Much as it is derided Alan, CPU speed is an important consideration. Not the only consideration of course, but important all the same. 4Ghz on a the PowerPC-like PU core of a mainframe is - impressive. Of course, IBM has always been a little "retentive" about stuff like that, because they always like to cap the processors for sales reasons. (One of which is of course, a customer never complains if it goes a little but faster than advertised, but will scream bloody murder if it ever runs slower than the maximum advertised speed. :)
Besides which- the competition here is small machines that have boatloads of processor cycles; so much so it is not a problem at all to run three, four, or even five to ten virtual machines on them running a full PC OS. That also tells you how inefficient and poorly written most PC'Oss are, because Windows, for example, does not run noticeably faster *even when it has the entire machine*. Amazing... On the Mainframe side, that is also true, but with a twist. On the mainframe it is because the OS's are very efficient indeed, and very highly optimized to the underlying hardware (well, virtual hardware in the case of z/Arch I suppose.) That's a very big difference when you get down to the underlying core. So sure - yack up the 4Ghz- sing out about it at the top of your voice! Be sure to mention that a mainframe can actually USE all that processing power, especially running z/VM or z/OS. Something that the competition just cannot do, period. :) -Paul > -----Original Message----- > From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On > Behalf Of Alan Altmark > Sent: Sunday, October 14, 2007 11:03 AM > To: IBMVM@LISTSERV.UARK.EDU > Subject: Re: IBM's Next Generation Mainframe Processor > > On Saturday, 10/13/2007 at 06:03 EDT, "McKown, John" > <[EMAIL PROTECTED]> wrote: > > >3) The cycle time is 4GHz+ and just as with MIPS, GHz is a > > >"meaningless indicator of processor speed". > > > > True, but we will be able to shut the Intel bigots up just by saying > > that our CPU is 4 Ghz and not a paltry 3 Ghz like their's. Granted, > > meaningless, but it still will be nice. > > (sigh) I feel like saying to them, "Is your fave machine so > one-dimensional that all you can ever talk about is the frequency of > the > oscillator on the CPU chip? Describe for me, please, the business > value > of 1 MHz. I will wait.... Still waiting... Hello? Is anyone there? > Hello?" > > So we probably *don't* want to play up the 4 GHz because in 6 months, > *they'll* have 4 GHz, and then where will we be? > > I suppose we should knuckle down and define a new eHG (e-holy grail) > that > is a function of CPU speed, disk access time, bus latency, memory > latency, > number of peripherals, wind direction, amount of local disk storage, > number of tape drives, power consumption (kwh), mean age of programmers > who wrote the code you're running, and the number of cubic meters it > all > occupies based on cabinet dimensions, the air-speed of an unladen > *African* swallow (in April, of course), and total length of cables. > > Warning: If we could have such a number, it would probably be ignored > in > favor of CPU speed. > > Alan Altmark > z/VM Development > IBM Endicott
