Omnimaga
General Discussion => Technology and Development => Other => Topic started by: fb39ca4 on December 13, 2010, 09:47:47 pm
-
Check out this (http://techeye.net/hardware/us-airforce-builds-supercomputer-out-of-1760-sony-playstation-3) article, on the Air Force constructing the 34th fastest supercomputer in the world for only 2 million dollars! This would be handy for rsa cracking!
-
Wow, that is awesome O.o I wonder if it was jailbroken with a TI-84 Plus ;D
-
so... anyone have 1760 PS3's lying around ;)
-
Yep, need 'em?
-
so... anyone have 1760 PS3's lying around ;)
We should make it 1761, so we have bragging rights over the military :P
-
We should make it 1761, so we have bragging rights over the military :P
++
-
I've always thought about this. (Assuming infinite amount of money)
Create a motherboard with 4 4GB slots of RAM (16 GB), with 4 quad-core CPU's, and a 2TB internal hard-drive, 5 network plugs in the back, with two usb plugs in the front. If you could deal with the immense heat, you would essentially have an awesome super-computer module.
Then, you could get a "tower" of those (16 of them stacked on top of each other), with a box at the bottom that combines and sorts all 80 network cables (5 from each box, 16 boxes) into 9 network cables by deciding which information is most relative. Then, you could get a "row" (8 towers put side by side) with a box at the end combining the 64 cable input (Technically 72, but the other 8 are special) into 8 network cables. The 16 open network cables combine together into 1 cable that feeds into the "aggregation unit".
So, say I wanted to start a program running on all cores. I could set my program up to be spread out over many cores, and the "aggregation unit" would then get most of the data. (I say most. Really it would be all of it, as decided by the row's and the tower's box that combined the network cables what information was needed.) The aggregation unit could then look at the data, and give further instructions if need be. I also say that the boxes could filter out information, but so could each individual module.
But wait, there's more! Assume for a second that it's all simplified, and that there's only one network cable out of every module, that goes into a box that only outputs another network cable that goes into a row's box, that is then given to the computer (We'll call this entire set up a "computer"). Assume also that a single CPU can go through 10,000 numbers in a second. That means that an entire box can go through 16*10,000=160,000. Also assume that N/sec means that the computer gan go through N numbers per second.
Okay, so assume that we have 256 "computers" (1 computer = 8 towers = 128 modules = 20,480,000/s) which means that 256 computers = 5,242,880,000/s. But wait, there's more. Assume that 256 "computers" equals a "room" (Okay, so maybe a foot-ball field, but it's shorter in name length). Actually, if you assume that a "room" covers an area of approximately 50 ft x 50 ft x 50 ft (Rough estimate, just to give an idea.) Say 65,536 "rooms" made a "country". That means a country can computer 343,597,383,680,000/s. So let's say that 65,536 "countries" equals a "nation". That means a nation can compute 4,071,254,063,142,930,000/s. So, assume 65,536 "nations" equals a "world". That means a world can compute 266,813,706,282,135,060,480,000/s.
So, after assuming all that, I hope you assumed I would come to some conclusion. Well, I have. As far as I can tell, I don't think we'll be allowed to build this thing. (Please don't hate after reading all of that, and hoping for some awesomeness at the end. Also, the game. ;-)
-
How many PS3s would we need to factor a 1024 bit RSA key in one year? :P
-
@Graph: O.O
-
How many PS3s would we need to factor a 1024 bit RSA key in one year? :P
I'd need to know something along how many numbers they can factor in one second. But probably around 2^500 PS3's if they could factor 1024/s. I think, my math could be awful there, but I'm pretty sure it's correct. (Granted, that's in one second.)
EDIT: In a year, it would be 4.15141×10^157 computers.
-
How many PS3s would we need to factor a 1024 bit RSA key in one year? :P
I'd need to know something along how many numbers they can factor in one second. But probably around 2^500 PS3's if they could factor 1024/s. I think, my math could be awful there, but I'm pretty sure it's correct. (Granted, that's in one second.)
EDIT: In a year, it would be 4.15141×10^157 computers.
Psh, we can do that O.O There are already 1×10^9 computers in the world, anyway. And since the number of computers will double in 7 years, we'll get there in about log(4x10^157/10^9)/log(2)*7=3455 years...
EDIT: The TI-Ncourage'll be discontinued by then :P
-
wow... at this rate a 1024 bit key won't be factored within the next 100 years o.o
-
How many PS3s would we need to factor a 1024 bit RSA key in one year? :P
I'd need to know something along how many numbers they can factor in one second. But probably around 2^500 PS3's if they could factor 1024/s. I think, my math could be awful there, but I'm pretty sure it's correct. (Granted, that's in one second.)
EDIT: In a year, it would be 4.15141×10^157 computers.
Psh, we can do that O.O There are already 1×10^9 computers in the world, anyway. And since the number of computers will double in 7 years, we'll get there in about log(4x10^157/10^9)/log(2)*7=3455 years...
EDIT: The TI-Ncourage'll be discontinued by then :P
But you also have to take into account the performance increase of new computers. Also, there are 6 cores available on the ps3 processor, and from what I've read, the ps3 has 20-30x the flops of a Pc.
-
Plus, aren't you using the crummy TD method? :P I can think of better brute-force methods alone than -2, check, -2, check.
Also, the ps3 using the OtherOS feature only allowed you to use 2 of its six cores, and I believe that 20-30x was based on the two cores, not all six. :) Now, If you simply JB it and install linux then, you should theoretically get 120-180x the flops of a computer :P
-
Would factoring the RSA key require a lot of RAM? I heard that the PS3 only got 256 MB...
-
Not really, no. Prime number stuff would only require the amount of RAM necessary for storing the two numbers (Prime and what you're testing) and whatever your multiple precision math function takes. Maybe 10 MB max depending on your numbers and MP function.
Sieves would probably take more though.
-
Not really, no. Prime number stuff would only require the amount of RAM necessary for storing the two numbers (Prime and what you're testing) and whatever your multiple precision math function takes.
Nobody factors large numbers by trying to divide by all possible primes. If you turned the entire Earth into a giant supercomputer cluster, with every gram of matter being another processor, and let it run for a trillion years, it still could not have factored even a measly 400-bit key this way. No joke.
Sieves would probably take more though.
Indeed. The recent factorization of RSA-768 couldn't even fit in RAM, it took about 5 terabytes of disk space as storage.
-
Oh, I meant semiprime, as in the public key, not all the primes.
Shudder at that disk space.
Here's an Idea. When I become fabulously rich i'll go buy it from TI :P
-
Oh, I meant semiprime, as in the public key, not all the primes.
Shudder at that disk space.
Here's an Idea. When I become fabulously rich i'll go buy it from TI :P
For the right price i'm sure TI would sell you the RSA key, but soon after they would release a patch changing it. <_< I have an idea, how about when you become fabulousy rich you buy TI. :hyper:
-
Oh, I meant semiprime, as in the public key, not all the primes.
Shudder at that disk space.
Here's an Idea. When I become fabulously rich i'll go buy it from TI :P
For the right price i'm sure TI would sell you the RSA key, but soon after they would release a patch changing it. <_< I have an idea, how about when you become fabulousy rich you buy TI. :hyper:
Oooh, much better idea.