Slashdot Mirror
Top 500 Fastest Computers
epaulson writes "The Top500 list has been released for the first half of 1999. The number one machine remains ASCI Red. The biggest Linux machine is cplant at 129, and Avalon is number 160. The list is a ranking of results from the LINPACK benchmark, which is a Linear Algebra code, so things like distributed.net and SETI@home don't count. "
No, It's really inaccurate. I know of two Fujitsu computers at the Department of Computing, Imperial College, London, UK that are big enough to make the list.
Thing is, people have to go and register themselves on the list & probably no-one knows it exists.
One of my friends told me it was for CEDAR, the document analysis/recognition research group, but I have my doubts. Perhaps they'll use it to find a remedy for Bell Hall's severe case of sick building syndrome. Ever spend a summer afternoon in there? Errrgghhhh ....
(sorry, this won't make sense unless you go to SUNY Buffalo)
>A Hitachi or some other computers company got busted for "Dumping". That is, selling the good
for less than it costs to produce.
Of course, this is exactly what practically every Internet software company's strategy is--give away the product to gain market share.
I don't care if it's 90,000 hectares. That lake was not my doing.
Does this mean that there are zero supercomputeres in China? That sounds likely.
The original spec does call for the ability to
run Win NT on the Teraflops machine. My manual
describes the details on how to do so. In the good
news, we did get Linux 2.0 running, although it never
could access the MRC mesh interconnect.
-- http://www.swcp.com/~hudson/
I assume the benchmark would run on Kaffe and the new IBM Java, if someone wanted to see what difference a JIT compiler makes under Linux.
fish and pipes
Posted by FrankGraphics:
You'll find the kits at:
http://www.associatedpro.com/apsprice.html
wow, the AMD *does* suck for FP... I got 6.25
mflops in java with my p200MMX..
(it could be the java interpreter though, I'm using netscapes on win98)
---------------
Chad Okere
ReadThe ReflectionEngine, a cyberpunk style n
Just ran the Linpack.java on the 203. fastest computer on earth (at least on the list) and it gave a whopping 8.3 MFLOPs. Guess java isn't the right tool for the job :)
ask: java Linpack
Mflops/s: 8.374 Time: 0.08 secs Norm Res: 1.43 Precision: 2.220446049250313E-16
They are probably vector processors, and vector processors are good for LINPACK. (On Hitachi's homepage, they called it "pseudo vector processing".) Thus they have good Rmax/Rpeak ratio compared to other superscalar processor based computers.
ASCI Red was Intel's supercomputing division's last hurrah -- the king is dead; long live the king.
Back in November, one vendor (don't remember if it was SGI or Sun) earned itself some animosity when its marketing department counted the number of entries it had in the Top500 and declared itself the leader in the supercomputing industry.
BTW, try using a table to get your columns to line up.
Christopher A. Bohn
cb
Oooh! What does this button do!?
Ah, but supercomputers fill the niche of "performance at any cost." The notion of price-performance wasn't heard in HPC circles until the Beowulf project in 1994.
And you're absolutely right about the scaling of applications. But the use of the LINPACK benchmark was a necessary decision to make the list accessible to most anybody who had a supercomputer, despite the benchmark being obsolete.
Christopher A. Bohn
cb
Oooh! What does this button do!?
Sorry, I don't go to SUNY Buffalo. I'm in Buff State sweating cause NY-MO just turned off the A/C. Maybe they are trying to find a cure for the 8% sales tax, what to do with the Aud, and how do we get a decent radio station around here?
I've actually run the program on both Linux and Windows using Navigator. For some reason, I get better marks under Win98.. It may be because of something I did wrong under Linux, but I don't know. Any takers?
Allen Cain
Just thought I'd add a little comment... Dr. Jack Dongarra is a Distinguished Professor of Computer Science and the University of Tennessee and Oak Ridge National Labs. He also runs (or co-runs, I can't remember exactly, at the moment) Netlib, one of the co-sponsoring agencies of the Top 500... Just thought I'd say Go Vols! :)
Allen Cain
Posted by FrankGraphics:
"The market for such CPUs would be very small, maybe a few hudred per year"
Actually there would be a heck of a lot more demand for such a CPU if the production cost could be economical enough for let's say CG or games.
