jab@romeo.cs.duke.edu (John A. Board) (04/12/89)
I read it in the _Durham Morning Herald_ so it must be true.... NEC has apparently announced the "SX-X" series with peak vector speeds of 20 Gflop. Story says NEC plans shipments for July-September 1990, head-to-head with the early Cray-3 shipments. Does anyone know anything about this beast? Earlier NEC supers (in the 1-3 Gflop class) have been based around multiple identical vector functional units sharing a common vector register set. Have they gone to multiple independent or quasi-independent processors for this one? The "knowledgable industry analyst" says "US commercial and government customers may feel they have no choice but to buy the Japanese machine [instead of a Cray]" - any comments? The AP story goes on to define a Floating Point Operation as a "high precision calculation on rows and columns of numbers". Sigh. John Board INET: jab@dukee.egr.duke.edu Assistant Professor or jab@duke.cs.duke.edu Dept. Electrical Eng'g and UUCP: ...!mcnc!duke!jab Dept. Computer Science BITNET: DBOARD@TUCC
eugene@eos.UUCP (Eugene Miya) (04/16/89)
Yes, this is what I am fishing for. We were given a brochure, it has a completely separate brochure on SX- Unix. The machine sounds impressive. I will learn more later in this week. It is a vector, shared memory multiprocessor. It is faster than the rumored Datamation article. People in CRI also were give the brochure, so they know about it as well. There are a variety of open questions about the nature of the architecture (there is a photo BTW, just a bunch of boxes in asymmetric pattern), but it is a multiprocessor (like the VP-2600, only more powerful). They have "all the right buzzwords." TCP, X support, vectors, (oh vectorizing C), FO[ARTN]*, etc. I don't know which Government agencies are going to get "forced" into buying this machine, it won't be us, we don't get "forced," but we would certainly like to know who. Another gross generalization from --eugene miya, NASA Ames Research Center, eugene@aurora.arc.nasa.gov resident cynic at the Rock of Ages Home for Retired Hackers: "Mailers?! HA!", "If my mail does not reach you, please accept my apology." Live free or die.
eugene@eos.UUCP (Eugene Miya) (04/16/89)
>SO WHAT?
Permit me to explain why I said this about small supercomputers.
Many people tend to equate "super-" anything with "BIG." That's called
the Texas mentality. [Not to dump on Texans, 3 uncles were officially made
honorary Texans after WWII.] All these SUPER-conducting, SUPER-colliders,
SUPER-tankers running around, etc. Nothing wrong with said mentality, okay?
Well the problem in computers tends to be the reverse situation. We are
making machines smaller (unless your only experience is on micros
where you are largely (pun intended) staying the same size, just increasing in
speed and memory capacity.
A 10 inch cube: well when I was first scanning, at first I thought I
was scanning 10 cubic feet, well, that's about the size of a Cray-2
and that's "over a composite GFLOPS." We've had that for a few years,
note too that NeXT boxes are a cube foot as well. Remember the measure
there is YEARS. People are already working all over the work to make
specialized as well as generalized supermachines (there's that stupid
prefix I hate) which will fit into a cube. We are not alone.
All DARPA is doing is announcing yet another political project.
Remember my favorite little 1941 scene? "There has to be a way to make
these things smaller?"
There was announced a project to build a "massive memory machine."
What it was: a VAX with a 1 GB physical address space. When this machine
was announced, there were 2 GB machines rolling off production lines in
two countries of the world. Needless to say, this announcement (they didn't
have a machine at the time), was regarded as something of a joke.
At a Salishan high speed computing meeting over a year ago, one of my
officemates open up his session by tempering the audience and told them
to question every speaker with the words: SO WHAT? You should take
his wise words, too.
So if there is anything I hope I have done in all these dumb little keywstokes
of mine [I can't touch type], I hope you got 1) a sense of perspective (size),
2) got a bit of skepticism (or cynicism), and 3) SO WHAT?
Another gross generalization from
--eugene miya, NASA Ames Research Center, eugene@aurora.arc.nasa.gov
resident cynic at the Rock of Ages Home for Retired Hackers:
"Mailers?! HA!", "If my mail does not reach you, please accept my apology."
Live free or die.
The computer industry also makes some great new words, one of my favorites
is vaporware. Otherwise, just ignore me.urjlew@ecsvax.UUCP (Rostyk Lewyckyj) (04/18/89)
In article <3216@eos.UUCP>, eugene@eos.UUCP (Eugene Miya) writes: > I don't know which Government agencies are going to get "forced" into > buying this machine, it won't be us, we don't get "forced," but we would > certainly like to know who. > I don't understand this statement at all. Who will do the forcing? why? how? and why would forcing of any kind be necessary? I thought the quiet word was buy USA technology, just as they quietly discourage buying US high tech items in Japan. I hope that you will explain, in public or private. ----------------------------------------------- Reply-To: Rostyslaw Jarema Lewyckyj urjlew@ecsvax.UUCP , urjlew@unc.bitnet or urjlew@tucc.tucc.edu (ARPA,SURA,NSF etc. internet) tel. (919)-962-9107