Japan Unveils Next-Generation, Pascal-Based AI Supercomputer (nextplatform.com) 121
The Tokyo Institute of Technology has announced plans to launch Japan's "fastest AI supercomputer" this summer. The supercomputer is called Tsubame 3.0 and will use Nvidia's latest Pascal-based Tesla P100 GPU accelerators to double its performance over its predecessor, the Tsubame 2.5. Slashdot reader kipperstem77 shares an excerpt from a report via The Next Platform: With all of those CPUs and GPUs, Tsubame 3.0 will have 12.15 petaflops of peak double precision performance, and is rated at 24.3 petaflops single precision and, importantly, is rated at 47.2 petaflops at the half precision that is important for neural networks employed in deep learning applications. When added to the existing Tsubame 2.5 machine and the experimental immersion-cooled Tsubame-KFC system, TiTech will have a total of 6,720 GPUs to bring to bear on workloads, adding up to a total of 64.3 aggregate petaflops at half precision. (This is interesting to us because that means Nvidia has worked with TiTech to get half precision working on Kepler GPUs, which did not formally support half precision.)
Pascal-based? (Score:5, Insightful)
Am I the only one that thought "LISP machines, okay, but Pascal [standardpascal.org]?"
Not alone (Score:3)
That is actually the first thing that sprang to mind, even though I had been looking specifically at Pascal based GPU's recently. :-)
Re: (Score:2)
Borland (Score:3)
Re: (Score:2)
Delphi.. Ohhh such wonderful memories..
I've heard its latest incarnation as Embarcadero Delphi is quite good but I'm no judge of programming environments
Re: (Score:3)
"Ditto here. I had some memories of long dead undergrad programming courses."
Yes, Pascal, really? If we're going to invoke the age of steam, why not go full Fortran with this project?
Re: (Score:2)
Fortran is still very much in use for mathematical programming.
Re: (Score:3)
Makes no sense. Pascal is a perfectly fine language, even classic Pascal. Fortran, especially classic Fortran 77, is really lacking in many ways (newer versions add newer features but that's like comparing Visual Basic to Basic and trying to call them the same language). Granted Modula-II or Ada is much more suitable than Pascal.
Re: (Score:2)
Fortran, especially classic Fortran 77, is really lacking in many ways
But perfectly adequate for solving differential equations, which is the main task of scientific and engineering programs.
Re: (Score:2)
Latin is older than English, but a thousand times better as a language.
The value of Latin is not that it's "better" in modern usage but that as a root of so many later European languages, it is easily used as a lingua franca for European speakers. An Italian MD, a British MD and a French MD all understand what bis in diem on a prescription vial means.
Re: (Score:3)
"English is a language that lurks in dark alleys, beats up other languages and rifles through their pockets for spare vocabulary."
--attributed to James D. Nicoll
The original, complete quote appears to be:
"The problem with defending the purity of the English language is that English is about as pure as a cribhouse whore. We don’t just borrow words; on occasion, English has pursued other languages down alleyways to beat them unconscious and rifle their pockets for new vocabulary."
Re: (Score:2)
Latin is older than English, but a thousand times better as a language.
Which must be why it's used so today. Oh, wait....
Re: (Score:1)
Your position might be viewed as an example of fallacy of the single cause [wikipedia.org] with elements of argumentum ad populum [wikipedia.org]. -PCP
Re: (Score:2)
Latin is older than English, but a thousand times better as a language.
Which must be why it's used so today. Oh, wait....
Saw an article on a study once about the value of either language standardization and regularness versus freeform and irregular usage given between the examples of French and English. It seems that the irregular and modifyable nature of English actually helps more towards usage and adoption than standardization.
Re: (Score:2)
Yes, I too would rather have a language with a million inflections to save a dozen pronouns and structured so inflexibly that you have to add a zillion articles and conjunctions.
Re: (Score:2)
I'm worried the animal activists might take offence at peta-flops.
Re: (Score:1)
I'm older school and keep hoping for an ALGOL supercomputer.
Re: (Score:1)
Perhaps APL should be revived. Imagine the money to be made by keyboard manufacturers.
Paai
Re: (Score:1)
APL was effectively superseded by the J language [wikipedia.org].
I taught myself some APL[1] using a standard keyboard and an APL implementation that ran on OS/2, if memory serves. Wasn't bad. Yes, you had to compose all the APL characters using Alt-key sequences[2], but APL is so terse that you didn't have to do it very often.
J uses ASCII rather than APL's grab-bag of symbols. I'm not sure I'd call it more readable, though.
It would be interesting to see a less-opaque Iverson-Backus language, with the matrix operators and
Re: (Score:2)
Now we have multi-threading, Algol68 might actually be the best choice. However, history supports C-more Cray.
Extra Crispy (Score:3)
I'm sorry, but I like my KFC hot from the fryer. This experiment should end immediately.
Re: (Score:2)
Well, there's your problem, right there. Why not try the appropriate product which was meant for that [amazon.com] instead?
No, no need to thank me. Where would the world be if us experienced types didn't pass along our wisdom to the clueless?
Re:Chicken vs. Swallow (Score:1)
Re: (Score:2)
Precisely so. Well played. :)
Re: (Score:1)
Re: (Score:2)
I was thinking "Mr Wirth would be very happy..."
Re: (Score:2)
Umm...is this the Pascal as in the programming language?
Re:Pascal-based? (Score:5, Informative)
Umm...is this the Pascal as in the programming language?
Does anyone still program in Pascal? The last time I saw it on a resume was more than 20 years ago.
Anyway, this supercomputer has nothing to do with the Pascal programming language. It is built using NVidia Pascal GPUs [wikipedia.org].
Re: (Score:2)
Some would say that everything old is new again; good design merits consideration.
I am not suprised by this. When faced with coming up with something better, why not choose something old (and proven) rather than try to create something new?
Re: (Score:2, Redundant)
Well first off the super computers aren't about the Pascal language but the Pascal chip. I'd disagree that Pascal was all that proven out. It seemed very quickly to have had structural flaws which caused other languages to overtake it. Pascal was fairly low level yet it lacked good low level interfaces. Which is why it lost out to C. Pascal supports admit this and one of the main directions of Turbo Pascal / Delphi was to introduce into Pascal handling for lower level code (example partial compilation)
Re: (Score:2)
Sorry but eveything you write about Pascal is wrong. But I'm to lazy to correct it, you can read wikipedia.
You clearly never used it, so why write that bullshit?
Pascal, especially as UCSD P-System was for decades the most widely used 'OS' and programming language on the planet.
Re: (Score:1)
Huh? When? What year? The p-system is released around the same time as the microcomputer. Pascal never approaches the ROM/Basic/OS combo those systems defaulted to.
Re: (Score:2)
You should read the relevant wikipedia articles :)
Re: (Score:1)
That's a longer answer than a single number or range of numbers.
Re: (Score:2)
As much as I love C, it is not a good fit on other architectures such as stack machines. Burroughs machines (now UNISYS) use an extended version of ALGOL as a system language with impressive results, especially when it comes t
Re: (Score:1)
Re: (Score:2)
OK, thanks.
I liked Pascal. It was very elegant.
Re: (Score:2)
Re: (Score:3)
Being a Pascal programmer I felt horribly bait and switched.
Re: (Score:1)
The real question is, did anyone NOT think that?
And, more importantly, just how calculated was that clickbait title, and who or what calculated it?
Re: (Score:3)
Re: (Score:2)
Pascal has always been exceptional performing math. Makes sense for a super computer IMHO. The fact that I wrote pascal for my math degree is bonus!
Are you sure you're not thinking of Fortran? I've never heard that Pascal is unusually good at math compared to other general purpose programming languages.
Re: (Score:2)
Re: (Score:2)
He meant Pascal, the human. Not Pascal, the language. ... or basically every pro edural language.
And Pascal as languge is as good as C in doing math
Re: (Score:1)
but Pascal [standardpascal.org]?
It's more like Pascal [freepascal.org] or, alternatively, Pascal [embarcadero.com].
Re: (Score:3)
Re: (Score:2, Informative)
Kepler Fluid Cooled, though Tsubame is a kind of bird and the computers are indeed immersed in oil.
Also Japan has an unusual tradition of eating KFC for Christmas with reservations made months in advance.
Re: (Score:2)
For some reason, Kernel Sanders made me think of Dr Fun: Kernel Panic [ibiblio.org]
For those that don't know, Dr Fun was the first webcomic [ibiblio.org] 520 weeks, or 10 years worth
Re: (Score:2)
I had a pretty deep fondness for Pascal back in the day, and messed around with Delphi, Modula and Oberon, but the reality is that these aren't exactly common languages anymore, at least not in commercial circles. It's a real pity too, because learning TurboPascal was my sort of "Wizard of Oz black-and-white to color" moment back in highschool, where I shed all the evils that I had learned through mucking around with various flavors of BASIC, and basked in the glory of structured procedural code.
Re: (Score:2)
I thought all Pascal use was for BBS DOOR games, and the internet killed it?
Re: (Score:1)
Faster hardware is not always a boon. Now we have Java monsters that eat up all the performance of even decent hardware. Easy to learn? Nah, with the myriads of different libraries and paths it is a conceptual mess. And it was supposed to be the cure-all for viruses, but that has not materialized either.
The good Lord gave us C, Bash and the CLI , and we frail humans should not presume to improve on His creation.
Paai
Re: Good ol' days (Score:2)
It is good that I am an atheist, then.
Re: (Score:2)
Faster hardware is not always a boon. Now we have Java monsters that eat up all the performance of even decent hardware. Easy to learn? Nah, with the myriads of different libraries and paths it is a conceptual mess. And it was supposed to be the cure-all for viruses, but that has not materialized either.
The good Lord gave us C, Bash and the CLI , and we frail humans should not presume to improve on His creation.
Paai
Those are tools of the Devil; on the 1st day, the Lord created binary, microcode and logic gates and saw that it was good.
Re:Good ol' days - blame p2c (Score:1)
One it was realized you could mechanically translate Pascal to C, then compile the C and get a 2-3x speed up (compared to interpreting p-code), Pascal started dying.
C has issues, but in practice it has less of them than Pascal, so the resulting code was easy to tweak to get even more performance.
Re: (Score:3)
Pascal usage quickly shrank. I'm not quite sure why, it was a fairly decent compiler-based language.
I am pretty sure why. More modern operating systems came up and nearly invariably they were all C based. If you wanted to call the OS libraries you were better off using C. That's why Pascal died off.
Re: (Score:2)
You can call a C library or OS functiom from Pascal, too.
There is no difference.
C is not something magically, you know ...
Re: (Score:2)
I don't know if there is any difference between calling conventions, but the mere fact you would have to write your own headers in Pascal seems pretty significant to me.
Re: (Score:2)
Ofc, there are different calling conventions.
Usually C pushes arguments from left to right onto the stack and the caller cleans up the stack (because of variable argument lists), in Pascal arguments get pushed from right to left and the called procedure/fuction cleans up the stack.
Interfacing with C you usually do via so called 'units'. Units have an interface section and an implementation section. In the interface section you define functions/procedures and call also define if they are written in a differe
Re: (Score:2)
It also allowed nesting of functions.
FWIW GCC allows that now in C.
Re: (Score:1)
The reason is both the reason for its success and the reason for its failure. The Pascal language makes a lot of compromises in areas of readability and organization to allow for small compilers. In the case of PCs, it was much easier to write a Pascal compiler that ran well off a 128k floppy than a C compiler. That stopped mattering pretty quickly.
Good for them... (Score:2)
These P100 come with sweet HBM2 and around 500GB/s in memory bandwidth... everything based on dense linear algebra (AI, but also physics simulations) is basically flying on them.
Re: (Score:3)
I am going to make make a supercomputer that runs on BASIC A.
Whom the gods would destroy, first they teach basic.
Re: (Score:1)
RTFA (Score:4, Insightful)
Re: (Score:3)
It does make one wonder what they were thinking - if they were at all - when they chose the name.
Re: (Score:2)
Probably true, but irrelevant. The point is that they didn't think to check if the word was already in use in a related context.
Re: (Score:3)
Pascal, the GPU design. Not Pascal, the language.
So... It won't run Delphi then?
Re: (Score:1)
Delphi closed, along with BIX, Prodogy and all the other AOL like online services of the early 90s.
Re: (Score:2)
Delphi closed, along with BIX, Prodogy and all the other AOL like online services of the early 90s.
They're all gone? Compuserve too? Sad!
One of few people (Score:1)
GPUs have limited applications (Score:4, Insightful)
As of now there are very few applications for massively souped up GPU processes. Fluid mechanics loves this GPU. Navier-Stokes is probably the most difficult equation to solve, agreed. But it is hyperbolic, with limited "zone of influence", and numerical equations are quite simple, just mass, momentum and energy balances in the control volume. It plays well in GPU, the calculations fit inside the teensy memory and processor. All time domain problems are hyperbolic and they all can be ported to GPU, theoretically. But try squeezing Maxwell's Equations into that teensy processor!
Graphics card companies are desperately looking for new markets and they keep pushing this. They might as well push a wet noodle across the table. It ain't gonna go nowhere it didn't wanna go.
Re: (Score:2)
I wouldn't call a modern GPU that constrained really. The only thing they lack is memory protection. It's also a lot easier to program on a GPU ever since the SIMT paradigm came out (i.e. CUDA, OpenCL). Also plenty of modern processors come with a GPU on the same die as the CPU. Like nearly all smartphones for example.
Re: (Score:2)
I wouldn't call a modern GPU that constrained really. The only thing they lack is memory protection. It's also a lot easier to program on a GPU ever since the SIMT paradigm came out (i.e. CUDA, OpenCL). Also plenty of modern processors come with a GPU on the same die as the CPU. Like nearly all smartphones for example.
Actually all modern GPU have had memory protection for several generations. The problem GPUs have is that don't generally have full support for demand paging and precise exceptions (SIMT makes that pretty expensive). For example, putting in hardware for highest possible performance and hardware to be able to hit a page fault and be able to clean up and restart multiple threads (that might be communicating or synchronizing states) are two different hardware optimization points. That being said, some limit
Re: (Score:2)
GPUs are problematic because there's not a good way to compute matrix factorizations on them.
Have you looked at this paper on large matrix factorization [arxiv.org]?
Re: (Score:2)
This one is targeted towards neural nets. Does that work ?
Re: (Score:2)
Re: (Score:2)
Pascal? (Score:1)
Making all the other supercomputers Wirth-less.
Re: (Score:2)
Making all the other supercomputers Wirth-less.
Or by name, Veert-ually worthless [thestandardoutput.com]...