From: norm.raphael@metso.com Sent: Tuesday, October 29, 2002 9:37 AM To: Info-VAX@Mvb.Saic.Com Subject: Re: If you want a VUP number for your alpha, this was submitted on Encompasserve by Jack Harvey on 7-SEP-1999. - - - - - - - - - - $ !VUPOMETER.COM - super simple VUP meter $ time_start = f$cvtime(,,"time") $ set noon $ set process/prio=12 $ n = 0 $loop: $ n = n + 1 $ if n .lt. 10000 then goto loop $ elapsed_time = f$cvtime("-''time_start'",,"time") $ mm = f$element( 1, ":", elapsed_time ) $ hss = f$element( 2, ":", elapsed_time ) - "." !hundredths of a second $ hss = (mm * 6000) + 'hss' $ vups = 29300/hss $ write sys$output "VUPs = ''vups' Elapsed time = ''elapsed_time'" - - - - - - - - - - He claimed about 20% acccuracy then, but it _is_ a metric. He used it mostly to compare performance on the same machine at different times, if he thought it was running slow. JF Mezei on 10/28/2002 08:52:56 PM Please respond to JF Mezei To: Info-VAX@Mvb.Saic.Com cc: Subject: Re: Paul Sture wrote: > The question is roughly "How many MIPS does your Alpha app need, so > that we can price that up in IBM terms?". You need to find one commeon metric between *an* IBM system and you. If for instance, your Alpha system does 15 tps, and IBM has a model with a published tps value of 25, and IBM also published MIPS numbers for that same model, then your could faily easily deduct how many "IBM MIPS" your Alpha has. Another possibility is to find some old VAX-ALPHA performance comparisons, at which point you may be able to get a VUP number for your current alpha and since VUPS are essentially a MIPS, you could give that to IBM. However, one has to be careful about the application itself. Will its IBM equivalent perform as efficiently or more efficeintly than the VMS version ?