"Toy" computers http://www.conmicro.cx/hercules

From: Raymond Moyers <rmoyers_at_nop.org>
Date: Mon Apr 29 14:47:35 2002

On Monday 29 April 2002 09:39, you wrote:
> > -----Original Message-----
> > From: Raymond Moyers [mailto:rmoyers_at_nop.org]
> >
> > A real box can total flood a bundle of gigabit fiber
> > the diameter of your right leg, its the I/O that sets
> > these monsters apart really, their CPUs are no
> > slouch, but not any faster than a common modern
> > PC chip.
>
> So, by "emulated," you didn't mean to imply that they were emulated
> in such a manner as to be replaced by the peesee -

 They don't pay those costly License fees and support contracts
 out of ignorance. todays mainframes are exteamly muscular,
 having taken advantage of the same technical advancements

 Take the old 360 for example, a very old machine, yet it had
 16 way interleaved memory, giving it thruput very respectable
 even when that ram was core.

> - that is the > misunderstanding here, I'm sure. :)

 Well, i had assumed most would know that a box that typically
 served 5000 seats or more was powerfull even as most might
 not know what it is that really sets these things apart.

 Channel I/O for example, translating to terms and concepts
 more familiar to those without OPER console time, imagine a
 "PC" where every orifice was pumped by its own dedicated
 DMA controller, where you can have 65535 of these devices.
 and fill em all up with no load on the CPU.

 Remember when the mickysoft press release parroting nattering
 nabob computer press was declaring the death of the mainframe ?

 and very humorous events like when the idiot press would read a
 product release about NT being ported to an FSIOP card, and run
 to print "NT Ported to the mainframe ! "

 A FSIOP card, File Server I/O Processor, is a PC on a card that
 plugs into the mainframe so that it can share mainframe DASD
 (disk) or have a faster channel for I/O to PC based middleware,
 it certainly isn't NT running on the mainframe in the manner
 the hapless readers of these sorry articles was led to believe.

 As for CPU power, PC's are certainly as fast per CPU in a general
 sense as a mainframe, but without the I/O capacity could never
 hope to replace a mainframe anytime soon.

 Those same press twits that reported the death of the mainframe later
 wrote articles that perhaps they would be around longer than
 expected and finnaly that IBM was selling more of them than ever
 before.

 The PC running something decent does rival the power of a
 mainframe or a supercomputer of yesterday however,
 its just that most cant tell when all that performance is bloated
 out with a utterly total garbage OS called winblows that is such
 utter garbage that at last time i checked w2k needed 64megs
 of ram for the installer to run ? how absurd !

 Compare with the size of bsd/linux/unix that will still run on
 a machine with 4megs or compare with a mainframe nucleus
 and you see that they on the other hand, have stayed small
 tight and fast. and assembler is still very mainstream
 on the mainframe where thruput of massive loads is still
 the focus.

 No, i really didnt think i would be misunderstood as saying
 it was practical or even passable beyond laughable that
 i was sugessting a PC could replace a mainframe.

 The computation power needed to emulate that hardware
 in any useable sense is not small however, on the bare metal
 the PC has caught up to those machines and surpassed
 all but the newest on a per CPU basis, the ability to get 5
 mips or so emulating 360/70/90 instructions is testament
 that the lowly PC has become very muscular in its own right.

> Anyway, I've always subscribed to the school of thought which says
> that the CPU should do only that which can't be accomplished with
> special purpose chips. It makes things much more elegant.

 Look at a mainframe stuffed full of channel controlers and you
 see very similar type of thing, and one of the reasons why big iron is
 so powerfull.

 a modern uP is the equal of a mainframe CPU today, but you get
 more out of the mainframe cpu computationally because it does
 not have to do anything else.

 Raymond
Received on Mon Apr 29 2002 - 14:47:35 BST