[Info-vax] Question about Alphaservers ES40 performances

[abrsvc]

On Tuesday, April 3, 2012 10:28:38 AM UTC-4, Mspec wrote:
> вторник, 3 апреля 2012 г. 12:19:01 UTC+3 пользователь Jan-Erik Soderholm написал:
> > Mspec wrote 2012-04-03 11:03:
> > > Hi, guys!
> > >
> > > I beg your help!
> > >
> > > Do you know what the performances of AlphaServers ES40 6/500, 6/667 and 6/833 in the VUPs measurements, both Int and Float?
> > >
> > > Please, help me and clarify these numbers for me!
> > >
> > > Thank you in advance!
> > 
> > A quick Google for "Alphaserver VUPS" gave for example:
> > 
> > http://www.compaq.com/cpq-alphaserver/performance/perf_by_name.html
> > http://www.compaq.com/alphaserver/performance/
> > 
> > I guess you could equaly easy have found them yourself.
> 
> Thank you, Jan-Erik!
> 
> But I need exactly the VUPs performance numbers. I've been seeking for several hours via I-net and didn't find these ones. And I haven't found any appropriate way to transform spec2000 or any other spec's numbers to VUPS Int and VUPS float. So how can I transform spec2000 to VUPS for ES40?

You have posted this request in numerous forums.  Let me summarize what i posted in an other one:

VUPs are a meaningless number for most things.  It is a comparitive measure based upon the performance of a number of workloads that ran on a VAX 11-780.  When a new VAX was marketed, the VUP rating was used with various workload descriptions to provide a relative performance level for te new machine.  Both the reported workloads and SPEC benchmarks are artifical workloads that may not represent the actual use of a particular machine.  While running the same workload on two different machines can guage the relative performance of the machines for that workload, there are other factors involved.  Specifically, the quantity of I/O for the workload.  Most VUPS ratings involved some level of I/O in the workload presented to the machine such that the relative rating included that I/O performance as well.  THe SPEC numbers are essentially raw computational throughput numbers.  Again, relative to the VAX 11-780, you can get an approximation just by scaling reported VUP values using the same ratio as with SPEC numbers for both machines.  This is a rough approximation and will not reflect the differences in caches etc.  

This is about as close as you will get to what you require outside of performing actual benchmarking of real applications.

Dan