本帖最後由 Puff 於 2011-12-2 12:59 編輯
2M4T 無 L3 真是好難樂觀

ivy 一定會推高入門獨顯性能
iGPU 性能帶來的優勢大幅減少 ...
usei 發表於 2011-12-2 11:07

L3 isn't a must in consumer processor. It won't hurt the performance too much, or it even could benefit it.
When Ivy delivers a 199% performance gain on IGP (compared to HD Graphics 2000), AMD can deliver a 30% also (compared to HD 6550D). It seems Ivy is powerful and have a greater performance gain, ehhh? However, in fact, the performance gap is still around 2x, because the initial value of the +199% is so small.

TOP

提示: 作者被禁止或刪除 內容自動屏蔽

TOP

回復 21# Puff

HD2000 -> 1
HD3000 -> 2 , G520M -> 4
ivy ->3
入門獨顯就要到 6 -> 升 50%
6 相當接近 A6 / A8

TOP

提示: 作者被禁止或刪除 內容自動屏蔽

TOP

提示: 作者被禁止或刪除 內容自動屏蔽

TOP

Llano = Husky + VLIW5
Trinity = Bulldozer + VLIW4

Each VLIW4 group should be as fast as an VLIW5 group (most of the case), so under the same SPU count, VLIW4 core actually has 20% more VLIW groups than the old architecture, which could be translated to 5% to 15% performance increase.

TOP

e+ AMD A8 3850 CPU part 會比 Piledriver CPU part 快幾多?
destiny2009 發表於 2-12-2011 01:00 PM


A8-3850 @ 2.9GHz = A10-38XX @ 3.8GHz
Just similar with Phenom II X4 960T > FX-4100 , Unlock 6 核穩跑 4GHz  連 FX-8150 都收皮

TOP

A8-3850 @ 2.9GHz = A10-38XX @ 3.8GHz
Just similar with Phenom II X4 960T > FX-4100  , Unlock ...
dom 發表於 2011-12-4 13:02


We will see. Trinity cut that relatively slow L3 and 2 mb L2 for each module. Per module wise, Trinity should be better than Bulldozer in desktop app.

TOP