CPUアーキテクチャについて語れ 11
ReadWorldTech掲示板でTheINQのCharlie "Groo" Demerjianがx86デコードのオーバーヘッドについて
IntelのPat Gelsingerから聞いたという話を開陳しているす。
http://www.realworldtech.com/forums/index.cfm?action=detail&id=92556&threadid=92525&roomid=2
---------------------
I directly asked Pat Gelsinger what the decode overhead was for similar performance on x86 vs
ARM about 2 years ago. He said, in many uncertain terms, that it was within the range of 0-50%
power use, basically you take a 25% extra power hit.
The way it was worded was roughly "at 200%, it wouldn't be worth it, 150%, it would be tough,
100% perfect, and we think we can do it."
It sounded like they did some very extensive modeling of the power costs, and that lead them to
dump the ARM division. :) You will recall that they had a very good ARM division a few years ago,
so they knew exactly what it would take to compete.
---------------------
25-50%の電力効率の不利をIntelの技術でなんとかする。。。ということの様す。
IntelのPat Gelsingerから聞いたという話を開陳しているす。
http://www.realworldtech.com/forums/index.cfm?action=detail&id=92556&threadid=92525&roomid=2
---------------------
I directly asked Pat Gelsinger what the decode overhead was for similar performance on x86 vs
ARM about 2 years ago. He said, in many uncertain terms, that it was within the range of 0-50%
power use, basically you take a 25% extra power hit.
The way it was worded was roughly "at 200%, it wouldn't be worth it, 150%, it would be tough,
100% perfect, and we think we can do it."
It sounded like they did some very extensive modeling of the power costs, and that lead them to
dump the ARM division. :) You will recall that they had a very good ARM division a few years ago,
so they knew exactly what it would take to compete.
---------------------
25-50%の電力効率の不利をIntelの技術でなんとかする。。。ということの様す。
|
|
|