"Needless to say, there would also be many technical difficulties. Feeding thousands of registers would require a very wide memory architecture, a few thousand bits might be a good start. "
Sony's next generation "Play Station" has a 2000+ bit bus. So the technical difficulties aren't really a stumbling block.
"SIMD can be used effectively only when there is one operation that is done to a big array of data. eg you have an array of 1024 bytes, and you want to increase the value of each byte by one. However, not all code is like this. "
True but you could use more execution units that can control several registers the ratio could be something like 2:1, 3:1, 4:1,...etc. The optimization of wafer real-estate verses execution cycles becomes the issue. Wafer size can be increased to accomodate more processing power which BTW is more power effcient than these clusters of PCs(Sarcasim) being used today.
The technology that comes to mind to at least experiment with such architectures would be FPGAs. I think I read somewhere were FPGAs were being used by a company to produce a massive parallel processing system.
Something to think about also is the needed R&D for such systems could come from talented engineering students working on MS or PhDs. They could use electronic CAD systems on campus to produce the liths for CPUs and have Intel or Motorola grow the chips. This would reduce the cost of R&D by billions. Giving access to Intel's or Motorola's facilities that are manufacturing chips anyway wouldn't cost much. Remember these companies only grow the chip. The Companies that support such a program get first rights of refusal for any new CPU architecture produced by a student. Almost like open source for hardware. Just a thought.
The hitachis are vector CPUs. You don't have much
use for them, but they sure do run optimized linear alrebra codes fast. Most other machines
on the list use commodity CPUs. Cray used to
have a bunch of vector CPUs in MPP configuration,
but now they just use Alphas. Doing a separate
CPU line for supercomputers is just too expensive.
Let's hope the vector stuff is gonna make a
comeback in next generation "multimedia" instructions or something like that.
> I read somewhere that american companies are _way_ out ahead in dealing with the millenium bug, followed by the UK & Germany, then rest of Europe, then Japan.
:-)
I believe Scandinavia and the Benelux region is in fact ahead of Germany when it comes to dealing with the Y2k problem, on par with the USA and the UK.
When it comes to IT usage, Sweden and Finland is in fact #1 and #2 in the world when you count per capita. In Sweden over 50% of the population used the Internet last month. Mobile telephone usage is WAY ahead. There is in fact a lot of interesting research done in northern Europe when it comes to wireless communication, for instance Bluetooth, and other high-tech areas. But American media is a bit bad on reporting on non-american news....
Do I have an inferiority complex? Hell yeah.
Being bitter is drinking poison and hoping someone else will die
Secondly; is Blue Mountain completely up to speed yet? I seem to remember reading that it was going to be the fastest (albiet not by much) when all the processors were finally added. I dunno, maybe I was just smoking something or reading SGI press releases....
----
Every year during my review, I just pray the words "slashdot.org" aren't mentioned.
Yeah, I have an account on the 13th fastest computer in the world. Nothing quite as fun as sitting down and allocating 512 Processors to work on the formation of a planetary system. The Unicos/MK OS on Cray's is a very interesting example of microkernel design that is well done. Of course, the optimizing compilers on that baby are even more astounding.
neutrino
History has the relation to truth that theology has to religion-i.e. none to speak of. - Lazarus Long
A think that a computer's power is reflected by its use, not by megaflops.
The results count, not only the potential.
Then the top 500 is useless, since it doesn't describe the task.
We should search for the most useful computer in the world. That would lead to a great debate.
I will find later.
So, what sort of numbers would a uniprocessor machine spit out under the linpack benchmark?
(i.e. my cyrix 166)
We made the list!!! Hurray, something besides snow!!! I don't think the UB site is finished yet. I'll bet that those government classified lisings are not there just to run a few stockpile calculations and a screensaver. I have some opinions, but I will keep them to myself.
