Where are PC / Mac / Linux benchmarks (programs) that compare performance to
Cray and other supercomputers ... must exist somewhere.

Vista has a benchmark system, but it makes no Cray comparison.

BOINC clients, in Graphical mode could provide Cray comparasons, but don't.

In comp.unix.cray Max Power <mikehack@washington.edu> wrote:
>
> Where are PC / Mac / Linux benchmarks (programs) that compare performance to
> Cray and other supercomputers ... must exist somewhere.

Linpack?
You can compile it on most systems. The real thing is Fortran
code, but there is a C version too. The C version I found years
ago gave about the same results as the Fortran version on single
CPU CISC and RISC systems. When benchmarking your own little
Cray (or other vector processor) in your basement, better use
the Fortran version.

Dennis

--
Don't suffer from insanity...
Enjoy every minute of it.

Max Power wrote:

> Where are PC / Mac / Linux benchmarks (programs) that compare
> performance to Cray and other supercomputers ... must exist somewhere.

> Vista has a benchmark system, but it makes no Cray comparison.

> BOINC clients, in Graphical mode could provide Cray comparasons, but don't.

I don't much see the point on bragging rights here unless you are running a
cluster. It is what it is. The hardware benchmarks are all over and for free.
After that it is just MS v linux. Your computer will be different of course but
not by more than a few percent. Unless you are doing something like BOINC that
is 24/7 you can probably get a greater performance increase by learning to type
faster. 24/7 programs benefit by using the time between keystrokes so it doesn't
matter how long between them.

I am not trying to talk you out of it. It is only now that you mention it II
can't remember seeing benchmark suites in years. There is no longer a point in
showing linux is faster the MS as that was done years ago. MS is getting slower
with every new release.

--
If Jews had wanted peace they would never have gone to Palestine.
-- The Iron Webmaster, 3961
http://www.giwersworld.org a1

Max Power wrote:
>
> Where are PC / Mac / Linux benchmarks (programs) that compare
> performance to Cray and other supercomputers ... must exist somewhere.
>
> Vista has a benchmark system, but it makes no Cray comparison.
>
> BOINC clients, in Graphical mode could provide Cray comparasons, but don't.

I tried to see if any Cray users were running BOINC, but found no evidence
of such. I notice there is a comp.unix.cray newsgroup, so I infer that one
can run UNIX on one. Therefore one could get a BOINC client to run on the
Cray I suppose by compiling it from source. OTOH, I do not know if the
actual BOINC applications are available for it. While I did not do an
exhaustive search for Cray users at the BOINC web site, I scanned the users
and found no Cray in the list of OSs being used.

--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 07:40:01 up 23 days, 12:42, 5 users, load average: 4.51, 4.31, 4.18

Jean-David Beyer <jeandav...@verizon.net> wrote:
>
> I tried to see if any Cray users were running BOINC, but found no evidence
> of such. I notice there is a comp.unix.cray newsgroup, so I infer that one
> can run UNIX on one. Therefore one could get a BOINC client to run on the
> Cray I suppose by compiling it from source. OTOH, I do not know if the
> actual BOINC applications are available for it. While I did not do an
> exhaustive search for Cray users at the BOINC web site, I scanned the users
> and found no Cray in the list of OSs being used.

UniCos is the standard (one of the standards?) Cray operating
system. It's quite vanilla UNIX System V.

Moore's law says at some point the laptops at the local store
will be as fast as the Cray's in the earily versions of the top
500 supercomputers list. Wanna bet the laptops won't run as
fast anyways? ;^)

Doug Freyburger wrote (in part):

>
> Moore's law says at some point the laptops at the local store
> will be as fast as the Cray's in the earily versions of the top
> 500 supercomputers list. Wanna bet the laptops won't run as
> fast anyways? ;^)

My current desktop has 8 GBytes of RAM. My first desktop had only about 1.3
GBytes of hard drive.

My current desktop has dual 3.06 GigaHertz Hyperthreaded Xeon processors.
The first computer I ever used had 32K words of RAM, no hard drives, and the
processor took about 2 1/2 refrigerators of space and ran at what would be
called 50 KiloHertz.

