總之先慢慢修吧....
先別提離NV35的product offical release還有一個月多
(standard--june/value--july)呢.

----
NOTE: nVidia engineers are currently investigating this issue, and will be letting us know what they discover. Once we have that information, we'll give you an update as to what's going on.
FutureMark, the developer of 3DMark2003, has told us that it is also aware of the issue with nVidia's 44.03 driver, and is currently investigating it as well.

大家都在找破頭啦....
總之, 對消費者而言, 出來的東西是對的就好了....

Quoted by Reverend

I'm not sure if I should say something about this in this thread or in the related thread at the 3DHW&T forum. Oh well, here, I guess.

This issue is a real smoking gun and gives the overall impression that IHVs (NVIDIA being the example here) can practically spend hours and hours studying popularly used synthetic and game benchmarks (studying every frame rendered during the demo benchmark run practically!)... and do what they can to boost performance in those specific benchmarks without translating into actual gameplay performance improvements. We're going to give NVIDIA the benefit of doubt and hope that they come back to us with credible answers on why this issue exists only on their hardware/drivers and not on ATis.

While continuous variety in game benchmarks is a possible thing to do to overcome any such similar issues, it then becomes extremely difficult for comparison purposes. Insofar as Futuremark and 3DMark2003 is concerned, while the general feeling is that this is something IHVs shouldn't do (duh!) and/or it's up to reviewers to bring such investigations to the public's knowledge, I have tossed up some ideas to FM wrt 3DMark2003 that are either extremely difficult to implement or it's virtually impossible when the end result is to have indisputable comparison amongst a variety of hardware.

I am not at liberty to reveal some of the things Futuremark is contemplating but I thought I'd just say that Futuremark is indeed looking at this issue.

繼續有問題下去的話,
不利的可能會是Futuremark,
而不是nvidia...
在nvidia仍是多數Game developer標準測試平台的狀況下...(無論是不是Nvidia的問題...)

有點覺得nvidia要搞倒Futuremark的威信不是難事...

當然,這純粹只是小弟神經質的感覺...不喜勿罵...

倒是A&N的策略蠻像的...
DX8時代M$和NVDA合作,
ATI就專心搞Open GL,
DX9由M$和ATI合作,
NVDA就跑去玩Open GL...

不過在市場策略上,
NVDA始終略勝一籌...
ATI的R9500殺敵一萬傷己八千的戰法實在不聰明...(當然user就爽到了...)

如果是如 Redsky大所言的那樣.....沒顯示的地方就省略不畫...我記得這似乎是 GeForceFX 當初的一項節省頻寬的功能的樣子耶.... 省略不會出現在畫面上的部分(所以不算作弊吧.......只要不被看到沒畫的地方就成了... )

現在的問題就是在測試的時候驅動程式是不是已經「預知」(當然是人為使然)
某些場景是不用計畫在內的
這些都還在合理的猜測範圍
3dmark這種固定循著某一個track的demo可以用人為"優化"
但是一般遊戲中可不行(user可依自己自由意志動作)
倒底結果會怎樣我覺得還是先等雙方解釋

我也試著解釋一下, 就是因為在 Game4 中, 在啟動一些如 early Z-tests, occlusion culling 等這些節省頻寬的功能時, 即使脫離原有路徑, 測試者在自由移動時, 這些功能仍然應該卻保顯示方面的正確性, 因為這是即時運算的

但用了 44.03, 這些功能像是突然 "失效了", 而造成顯示錯誤

說得簡單一點, 就是 ExtremeTech 認為在 44.03 下, 這些 clipping/culling 的資料是早被 "準備好的", 而不是運算出來的, 好處就是節省了運算時間, 進而增加分數

牛頭不對馬嘴，一看到NV驅動爭議就急著要罵你這幾段話是吧？
怎麼？從一個你根本拿不到的測試程式上看到問題
閣下自從FX5800後以來的論調又出現，這麼High嗎？

系統Lag多po一篇，刪

嗯....如果照之前的說法是看不到的就省略
那可能是藉由Z軸的資訊來判斷要不要顯示(算圖)與否....
不過沒有一家顯示卡的Z軸效率改進功能會作成這樣的，因為不知道什麼時候視角會切換...
如果真的是發生在只有"開發版本"的3DMark2003，那麼未預期的情況本來就會比較多
==
這篇討論串的作者也真熱心，看到這消息馬上(再?)註冊一個pcdvd會員帳號XD