep 29, 2004 9:24 am

thanks for that. It looks like a new machine is the answer as the motherboard is presently at full capacity. Good, time to talk to my supervisor..

cheers again

Andy
andy
 
Posts: 2
Joined: Wed Sep 29, 2004 5:23 am
Location: UK


Return to software problems/news

Who is online

Users browsing this forum: No registered users and 13 guests