jasonyang 說:
西瓜靠大邊,有寫軟體的都知道,主流的 compiler 就算有開最佳化,default 也都是對 intel 的 cpu 最佳化的,amd 需要下特別的選項。
此外,同樣 adobe acrobat 也沒對 AMD 最佳化,不過 AMD 讓人覺得比較強的是,許多軟體的效能都很不錯,就算輸給 p4 或 dothan/yonah 的地方,也只是小輸,不會一輸就是慘敗,不像 p4 整數運算與 x87 弱,在 server、game、compiler、科學運算上大敗,或是 dothan/yonah x87 弱,一用到就大敗。
比較可能的是對於浮點運算或許沒有用到 SSE/SSE2 來最佳化,不過即使用到了,還是勝負難料,因為 AMD 的 SSE/SSE2 性能表現也不差,由 sciencemark 2.0 上可看出來。
as I mentioned before, I run the test in my 1.8GHz Dothan, and finished it in about 120min. When running the test, the CPU utilization is always 100%. But strange enough, the processor is very cold, the fan of the T42 is not even start. my experience is even when CPU utilization goes up to 70% for a few second, CPU fan will start.
So, this means the processor core is actually idle. it's not a computational intensive benchmark. but I can't explain why Dothan lags so much.
But, the 2.6G Dothan result is not consistent with my 1.8G result. 2.6G Dothan should be a little faster.