On Tuesday, 10/10/2006 at 05:20 EST, "Ward, Mike S" <[EMAIL PROTECTED]> wrote: > The thing that gets me is the MIPS. You can go from a machine that has a > single processor at 100 mips. To a processor that has 4 cpus at 100 mips > each and they call it 400 mips. Depending on you workload type that may > not buy you anything and you are still chugging along with just more > mips. What I would rather see is the actual cycle rate of the processor > go up. Give me the mips 400 but with two processors at 200 mips each, or > even 1 400 mip processor. This mips rate is underrated unless you know > the true meaning of mips.
You are hitting on exactly why we don't publish MIPS any more. It simply leads you down the proverbial garden path. You notice, too, we don't make a big deal about the cycle rate (GHz), either. Both are pretty much irrelevant in estimating the affect on your workload. And each generation of processors changes the mix of silicon, microcode, and millicode so you can't really depend on knowing how many cycles an instruction takes. Add pipelining into the mix and it gets worse. The best measurement of performance is the one you use to determine how much value you are getting from the box. You bought it for a reason. Are you getting the number of transactions per second from YOUR applications that you need to make the box worth the investment? Are you getting the qualities of service you require? Is it helping you meet your IT spending goals? Alan Altmark z/VM Development IBM Endicott ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
