[rescue] cray sx-6

[Francisco Javier Mesa-Martinez]

On Tue, 9 Oct 2001, Joshua D Boyd wrote:

> On Tue, Oct 09, 2001 at 09:52:19AM -0700, Francisco Javier Mesa-Martinez wrote:
> > Specially when there is only 1 American vector company vs 3 Japanese
> > vector manufacturers..... :-)
> 
> Who are the other Japanese vector makers?  Also, any japanese machines ala
> the MTA2?

You have Fujitsu (VPP/VPX), and Hitachi (SR). And then a bunch of
universities doing a lot of application specific supercomputing (ala
GRAPE).

The short question to the MTA2, would you do a sequel to a movie that was
a flop? :-) It is a great idea in paper, but they seem to get no where.
Although I think we need companies like TERA that try new approaches,
certainly it will never be as exciting as the mid/late 80s in the area of
computer architecture.... but....
  
> > I still find it ironic that Cray (well it really is TERA anyways) is now
> > selling SX's which were NECs response to American supers.... 
> 
> Me to.  Especially since they are also pumping loads of cash into the much
> sexier SV2s.

Well the SV2s are being subsidiced big time by the US government. W/o it
the US has no modern vector technology. Although the SX-5/6 are plenty
sexy, specially if you take into account that a single SX-5/6 processor
can do 8GFLOPS, whereas the SV1 needs to cluster 4 processors to reach the
same 8GFPS. And you can get a SX5 now, SV2 is pretty much vapour right
now.

One funny thing is that most of the new machines are CMOS, and I have a
ton of old papers in my drawer that claimed that CMOS was dead, and why
more research was needed on GaAs, ECL, opto computing, etc.... I guess
this goes in line with all those people who said that Unix was dead, that
NT was the future, or that the mainframe was dead and NT could do the same
on cheaper infraestructure. So maybe, just maybe this means that Microsoft
will go the way of the GaAs, I can only dream.... can I? :-)

Cheers.