This list is accurate at what it is trying to measure. The list is for general-purpose supercomputers and so this eliminates a number of machines that are much faster for dedicated tasks. The NSA, for instance, has machines that are very good at the task of factoring. These machines, however, are not capable of performing general tasks, and thus are not candidates for this list. In addition, you may say that there are other general purpose supercomputers out there that are faster, but not on the list. This is doubtful, as the machines have to be made by someone with experience at the rather specialized task of making ultra-fast systems. All of the major producers of supercomputers recognize the importance of such benchmarks and thus work with people to do Linpack tests. The importance of these benchmarks is that the company that produces high results will get more sales. The only computers now left out of this list would be those produced in secret by governments or other such entities. These groups fall behind because they do not have the existing knowledge and resource base needed to produce such computers. The end result: This list is quite accurate with perhaps a few Ultra Secret machines left out.
BTW, I work on two of the machines on this list.
neutrino
History has the relation to truth that theology has to religion-i.e. none to speak of. - Lazarus Long
The most useful computer in the world is the one you use.
----
Every year during my review, I just pray the words "slashdot.org" aren't mentioned.
I'm very surpized at how many Industry computers are on the list and how early they dominate. From 21 on industry starts to make a strong showing. For many years Floating Point super computers were strictly reasearch/university material...
"There is no spoon" - Neo, The Matrix
"SPOOOOOOOOON!" - The Tick, The Tick
I just went and got the 1000x1000 double precision benchmark from netlib.org. I grabbed the lapack library and g77 from the debian website (Debian 2.1/slink versions)
:)
On my 400MHz K6-2, I get 16 Mflops without optimization, 20 with -O3. Not quite what was listed in the performance document, but that might have been with a hand-tuned library.
For comparison, my home machine (a 300 HHz K6-2) gets 13 Mflops unoptimized, 20 with. It's running Debian 2.2pre/potato which uses egcs, so the optimization is probably better. Both machines have 100 MHz fsb and 1 MB L2 cache.
There's a fun java version on the LINPACK benchmark as well. I get 1.4 Mflops.
Where do I get one of these fine computers, do you think best buy carry them ?? :)
-----
-----
just another lover of Linux and VIM
I'd LOVE to put Windows 98 on ASCI Red !
:)
how many teraflops would we get, then?
----------
"They misunderestimated me." --George W Bush, Nov. 6, 2000
Posted by FrankGraphics:
.18 microns, say it's .35 and the wafer is one inch square. The added benefit of proximity and architectural design along with the new machine code(software) would make ground breaking research. This use of clunky CPUs doesn't impress me. The Mercede will use multiple registers and execution units and will do more for the development of parallel processing software research than any of the listed supercomputers!
Why use thousands of antiquated CPUs? Wouldn't it be better to build a multi register CPU similar to MMX but on a scale in the thousands? So what if the chip density isn't
According to IC's web site, they have one Fujitsu VX/1 vector supercomputer and one Fujuitsu AP3000 massively parallel server, which is effectively a nest of 48 independent UltraSPARC systems linked together Beowulf style.
But when you look up Fujitsu on the Top-500 database, it turns out that only the vector supercomputer (VPP) series make the list, and none of their AP-xxxx systems.
For the VPP series, the entry-level for the top 500 is a twelve processor system rather than IC's single vector processor.
On the other hand, the AP-3000 would have enough total throughput at 45.6 Gflops to get on the list at number 172. But my guess is that it can only achieve that for problems that split into relatively big independent chunks.
That might be OK say for servers and big CFD models, but I suspect that the LINPACK test suite needs a much more fine-grained parallelisation, and would be much harder hit by communication latencies between nodes.
That's just a guess: perhaps any real supercomputer experts out there could say whether this sounds right ?
Even though I believe they modified the souce of Linux to run on all them processors, it is one of the advantages of Linux.
I am awaiting a press release from Redmond.