My current desktop may have cost around $6000 (including 6 SCSI hard drives,
LCD monitor, etc). The 704 cost around $600/hour to rent (it came with 10
tape drives).

If the Crays were anything like the Control Data machines, they had 60-bit
word size which is substantially more than the 32-bit size of my machine,
although nowadays you can get 64-bit machines at reasonable cost. People in
those days needed bigger word sizes because the roundoff in large floating
point calculations (think analytical weather prediction) was too much even
in double-precision of 32-bit binary (or worse, 32-bit hexadecimal)
machines. THese days, we do not do all that much computation on our machine,
but just push bytes around, so this is less important.

--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 15:40:01 up 24 days, 20:42, 5 users, load average: 4.22, 4.31, 4.27

Max Power wrote:
>
> Where are PC / Mac / Linux benchmarks (programs) that compare
> performance to Cray and other supercomputers ... must exist somewhere.
>
> Vista has a benchmark system, but it makes no Cray comparison.
>
> BOINC clients, in Graphical mode could provide Cray comparasons, but don't.

Just for reference, cray-cyber (www.cray-cyber.org) measured the
performance of thier Cray EL and found "460 MFlops (A 2.4 GHz P4/RDRAM
is 770 MFlops on the same benchmark problem, huge matric multiply, with
Linux gnu cc -O3)". Now, the EL is an old, bottom of the range, machine
(mid 90's) and is clocked at 33MHz (yes, thirty three megahertz). The
key thing with the Cray is that it's a vector processor - it has 8
vector registers, each of which is 64 words where a word is 64 bits. No
x86 vector extensions (MMX etc), or even the Cell, come close to that.
So long as the problem you are trying to solve maps well to that
hardware then it will fly. For general purpose desktop apps the PC will
easily out perform the Cray.

James
http://www.machineroom.info

In article <2be96590-1283-4d9e-bc64-3cf2b836337f@k13g2000hse.googlegroups.com>,
Doug Freyburger <dfreybur@yahoo.com> wrote:
>Jean-David Beyer <jeandav...@verizon.net> wrote:
>> I tried to see if any Cray users were running BOINC,

What's BOINC?

>> I notice there is a comp.unix.cray newsgroup, so I infer that one
>> can run UNIX on one.

Yeah, it was proposed and approved by people who didn't have one/any.

>UniCos is the standard (one of the standards?) Cray operating
>system. It's quite vanilla UNIX System V.

It's approaching 25 years old. Next week we celebrate the founding of
one of our supercomptuer centers.

>Moore's law says at some point the laptops at the local store
>will be as fast as the Cray's in the earily versions of the top
>500 supercomputers list. Wanna bet the laptops won't run as
>fast anyways? ;^)

Erich's list post-dated the retirement of all the numeric numbered CRI
models and most of the alphabetic MP models. Just give a laptop an I/O
intensive big data problem if you want to see a slow machine.

--

In article <fY6dnbInnOPf-p3VnZ2dnUVZ8s-qnZ2d@pipex.net>,
James Wilson <news@machineroom.info> wrote:
>Just for reference, cray-cyber (www.cray-cyber.org) measured the
More of historic interest.
>performance of their Cray EL and found "460 MFlops (A 2.4 GHz P4/RDRAM
>is 770 MFlops on the same benchmark problem, huge matric multiply, with
>Linux gnu cc -O3)". Now, the EL is an old, bottom of the range, machine
True.
>(mid 90's) and is clocked at 33MHz (yes, thirty three megahertz). The
>key thing with the Cray is that it's a vector processor - it has 8
Historic. Current models are a bit more complex.
>vector registers, each of which is 64 words where a word is 64 bits. No
>x86 vector extensions (MMX etc), or even the Cell, come close to that.
>So long as the problem you are trying to solve maps well to that
>hardware then it will fly. For general purpose desktop apps the PC will
>easily out perform the Cray.
Old Crays.

PCs will also tend to out perform the ENIAC as well.

--