Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!mintaka!ogicse!littlei!intelisc!iSC.intel.com!hays
From: hays@iSC.intel.com (Kirk Hays)
Newsgroups: comp.sys.intel
Subject: Re: New supercomputer
Message-ID: <1394@ssdintel.isc.intel.com>
Date: 5 Jun 91 21:46:52 GMT
Article-I.D.: ssdintel.1394
References: <1991Jun2.153238.24866@magnus.acs.ohio-state.edu> <1382@ssdintel.isc.intel.com> <1386@ssdintel.isc.intel.com> <1991Jun5.185850.988@leland.Stanford.EDU>
Sender: news@isc.intel.com
Reply-To: hays@iSC.intel.com (Kirk Hays)
Organization: Intel Supercomputer Systems Division
Lines: 67
Nntp-Posting-Host: roadkill

In article <1991Jun5.185850.988@leland.Stanford.EDU>, izahi@leland.Stanford.EDU (Raul Izahi Lopez Hernandez) writes:
|> In article <1386@ssdintel.isc.intel.com> hays@iSC.intel.com (Kirk Hays) writes:
|> >it's 32 GigaFlops + 17 billion integer IPS, for a total of 49 GIPS, which
|> >is > 6000 times faster than the 386-33.
|> 
|>    I don't think it is fair to make such comparison.

Lighten up, Raul - I didn't make the original comparison - some person at Ohio
State did.

I was correcting my original incorrect correction of *his* comparison in the
above message.

For the record, I believe it is a somewhat silly exercise to compare PCs to
supercomputers, too.  On the other hand, it is interesting to see how large the
span of available computational power remains.

|>    Can you mention an application that runs routinely both in the 386-33 and
|> in the Delta?

An application?  Not without naming a customer, which I won't do.

However, PERFECT club, SPECmarks, linpack, and many other benchmarks certainly
will, which indicates that many scientific/business applications could.

The ORACLE database runs on PCs, and one of our competition, nCube, is porting
ORACLE to their machine, as reported in the trade press.

|>    Applications that benefit from being run in supercomputers are never run
|> in a PC (be it 386-33 or not), since the PC does not have similar memory or
|> I/O resources.

This is not true - I am aware of customer codes that run on our machines, which
are supercomputers, and that are also run on minicomputers, workstations, and
PCs, as well.

The supercomputer allows you to handle a larger problem set, perform more
detailed analysis on the data, and/or do the calculations in much less time than
would be required on a smaller/slower machine.

This is why they exist.

But their existance doesn't automatically mean that an application can *only* run
on such a machine, just as the existence of Formula One race cars does not preclude the use of a Hyundai for commuting.

|>    It is irrelevant to say that your LOTUS 1-2-3 spreadsheet would recalculate
|> 3000 or 6000 times faster since it would not be noticeable beyond 10 or 20
|> times faster, because the spreadsheet is not big enough to even tax the 
|> abilities or[sic] the new 386-33 or 486 based PCs in any business or home 
|> application, while in engineering or the sciences serious researchers use 
|> RISC based workstations.

Nice strawman.  Because you can't imagine an application that spans three or four
orders of magnitude in performance requirements, it must not exist.

Business applications use supercomputers routinely today, and have for at least a
decade.

*Serious* researchers use supercomputers, and have for at least three decades. 
Workstations are for software development, reading netnews, and running X and
EMACS. :^)

[Ever known any *goofy* researchers?]

-- 
Kirk Hays - NRA Life.
Message for Timothy Fay - "Do not eat/wear/exploit things you will not kill."