--
If you go to the very end of the LinPack ratings you'll see that the AtariST beats the MacIntosh as well as the IBM 286/287 combo
:-)
I knew it
send + more == money?
As you would expect:
1) By number of computers on the list, the top 6 countries are G7 members;
2) The G7 all have a computer in the top 47, while no non-G7 state has a computer in the top 52.
That machine had LOTS of chess-specific hardware acceleration.
retrorocket.o not found, launch anyway?
Posted by FrankGraphics:
Just saw a FPGA kit that slips into a PCI slot in a PC for some US $300. I hear some FPGA chips have up to 1,000,000 gates and operate at 200 MHz! Given that it would take two gates to produce one register for one bit, I figure the possibility of a 32 bit processor with 200 or maybe even 500 active registers with a humble instruction set of 17. XOR, OR, AND, conditional jump, jump, Shift Right, Shift Left, Integer Math: Add, Sub, Multiply, Divide, Floating Point Math: Add, Sub, Multiply, Divide, Load from memory and Block Transfer. Pretty basic but should get the job done. Each register will have it's own execution unit. Now the number of cycles to complete an operation should range from one to four with the mean being about 2. Lets see 200 MHz average 2 cycles per op that's 100 million times five hundred, that's 50 billion instructions per second! Give me a few months and My supercomputer will be on this list and will only take a thousandth of the space of those other monsters!
Um, number 16 or 19 (can't remember) on the list was in Chippewa Falls, WI. The only important thing I can think of in Chippewa Falls is...Leinenkugel's Brewery! Maybe those of you who aren't from the Upper Midwest aren't as familiar, but maybe they do a whole lot more volume than I thought...hehe. Being a Minnesotan, I think that Leinie's is close to the only useful thing to come out of Wisconsin. On a side note, 14 of the top 500 are in good old Minnesota, mostly because of the University and Cray. That's a pretty decent total for one state. And the U has some pretty kickin' machines it would seem.
Where did you see this nifty device?
/., anybody with
Just curious . . .
If you don't want to clutter
nifty info on FPGA stuff, mail me at
johng@NOSPAM.eng.auburn.edu
And, of course, s/NOSPAM//
(Isn't it sad we have to mangle?)
Deep Blue is a member of IBM's SP family. IIRC, it was more souped-up (suped-up? :-) than a standard SP-2 but not quite up to par with ASCI Blue Pacific.
Christopher A. Bohn
cb
Oooh! What does this button do!?
A machine cannot be included on the list if the owners don't submit the LINPACK results for consideration.
Christopher A. Bohn
cb
Oooh! What does this button do!?
It's a shame. Sure was nice for 7 months to be able to point out that a Linux cluster was one of the 100 fastest supercomputers (C-Plant was ranked #92 on November's list). We'll just have to wait until the port to ASCI Red is finished (I seem to recall a University team (UVA?) was working on that. Of course, I could be completely mistaken.
Christopher A. Bohn
cb
Oooh! What does this button do!?
That I think it is pretty cool that you can get an account on the 129th (CPLANT) fastest computer, if your work is valid. I just think that is too frigg'n cool.
Linux O Muerte!
"The only computers now left out of this list would be those produced in secret by governments or other such entities." - not true. I've worked on a number of systems that would spec on the list and aren't there.. Why? A lot of major corporations don't like to tell competitors what they are using.
Okay, I was bored, so I went through and counted some stuff:
/. won't let me).
The numbers won't add up correctly because several of the machines were credited to two co-builders. Or I could have made a mistake.
Company: total, # out of the top 10, highest rank
(I tried to make this line up but apparently
SGI: 182/500, 7/10, #2
IBM: 118/500, 1/10, #8
Sun: 95/100, 0/10, #54
H/P: 39/100, 0/10, #150
Fujitsu: 23/500, 0/10, #26
NEC: 18/500, 0/10, #29
Hitachi: 12/500, 1/10, #4
Compaq: 5/500, 0/10, #49
Intel: 4/500, 1/10, #1
Self-made: 3/500, 0/10, #129
SNI: 2/500, 0/10, #66
Tsukuba: 1/500, 0/10, #18
Siemens: 1/500, 0/10, #355
This ranking above looks very different than the ranking of the top five computers. For example, Intel, who is #1, is basically a non-factor in the supercomputer market, with a mere three other computers on the list. H/P and Sun, which don't even make the top 50, seem to have the mid-level supercomputer market locked up, with 134 computers between them. SGI, however, is still the undisputed leader, from the high end (7/10) to the mid and low ends of ths list.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
A Hitachi or some other computers company got busted for "Dumping". That is, selling the good for less than it costs to produce. This drives others out of business and leaves the market for the Dumper. Of course, they have to be able to still stay in business while "dumping". When your Hitachi you still have other products that are very profitable.
Linux O Muerte!
Wouldn't it be better to build a multi register CPU similar to MMX but on a scale in the thousands?
Intresting idea, but it does have it's flaws. For one, designing a new CPU is _really_ expensive. And as you add more parallelism, it gets even more complicated and expensive (look at Merced). The market for such CPUs would be very small, maybe a few hudred per year. As you may have noticed, even supercomputers are made as cheap as possible these days (eg. beowolf).
Needless to say, there would also be many technical difficulties. Feeding thousands of registers would require a very wide memory arcitecture, a few thousand bits might be a good start. I sure wouldn't want be the engineer responsible for designing a mobo for those CPUs..
Few architectural problems also. SIMD can be used effectively only when there is one operation that is done to a big array of data. eg you have an array of 1024 bytes, and you want to increase the value of each byte by one. However, not all code is like this. You might want to inrease the value of the first elemnt by one, the second element by two and so on. MMX just became useless, there is no paralelism here. Now we have a CPU that is working at a fraction of its full potential: of the 2000 or so registers, only two are used. There is other stuff too, but I lazy so..
>Guess java isn't the right tool for the job :)
It's not the language that's the problem, it's the fact that you're running the program via an interpreter (aka your Java VM). Compile the Java to native code and you should get better results.
I don't care if it's 90,000 hectares. That lake was not my doing.
Anybody know about highly-pentium2 optimized lapack/blas libraries? Anybody know if compiling
the BLAS under pgcc/pg77 breaks the test cases? Are there binaries in rpm format?
Okay, I was really bored, so I did more stats. This time by country.
USA: 292/500, 7/10, #1
Japan: 56/500, 1/10, #4
Germany: 47/500, 0/10, #15
UK: 29/500, 2/10, #7
France: 18/500, 0/10, #47
Canada: 8/500, 0/10, #29
Sweden: 7/500, 0/10, #71
Netherlands: 6/500, 0/10, #146
Switzerland: 6/500, 0/10, #339
Italy: 5/500, 0/10, #36
Australia: 5/500, 0/10, #102
Korea: 3/500, 0/10, #78
Denmark: 3/500, 0/10, #275
Belgium: 3/500, 0/10, #286
Spain: 3/500, 0/10, #314
Finland: 2/500, 0/10, #53
Norway: 2/500, 0/10, #193
Austria: 2/500, 0/10, #392
New Zealand: 1/500, 0/10, #64
Luxembourg: 1/500, 0/10, #247
Mexico: 1/500, 0/10, #436
Summary: United States 292 vs. Everybody Else 208.
In the top ten, it's United States 7 vs. Everybody Else 3.
If you compile the stats by the country in which the corporation that made the computer is based, American companies are responsible for over 400 of the top 500 supercomputers (just about everything except the Japanese stuff).
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Anybody know what kind of processors the Hitachi supercomputers at #4 (128 processors) and #12 (64 processors) get? They seem to be the fastest per-processor for LINPACK...
The MFLOPS/$ for a processor goes down for the top end. The problem is that you can't just stick 1000 CPUs in a computer and then say it's 1000 times faster than a single CPU. For some problems, it might not be faster at all.