Chinese supercomputer is the world's fastest — and without using US chips - Polidicks

Just a note that the source's source is top500.org. top500.org being the same organization that listed Craig Wright's supercomputer which to the best of anyone's knowledge doesn't actually exist and may have only been an invention to manipulate tax rebates for research purposes [url]http://www.top500.org/system/178468[/url]

By reading the stats (10 million cores etc) it seems they just went for a shitload of weaker processors instead of fewer better ones. Or am I reading this completely wrong?

[QUOTE=ItsMozy;50561072]By reading the stats (10 million cores etc) it seems they just went for a shitload of weaker processors instead of fewer better ones. Or am I reading this completely wrong?[/QUOTE] Depends on their endgame. You can have insane single core performance in one of these machines and still lose out in the computation game due to the fact that another supercomputer has a shitload more processors going at once, doing multiple instances of the same computation with different input data at once, generating more results at once. The trade offs between lots of weaker cores, and fewer stronger cores are entirely down to what they want to simulate, how many users they want to support at once, and how well parallelised their software is (and seriously, fuck parallelisation, shits HARD). A standard cluster will usually support a decent number of users running their own jobs at the same time. So it needs more cores, a supercomputer may not need to support more users and it could be a much more single purpose machine.

Finally, a computer that can keep up with my porn addiction.

[QUOTE=Demache;50557054]Port Dosbox over but simulate an 80486 machine at transistor level. Then run OG Doom. :v: Truly a noble and useful purpose for this machine.[/QUOTE] now the true test you could give this machine would be how many revenants you could stick in a WAD until the supercomputer freezes with how much processing power this thing churns out, I'm expecting at least a googol

[QUOTE=hexpunK;50561115]Depends on their endgame. You can have insane single core performance in one of these machines and still lose out in the computation game due to the fact that another supercomputer has a shitload more processors going at once, doing multiple instances of the same computation with different input data at once, generating more results at once. The trade offs between lots of weaker cores, and fewer stronger cores are entirely down to what they want to simulate, how many users they want to support at once, and how well parallelised their software is (and seriously, fuck parallelisation, shits HARD). A standard cluster will usually support a decent number of users running their own jobs at the same time. So it needs more cores, a supercomputer may not need to support more users and it could be a much more single purpose machine.[/QUOTE] Parallelisation can be easily solved by using functional language... pure functions can jump from thread to new thread then join and return result. And messaging queue + callbacks can turn hard problems to easy :).

[QUOTE=proboardslol;50558366]The NSA uses quantum computers[/QUOTE] Source? As far as I'm aware, no one has built a true quantum computer yet.

[QUOTE=SashaWolf;50563945]Source? As far as I'm aware, no one has built a true quantum computer yet.[/QUOTE] Yes, as far as we know there exist no quantum computers.

[QUOTE=SashaWolf;50563945]Source? As far as I'm aware, no one has built a true quantum computer yet.[/QUOTE] [URL="https://en.m.wikipedia.org/wiki/D-Wave_Systems"]https://en.m.wikipedia.org/wiki/D-Wave_Systems[/URL] It's early days, but NASA has one

[QUOTE=SashaWolf;50563945]Source? As far as I'm aware, no one has built a true quantum computer yet.[/QUOTE] Google has been (through a subsidiary) manufacturing them for a while now. NASA just bought one, amazon has one, google has a lot of them, presumably the NSA has a lot of them. They're very early, i think the most one has now is 1000 quibits (which is the quantum equivalent of a bit) so we have a ways to go to catch up to current processors in terms of numbers, but they far surpass conventional computers in terms of processing power to size for their specific subset of problems theyre marketed under the name D-wave

[QUOTE=Sableye;50565086]Google has been (through a subsidiary) manufacturing them for a while now. NASA just bought one, amazon has one, google has a lot of them, presumably the NSA has a lot of them. They're very early, i think the most one has now is 1000 quibits (which is the quantum equivalent of a bit) so we have a ways to go to catch up to current processors in terms of numbers, but they far surpass conventional computers in terms of processing power to size for their specific subset of problems theyre marketed under the name D-wave[/QUOTE] d-wave's chips are still worthy of skepticism. Their value for the specific problems they are designed for (quantum annealing) haven't been entirely proven and D-Wave have a thing for blowing out their numbers.

[QUOTE=.Lain;50567757]d-wave's chips are still worthy of skepticism. Their value for the specific problems they are designed for (quantum annealing) haven't been entirely proven and D-Wave have a thing for blowing out their numbers.[/QUOTE] they have been proven to use quantum annealing, the rub is that they themselves aren't quite sure how it works nor does it perform universal calculations yet [url]http://www.wired.com/2013/06/d-wave-quantum-computer-usc/[/url]

[QUOTE=CakeMaster7;50564616][URL]https://en.m.wikipedia.org/wiki/D-Wave_Systems[/URL] It's early days, but NASA has one[/QUOTE] Aren't D-Wave's processors too specific for other uses?

[QUOTE=eirexe;50571366]Aren't D-Wave's processors too specific for other uses?[/QUOTE] From what I can tell, they're not universal yet, but the fact that NASA, Google, Lockheed Martin and other large organizations have bought them despite their massive costs is proof enough that they're on to something.

[QUOTE=CakeMaster7;50575001]From what I can tell, they're not universal yet, but the fact that NASA, Google, Lockheed Martin and other large organizations have bought them despite their massive costs is proof enough that they're on to something.[/QUOTE] Not necessarily on to something that actually works however. It's quite common to look at a competitors technology when researching your own. And D-Wave aren't about to just give that stuff up. Google, etc. are probably using the machines, but I doubt they'd be using them in a serious manner. Mainly investigative.