Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Hua Peng <hpeng@stata.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Stata MP performance |
Date | Mon, 26 Jul 2010 09:29:42 -0500 |
Alfonso Miranda <A.Miranda@ioe.ac.uk> has an analysis which is faster when run in Stata/SE on a 2.8GHz Mac Pro as compared to Stata/MP8 on a 2.66GHz Windows 2008. Alfonso should contact Stata technical support at tech-support@stata.com with the program and the dataset since the speed differences on two machines may very well depend on the nature of the program. For example, if the main task of the program is to merge datasets, then disk I/O speed becomes the dominating factor. Of course, this is pure speculation without seeing Miranda's program and dataset. I will report back to the list for any findings. Hua --hpeng@stata.com On 7/26/2010 8:10 AM, Alfonso Miranda wrote:
Dear All, I just notice something strange about Stata/MP performance. I have access to two different systems: SYSTEM 1 (Desktop) - Hardware: Mac Pro computer with 2 Quad-Core Intel Xeon processors (2.8 GHz, 12 MB L2 Cache per processor, 1.6 GHz Bus Speed), and 4 GB in memory. - Operating system: Mac OS X 10.5.8 (Leopard). - Stata Stata/SE 11.1 for Mac (64-bit Intel). SYSTEM 2 (MP Server) - Hardware: Dell PowerEdge R900 with 2 x 6 core Xeon X7460 (2.6GHz, 16MB L3 cache, 1066MHz FSB, 130W TDP) 64GB memory ((16x4GB Dual Rank DIMMs) 667MHz FDB) The OS and page file are running on RAID1 2x300GB SAS 15k 3.5" disks run by a PERC 6/I internal RAID controller. The data (F and G drives) are sitting on an MD3000 external SAS array, which contains a RAID1 implementation of 4 x 1TB SATA II 7.2K 3.5 inch HD - Operating system: Windows 2008 (NOT R2), including all latest critical security patches - Stata / 8-core MP for Windows (64bit x86-64) I put a process in my Mac desktop and it takes 244.3750 seconds to conclude (i.e., around 4min). Running the same process (same programme + same data) in my MP server takes 717.7790 seconds to conclude (I.e., around 12min). In other words, Stata/MP with 8 processors on SRC1 is at the moment 3 times slower than my Stata/SE in my Mac desktop, which I understand can only exploit the resources of a single core. This is strange indeed. Have you experience something similar? Any comments? Best wishes, Alfonso. * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/
* * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/