NVIDIA GT200 Rumours & Speculation Thread

Pretty much everyone has had cards since around Editor's Day (or whatever the nV event was called). If they were there they got one and had ample time to test. Since printed tends to go out on fixed dates, they had the choice of either coming out now or waiting another month in order to follow the 16th NDA lifting date. They might've been cut some slack by nV too.

 

Well, you're a nut, so why should we believe you?

 

hahahaha

 

Time to update your sig Geo, a F@H client has been sighted.

 

He was talking about GTX280 SLI...

 

edit: Yes it should be cuda based, the slides from the presentation however only show g200 results.

You still wonder what superb kind of developer relations causes a delay of almost two years.

 

It looks like all G80 and later GPUs will work, which is pretty cool.

Jawed

 

This is new ?

http://www.pcgameshardware.de/screenshots/original/2008/06/1040472.jpg

http://www.pcgameshardware.de/screenshots/original/2008/06/1040491.jpg

http://www.pcgameshardware.de/aid,6...enu=browser&image_id=838197&article_id=647500

 

No ;(

But this?
http://www.tweaktown.com/news/9668/medusa_stars_in_nvidia_s_upcoming_gt200_tech_demo/

 

Why Germans?
Here is the source: http://publish.it168.com/2008/0611/20080611033801.shtml

But if you mean "why did they bother to write a news about this chinese test", you know, every visitor means money.


€: I should mention, that the most Germans don't read strange chinese websites with strange letters, but they read PCGH and Computerbase and some german forums.

 

Updated.

I'll celebrate when I can delete that line from my sig completely. . . .

 

Wont be much longer. I'm not entirely sure the nature of the standford NDA. But its been distributed to the press already and Nvidia has working drivers for it.

 

Every game and benchmark they tested did have at least one piece of data at 1920x1200...

 

Here is a summary of scores:

9800GTX/HD4850/HD4870/GTX260/GTX280:

Lost Planet (DX10)
Test1 (19x12): 48/41/43/69/86
Test2 (19x12): 63/54/56/63/63
Test1 (12x8): 94/69/75/128/147
Test2 (12x8): 63/61/61/63/63

Call of Juarez (DX10)
Custom Demo (19x12): 19.7/29.3/32.5/27/31.1

World In Conflict (DX10)
Very High (19x12): 17/24/26/30/32

3dmark Vantage
Extreme (19x12): X2096/X2573/X2773/X3717/X4817
Performance (12x10): P5549/P6235/P6688/P7858/P9267

Summary of this data:
--> 48xx cards are very strong in Call of Juarez
--> 48xx cards are good in World In Conflict, but not as strong as GTX 260/280
--> 48xx cards are poor in Lost Planet compared even to 9800 GTX
--> Not very much difference in performance between 4850 and 4870, even in 3dmark Vantage (performance difference is always less than 10% between the two cards)

 

Seems like the drivers need more work if those numbers are to be trusted.

 

Yeah, probably. I just can't believe that there would only be 10% or less difference in performance between 4850 and 4870...

 

There should be at least 15-20% difference between 4870 and 4850, due to clock alone, because the bandwidth is 90% higher in 4870 so it should not limit anything. So only a 7% scaling is very suspect (in a generally very scalable benchmark). Probably a driver problem (or maybe they are testing a 4850 in overclocking, in this case it will be probably very bandwidth limited).

 

Yeah, we'll have to wait for more comprehensive reviews coming early next week to get a better picture I think

 

I still think COJ is a very biased benchmark that shouldn't be trusted at all.