|
|
Windows vs. Mac: Mac 3 times slower |
|
Message/Author |
|
|
Hello there everyone, As the title suggests, I've found this happening with several analysis (the latest of which was an LPA using mixture modeling with 3 classes). Mplus version 8.2 on both machines. The exact input file runs almost 3 times faster on a Windows 10 machine (Dell Latitude), with a 2.3GHZ i5 processor processor and 8 gigs of DDR3 RAM. This is compared to a Mac Pro 2013 (cylinder) with a 3.7 GHz Xeon processor and 32 gigs of DDR3 RAM. The Mac Pro has more L1, L2 and L3 cache as well. Both machines have SSD drives. My question is, why would this be? I would expect the Mac Pro to be much much faster, but the complete time for the script below is approx 16 mins, whereas the Dell is approx 5 mins. This occurs on multiple runs, after restarts, etc. I have also changed the "processors" option to both 4 (cores) and 8 (logical) on the Mac pro, to no avail. Any thought would be appreciated, as I am in the process of purchasing a new MPlus-heavy machine, and would prefer to stay with Macs (all my other machines are), but this is a serious penalty. Thanks in advance. |
|
|
INP file: TITLE: Profiles 3 Classes DATA: file is 'LPA2.csv'; VARIABLE: names = ID SES age a b; auxiliary = SES age; !Includes ID in new file IDVARIABLE = ID; usevar = a b; missing are all (-999); classes = c(3); !number of latent classes. ANALYSIS: type = mixture; processors = 4; starts = 500 50; stiterations = 50; lrtbootstrap = 500; lrtstarts = 50 20 50 20; !optseed = 813779 for 3-class; Set afterward PLOT: type = plot3; series = a(1) b(2); OUTPUT: tech14; SAVEDATA: file = 'Test Windows.dat'; save = cProbabilities; format = free; |
|
|
We usually don't see significant timing differences between the Mac and Windwos versions. Please send your Mac and Windows outputs including tech1 and tech8 to Support. |
|
|
Also send the data if that is possible so we can check on our computers. |
|
|
Hello Dr. Muthen: files sent. Please let me know if there are any updates. Thanks. |
|
Back to top |
|
|