Wow, what a powerful computer! I would love to have this to play Call of Duty at 60 fps.
Firephinx
If anyone was wondering what the first 2 supercomputers were, #1 is located in Wuxi, China and called Sunway TaihuLight. It uses 40,960 Chinese designed chips that amount to a total of 93.01 Petaflops of computing power, which is more than 5x as powerful as the 17.59 Petaflops that Titan has. #2 is located in Guangzhou, China and is called Tianhe-2, which has 33.86 Petaflops of computing power.
PhiLo
Since supercomputers have so many cores, I'm wondering what's its efficiency. I expect it to be very low because to reach full utilization you have to make all ALUs on all cores on all machines to be working. I guess this won't be very often. Besides, there must be huge synchronization and communication overhead.
asd
Just a thought: While parallel processing is definitely the way to go to speed up our computations, and there is a lot of potential in this domain, what would be finally limit the amount of parallelism that we can achieve in the future other than the sequential bottleneck, that would coerce us to look for even newer methods?
What comes after we have exhausted multi-core, SIMD, parallel compilers?
Wow, what a powerful computer! I would love to have this to play Call of Duty at 60 fps.
If anyone was wondering what the first 2 supercomputers were, #1 is located in Wuxi, China and called Sunway TaihuLight. It uses 40,960 Chinese designed chips that amount to a total of 93.01 Petaflops of computing power, which is more than 5x as powerful as the 17.59 Petaflops that Titan has. #2 is located in Guangzhou, China and is called Tianhe-2, which has 33.86 Petaflops of computing power.
Since supercomputers have so many cores, I'm wondering what's its efficiency. I expect it to be very low because to reach full utilization you have to make all ALUs on all cores on all machines to be working. I guess this won't be very often. Besides, there must be huge synchronization and communication overhead.
Just a thought: While parallel processing is definitely the way to go to speed up our computations, and there is a lot of potential in this domain, what would be finally limit the amount of parallelism that we can achieve in the future other than the sequential bottleneck, that would coerce us to look for even newer methods? What comes after we have exhausted multi-core, SIMD, parallel compilers?
http://computer.howstuffworks.com/worlds-fastest-supercomputer.htm
An interesting view into what is the world's most powerful supercomputer used for?