tempter

我張570用左1年   睇黎仲有1年用

Puff

SemiAccurate 表示 Kepler 將會繼續 Fermi 的 SM Configuration Split 的策略，頂端的 Chip 將會是 Compute-oriented (Fermi: 2 CUDA Array per SM)，排行第二的 Chip 將會是 Gaming-oriented (Fermi: 3 CUDA Array per SM, more texture units)。

willy930

SemiAccurate 表示 Kepler 將會繼續 Fermi 的 SM Configuration Split 的策略，頂端的 Chip 將會是 Compute ...
Puff 發表於 2011-12-2 18:40

Semi also said GK102 failed, that's why there is a GK112 on the roadmap.

peterma

回復  ricky1992
睇下幾時有人研發d ram可以自己插入去張display度
其實應該一早就知點搞,但係冇人做 ...
cyy 發表於 2011-11-30 21:48

好多年前的display卡一早有呢樣野啦, 好似係1M-4M的年代, 可以自己加RAM上去, 係後來冇左啫。

dom

你確定 AMD 有 license XDR/XDR2 solutions?
Puff 發表於 30-11-2011 10:28 PM

license 係有
不過唔一定拿去 Discrete Graphic 度玩啦
你唔比佢用去整 XBOX 720 粒 GPU

dom

Semi also said GK102 failed, that's why there is a GK112 on the roadmap.
willy930 發表於 2-12-2011 10:53 PM

又 FAIL ?

idle.man5216

Notice there isn't a GK100...

GF110 A1 stepping = GF100 B1 stepping...
While GK112 has the same specifications listed on the 4Gamer roadmap...

So, NVIDIA fixes the failed chip, and then launches the dual-chip variant. Only this time, instead of insisting on launching the failed chip first, NVIDIA has skipped that and directly go for the fix... The underlying strategy is the same...

So GK110 (and GK112) will still be a "success on first try"...

willy930

又 FAIL ?
dom 發表於 2011-12-3 01:39

Well, depend on which angle you come from.

High end consumer market wise, nV will do something that AMD is doing for awhile by putting two mainstream GK104's together on a same board.

Professional market wise, it will be a fail, GK112 won't be here until Q4, 2012.