![]() |
|
#8
|
||||
|
||||
|
Quote:
The desktop gets on average 25.2GFlops using 400W. (CPU only) The server gets on average 11.2GFlops using 200W. Desktop: 25.2/400 = .063GFlops/W Server: 11.2/200 = .056GFlops/W Although it's not that black and white. This both proves and disproves what you are saying. It is true that faster and more power-hungry systems (in this case, a 6-core 4.25GHz Phenom II X6) are actually more efficient than large numbers of slower systems (A dual core Athlon II X2 at 2.80 GHz). However, the argument that performance doesn't scale with power consumption is not always true. If I used 50 servers with the same spec, I'd still have the same performance/power ratio, because I'd use more power but also get more performance in a linear fashion. This is how most large computers are in fact constructed, with maybe 500 computers, each containing 2 or 4 CPUs. (With the exception of the IBM System z10, which fits somewhere in between a single box and a small supercomputer, with its 17 CPUs). Last edited by Sight Unseen; 09-21-2011 at 03:49 AM. |
|
|