qcmadness

引用:

原帖由 dom 於 2012-4-3 22:06 發表
一 Disable "HT" 就現形

只有  270 級數

e, 有人話慢過270既

qcmadness

引用:

原帖由 dom 於 2012-4-3 22:20 發表
有人話 Trinity 好大進步喎

Single Thread  A10 只係 Athlon II X2 @ 3.4GHz 級數
咁都叫 "進步"
人地 Intel Ivy Bridge 又推多 10% 啦!

你仲想點?
AII x2仲有得推高?
Bulldozer有無得推高?

qcmadness

引用:

原帖由 dom 於 2012-4-4 01:43 發表


After the test , plus reading lots of Bulldozer Review
AMD really "FAIL" , Bulldozer in concept sounds "awesome"
but in reality , AMD they screwed up the front end ........

One module have OK perfo ...

咁我問你
1. 唔用bulldozer, K10仲可以點改?
2. K10仲可以升到幾多頻?
3. 點解bulldozer要decouple INT同FP?

qcmadness

引用:

原帖由 dom 於 2012-4-3 22:30 發表
$ 900 同價位
FM1
A6-3670K 會提供更佳 效能 , 更好的 Value 給你, 真正4核 連 HD6530D , Full load 用電比 Faildozer "Quad" FX-4100 還少
加上 A75 晶片組 可提供原生 USB 3.0  + SATA 3 支援 , 配 SSD 更如虎添 ...

咁我問你, 係咪將bulldozer cut剩1 module
bulldozer就無得再升?

qcmadness

引用:

原帖由 Puff 於 2012-4-5 01:11 發表

Are you talking about cores? This is an issue of how AMD is telling Microsoft to profile its Bulldozer for desktop applications. And the flaw is not about the front-end at all, as one core is still e ...

for certain there is something wrong in front-end

qcmadness

引用:

原帖由 Puff 於 2012-4-5 22:46 發表

I only know that there is some limitation related to AVX-256. Is there anything else?

possible location of problem
1. pre-fetch
2. l1 cache
3. decoder