[geeks] New Itanium machines from SGI

[Chris Byrne]

4-6 ghz was the estimate that John "Hannibal" stokes used based on
current materials technilogies ability to shrink their fab process, and
the scaling of speed dependant on heat dissipation etc...

Bascially anything below a certain level (I think it ws .03 but I don't
remember) and silicon cant be fabbed any smaller and still take the
voltages necessary to support the speed without serious electromigration
etc...

Neither can germanium, gadolinium, rubidium etc... Which are the other
main semiconductor materials. So unless we come up with some new
materials or a new way of fabbing circuits we're gonna have a hard limit
somewhere between 4-6ghz (some have estimated as high as 10 but most say
that's too high)

And there's already a lot of work on more efficient processors rather
than higher clocked processors. Hell theres always been a lot of work on
it. Look at the IPC of a P-4, Itanium, athlon, Itanium2, p3, US-III,
POWER-4, and MIPS. Notice the IPC count goes up as the sentence goes
along (though the POWER-4 kind cheats having two procs on a single
chip). Or a mainframe for that matter (Same kind of cheating as the
POWER-4). Lower clocked processors have never really meant lower
performance. Its always been a matter of the work the processor can do.
And by work I mean actual work not just MIPS/MFLOPS. Mainframes have
never had that great a MFLOP in comparison to high end technical
computing processors but they have always been great at getting units of
work done. 

Also we have never really used the potential of massivley threaded
applications. Most of our proceeses are still largely monolithic or at
least only partially threaded. With better multithreading and better SMP
implementations we could really be doing a lot more with the processers
we have. 

Chris Byrne 

> -----Original Message-----
> From: geeks-bounces at sunhelp.org 
> [mailto:geeks-bounces at sunhelp.org] On Behalf Of Mike Meredith
> Sent: Wednesday, January 08, 2003 12:46
> To: The Geeks List
> Subject: Re: [geeks] New Itanium machines from SGI
> 
> 
> On Wednesday 08 January 2003 6:27 pm, Dave McGuire wrote:
> > When process technology hits the laws-of-physics-imposed barrier of
> > ~4GHz or so (by the last estimates I read) who's gonna be faster?
> 
> MIPS of course.
> 
> I wasn't aware of the ~4GHz barrier (I'm not a h/w person). 
> It's going 
> to be interesting to see what happens when people have to do 
> something 
> other than just turn the handle faster to get more performance.