The fastest supercomputer in the world named (November 2011)

I can see Apple fan boois wondering why Mac OS was not used.
 
so what do they do with it, thats what I want to know. Solitaire must run lekka on those windows boxes.
 
haha, Really.......what do they use these things for? U need a whole building for it!
 
It is a pity this article does not list South Africa's CHPC (Centre for High Performance Computing), as it has made the list of the top 500 at place number 329. See below:

http://www.top500.org/list/2011/11/400

To get a place in the top 500 is an achievement in itself, never mind being number 1! To be at the top you need to do really ridiculously expensive things!
 
Fujitsu K? I eat Fujitsu K for breakfast! :twisted:

10x0928jhib74wefdvs.jpg
:)
 
No specs on the power consumption of these mega thirsty showpieces?
 
saksakgp said:
how long till we get this sort of power in our smartphone?
Click to expand...

I would do the maths, but I'm feeling mega lazy.

Moore's law says that processing power doubles every 18 months.

10.51 petaFLOPS = 10 510 000 000 000 000 FLOPS
1GHz Snapdragon = 2 100 MIPS (million instructions per second)

I'm not sure here but "Assuming" FLOPS ∝ MIPS (Proportional to)
Then I'm guessing 2 100 MIPS ≈ 2 100 000 000 FLOPS

So... For every 1 FLOPS your smart phone does the super computer will do 5 004 761.9048 FLOPS

To Excel!!!
Z8W0O.png

Excel File Download:
Google Docs Link

Take line of best fit to Wolfram Alpha...
Graph Plot on Wolfram Alpha
...and Solve at y = 10 510 000 000 000 000

therefore x ≈ 23.3268 eighteen month jumps
therefore x in months = 23.3268 * 18 = 419.8824 months = 34.9902 years ≈ 35 years

SOOOOO..... (if my maths is any good)
by around 2046 or so we should have smart phones as powerful this beastly super computer.
 
I got a bit carried away and did the maths anyway.... :D

Edit 1: Okay, wow. It just sunk in what this means if my maths is right... That'll be like having a death star to make a phone call. Dammit time go faster! 2045 now!
 
Last edited:
bored at 3am? :)

even if your assumption of FLOPS ∝ MIPS is correct I think 34 years is a bit optimistic ... looking at PC's over the past 30 years saw us moving from 2.5 MIPS (286) to 180,000 MIPS (Core i7 3960) which represents a 72000 times increase in MIPS performance (wiki)

Apply same logic to today's 9,900 MIPS Cortex A9 mobile processors and multiplying that number by 72,000 only gives me 712 800 000 000 000 FLOPS (once again using your assumption) which if you compare it to 10.5 petaFLOPS shows that it is at least 2 orders of magnitude slower!

That is if any of the assumptions used is correct in the first place :)

Edit: Hmmm, applying same logic to Cortex A15 quad core processor does get us 1/4 of the way there! (2.5 petaflops on my cellphone does sound awesome ... just not sure how it will help me with Angry Birds!)
 
Last edited:
Can someone put a place marker on this thread for year 2045. ;)
 
Deezil said:
bored at 3am? :)
Click to expand...

No, procrastinating hard.

Deezil said:
even if your assumption of FLOPS ∝ MIPS is correct I think 34 years is a bit optimistic ... looking at PC's over the past 30 years saw us moving from 2.5 MIPS (286) to 180,000 MIPS (Core i7 3960) which represents a 72000 times increase in MIPS performance (wiki)

Apply same logic to today's 9,900 MIPS Cortex A9 mobile processors and multiplying that number by 72,000 only gives me 712 800 000 000 000 FLOPS (once again using your assumption) which if you compare it to 10.5 petaFLOPS shows that it is at least 2 orders of magnitude slower!
Click to expand...

You also need to take into account that it is an exponentially increasing function of power. So straight division won't work in this scenario. The function I derived is y = 1*10^(9)*e^(0.6931*x) where y = 1.051*10^(16). Because processing power doubles every 18 months it must be an exponential function, similar to how the Fibonacci sequence grows in size relative to number of steps.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter