GeForce4 Occlusion Culling

[MikeC]

Having had only 4 days to put a GeForce4 preview together about killed me. I wasn't satisfied with my work so I'm going to re-write the entire article and include much more detail. It might take a month or so, but I'm going to release sections at a time. I just don't understand how those guys at Digit-Life can pump out all that info :smile:

Anyway, my most recent test was to underclock the GeForece4 Ti 4600 (300MHz/650MHz) to match the clockspeeds of the GeForce3 Ti 500 (240MHz/500MHz). I then ran a series of Quake 3 benchmarks using maximum quality settings (1024, 1280, 1600, and 2048@16bpp) to find out how effective the enhanced occlusion culling is. Anandtech got a 25% increase in Villagemark by doing the same test.

http://www.anandtech.com/video/showdoc.html?i=1583&p=7

I'm getting ready to call it a night, and thought some of you might want to take a SWAG. I'll post my results tomorrow.

 

[arjan de lumens]

Umm, IIRC, occlusion culling in GeForce3 was effective only at 32bpp and not 16bpp .. ?

 

[Mephisto]

It might take a month or so, but I'm going to release sections at a time. I just don't understand how those guys at Digit-Life can pump out all that info :smile:

Because they can calculate 1+1 and may start writing their articles before Nvidia sends them their "white papers" full of confusing FUD.

 

[Kristof]

Run Villagemark as well and use the latest version at http://www.pvrdev.com that means both D3D and OpenGL ones.

Also see if you can find some kind of reg entry to enable disable the effect (I think one site has been using something like that ?). Would be interesting to see if its really the Occlusion Detection at work or just other efficiency optimisations.

 

[MikeC]

On 2002-02-10 09:32, Mephisto wrote:

It might take a month or so, but I'm going to release sections at a time. I just don't understand how those guys at Digit-Life can pump out all that info :smile:

Because they can calculate 1+1 and may start writing their articles before Nvidia sends them their "white papers" full of confusing FUD.

I find that the information they publish goes far beyond what's contained in NVIDIA's press release material.

 

[echo]

Mike,

I have seen your articles. Basically you try to use many, many different benchmarks, but without truly indepth analysis, pseudo indepth. I don't think you can ever break out of the circle you're in. At least, you have some originality.

On 2002-02-10 12:06, MikeC wrote:

On 2002-02-10 09:32, Mephisto wrote:

It might take a month or so, but I'm going to release sections at a time. I just don't understand how those guys at Digit-Life can pump out all that info :smile:

Because they can calculate 1+1 and may start writing their articles before Nvidia sends them their "white papers" full of confusing FUD.

I find that the information they publish goes far beyond what's contained in NVIDIA's press release material.

 

[MikeC]

On 2002-02-10 19:37, echo wrote:
Mike,

I have seen your articles. Basically you try to use many, many different benchmarks, but without truly indepth analysis, pseudo indepth. I don't think you can ever break out of the circle you're in. At least, you have some originality.

I guess that's a compliment :smile: My motto has always been that it's better to write nothing at all then to write about a topic you know nothing about. I'm certainly not a 3D graphics expert and don't pretend to be, but I think I have a knack for disseminating information that our visitors can comprehend. I do make mistakes however.

One way I tried to break out of the circle is to dabble in OpenGL. I was pretty excited after writing a program to draw, err, render a square :smile: I've been a programmer for 20+ years, although not in 3D graphics. I figured that by learning the basics of OpenGL, I could provide more analysis. But this plan went bad as there were so many other things that occupy my time. I haven't given up, but it's been put temporarily put on hold :smile: