TOP500's most powerful supercomputers in the world

No surprises that the Chinese are pwning the rest of the world in this department.
 
I guess these computers are bit of an overkill to balance your books. ;)
 
IBM has Quantum Computing open for beta testing. It may prove a threat for current encryption technologies.

https://www.research.ibm.com/ibm-q/

By thinking beyond ones and zeroes, the platform can already solve problems that were considered too complex for classical computer systems to handle -- which means it can solve problems that we haven't even thought of yet in fields like pharmaceuticals, artificial intelligence, financial services and logistics.
Click to expand...

https://www.engadget.com/2017/05/17/ibm-quantum-q-experience-qubits-most-powerful-processor-yet/

Kevin & co can thank me later :)
 
Last edited:
Sabre Wulf said:
No surprises that the Chinese are pwning the rest of the world in this department.
Click to expand...

Not really true. There are a lot of supercomputers not on this list. I don't know if any are near 93 petaflops, but there are a hell of a lot of them (primarily USA and Europe). Having one big one is just a matter of cost, and they start to become cost inefficient and computationally inefficient above certain sizes (usually limited by power).

Amdahl's law starts to make many smaller supercomputers a more efficient option for most applications (93PFlops is nice, but in practice it's very hard to get close to this peak). Measuring involvement/investment in supercomputing by "how many supercomputers with more than say 10 petaflops does a country have", would show a radically different scenario, especially if the private ones could be incorporated too.
 
Is there any point to this list really? Like cguy says it's a matter spent the most money to buy the most modules to jam together. It's not like there's any novel breakthrough tech that makes one better than the other.
 
cguy said:
Not really true. There are a lot of supercomputers not on this list. I don't know if any are near 93 petaflops, but there are a hell of a lot of them (primarily USA and Europe).
Click to expand...

Not following. It's either true or not. If you don't know if any are near 93 petaflops, how can you say not really true?
https://en.wikipedia.org/wiki/TOP500
The TOP500 project ranks and details the 500 most powerful non-distributed computer systems in the world. The project was started in 1993 and publishes an updated list of the supercomputers twice a year. The first of these updates always coincides with the International Supercomputing Conference in June, and the second is presented at the ACM/IEEE Supercomputing Conference in November. The project aims to provide a reliable basis for tracking and detecting trends in high-performance computing and bases rankings on HPL,[1] a portable implementation of the high-performance LINPACK benchmark written in Fortran for distributed-memory computers. In the most recent list (November 2016), the Chinese Sunway TaihuLight is the world's most powerful supercomputer, reaching 93.015 petaFLOPS on the LINPACK benchmarks.
Click to expand...
 
Sabre Wulf said:
Not following. It's either true or not. If you don't know if any are near 93 petaflops, how can you say not really true?
https://en.wikipedia.org/wiki/TOP500
Click to expand...

I'm not saying that their claim of 93PFLOPS being the fastest is necessarily false (it may be true, but certainly isn't necessarily true - the NSA, Investment Banks, Defence, etc. certainly aren't going to advertise their hardware). What I am disputing, however, is the statement that they are "pwning the rest of the world in this department". If by "this department" you mean "having a single supercomputer cluster that has more PFLOPS than any other", then ok, however, my point is that this is a near meaningless measurement. If by "this department" you mean, "the super computing sector" (which is a more sensible target), then I disagree entirely, since there are countries with far more extensive supercomputing resources by more meaningful metrics.
 
Last edited:
A good personal computer has a performance of around 100 gigaflops
Click to expand...

More like ~2x that. Also, an additional ~4x for each high end GPU and ~40x that if you have a new Tesla board.
 
cguy said:
I'm not saying that their claim of 93PFLOPS being the fastest is necessarily false (it may be true, but certainly isn't necessarily true - the NSA, Investment Banks, Defence, etc. certainly aren't going to advertise their hardware). What I am disputing, however, is the statement that they are "pwning the rest of the world in this department". If by "this department" you mean "having a single supercomputer cluster that has more PFLOPS than any other", then ok, however, my point is that this is a near meaningless measurement. If by "this department" you mean, "the super computing sector" (which is a more sensible target), then I disagree entirely, since there are countries with far more extensive supercomputing resources by more meaningful metrics.
Click to expand...

I provided a link, feel free to do the same ;)
 
It will be interesting to see what they eventually develop for the SKA project. It will generate truly massive amounts of data that will need fast processing.
South Africa could, one day, be home to an impressive "beast".
 
Sabre Wulf said:
I provided a link, feel free to do the same ;)
Click to expand...

One can't really provide a link to private information, but if you do the math, at roughly $1-1.5k per TFLOP (rough going rate for mass purchases), one doesn't need more than $10-15m or so to get into the top 10. This should be a strong clue that the top500 list is far from complete. I personally know of several machines large enough to be in the top ten, that aren't on the list at all.
 
I personally think in the public space the Chinese are absolutely killing it on the supercomputer front for now. HP's new tech might leap frog that when it's ready for prime time. On the private space I think the Chinese are probably lagging quite spectacularly but we'll never actually know
 
Syco said:
It will be interesting to see what they eventually develop for the SKA project. It will generate truly massive amounts of data that will need fast processing.
South Africa could, one day, be home to an impressive "beast".
Click to expand...

For reference right now our national facility is estimated at 800Tflops.
I think there are plans to run data from SKA to that, but who knows now. Storage is an issue.
The handful of local particle accelerators run analysis on it now though.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X