f



SAS parallel processing #2

I'm lucky enough to have access to SAS 9.3 on an Intel Xeon X5570 with
8 cores (at least there are 8 cpu-usage boxes in task manager).  I'm
using this mostly for running statistical simulations.  I generate
1000 datasets and then run my macro on each of them and store the
results.  The macro is an alternating sequence of DATA and PROC
LIFEREG steps.  One thing I noticed that if I batch submit a job to
process all 1000 datasets sequentially, and I open up the task manager
I see the CPU usage is steady around 15%.  All 8 cores have some
activity, so I assume that SAS is automatically distributing the work
among the 8 cores.  But I still wondered why only 15% of the total
power, why not 100%?  So I cut my 1000 dataset job into 5 exclusive
jobs of 200 datasets and submit all 5 to run simultaneously.  Now the
task manager reports things the CPU is working around 90% and I can
process all 1000 datasets in about 1/4 of the time this way.  My
question is why is this necessary to go through this process of
cutting up the data and submitting 5 jobs to get full advantage of the
CPU resources?

Another question, I'm looking to buy a new laptop and I'm wondering
which processor will give me the best performance in SAS for the money
( I do not want to spend over $1500).  I thought that I should go for
the Intel=AE Core=99 i7-2675QM Processor, and get something with 8GB ram.
It seemed to me that number of cores is more important in SAS than
clock speed.  I'm also wondering if 64-bit vs 32-bit SAS will make a
difference in SAS run-times.
0
4/14/2012 1:57:12 PM
comp.soft-sys.sas 142828 articles. 3 followers. Post Follow

0 Replies
609 Views

Similar Articles

[PageSpeed] 47

Reply: