Futuremark Announces Patch for 3DMark03

Image isnt dropping, or if it is, not by much.

You would actually need to use some form of digital capture device in order to objectively make this determination.

NVIDIA has already been accused of performing a single frame re-render when a frame buffer lock is detected, this invalidating screenshot grabs from comparison. This makes it so even 3DMark's own image comparisons may be comparing non-valid captures.
 
It would be nice if we could, as one example, take the camera off the rail for ourselves and note any performance and/or visual rendering descrepencies that might result with varioud IHV driver versions.

This remains one of the tools Futuremark could conceivably provide us with to make our information about defeated cheats (and the questions about the possibility of cheats that aren't defeated) more complete, but it is one they have apparently decided not to provide to a wide selection of consumers.
 
Forceware (Detonators) 53.03 for win2k/XP that I got the heads up from off of nForceMan over at nVnews.
Driver PR hype department said:
These drivers correct a number of problems found in the 52.16 release of the display drivers for Windows XP and Windows 2000. These drivers are not WHQL certified and you may see a warning message from Windows to that effect. Despite this, these drivers are safe to install and use.

Updates to this driver:

Improved support for HDTV Modes

Corrected the following issues from 52.16:

Clone mode may be set for the GeForce FX 5600 although only one display is connected.

GeForce FX 5950 Ultra: When 4x FSAA is enabled, half of the screen is black in Tiger Woods 2004 introductory video.

GeForce4 MX 440 and GeForce4 Ti 4400, Windows XP: Blue-screen crash during Battlefield 1942 Desert Combat.

GeForce FX 5950 Ultra: Homeworld2 antialiased performance is slow.

GeForce FX 5950 Ultra and GeForce4 Ti 4400, Windows XP: Rendering corruption in Battlefield 1942 Desert Combat.

GeForce FX 5950 Ultra, Windows XP: “X2: The Threat” benchmark is choppy or jerky in some places.

All GeForce FX: Intermittently, Warcraft III videos are not displayed properly on some systems.
I'm just wondering if these "fix" that 13-15% loss FX owners have been getting since the 3dm2k3 3.40 patch came out...
 
digitalwanderer said:
Forceware (Detonators) 53.03 for win2k/XP that I got the heads up from off of nForceMan over at nVnews.
Driver PR hype department said:
These drivers correct a number of problems found in the 52.16 release of the display drivers for Windows XP and Windows 2000. These drivers are not WHQL certified and you may see a warning message from Windows to that effect. Despite this, these drivers are safe to install and use.

Updates to this driver:

Improved support for HDTV Modes

Corrected the following issues from 52.16:

Clone mode may be set for the GeForce FX 5600 although only one display is connected.

GeForce FX 5950 Ultra: When 4x FSAA is enabled, half of the screen is black in Tiger Woods 2004 introductory video.

GeForce4 MX 440 and GeForce4 Ti 4400, Windows XP: Blue-screen crash during Battlefield 1942 Desert Combat.

GeForce FX 5950 Ultra: Homeworld2 antialiased performance is slow.

GeForce FX 5950 Ultra and GeForce4 Ti 4400, Windows XP: Rendering corruption in Battlefield 1942 Desert Combat.

GeForce FX 5950 Ultra, Windows XP: “X2: The Threatâ€￾ benchmark is choppy or jerky in some places.

All GeForce FX: Intermittently, Warcraft III videos are not displayed properly on some systems.
I'm just wondering if these "fix" that 13-15% loss FX owners have been getting since the 3dm2k3 3.40 patch came out...

I think there are a few posts around that state that users have gotten back a good amount of the lost points. Considering the seeming simplicity of Futuremark's changes (even I understood a good amount of what they did), I guess it wouldn't be too difficult to add a second set of shader fingerprints for 3dmark.

Now, my following statement is contingent on the following things being true--if these performance gains in 3dmark are present, and if they have been gained through replacement shaders:

Nvidia driver department--you are hacks. I'm sorry to say, that regardless of whether some of you do not work on pointless driver bloating replacement shaders, you are by association members of a first rate collection of software hacks. Not hackers, who do things exemplified by a mastery of code and elegant examples of cleverness, but hacks in the sense of lousy after school special writers, or a thousand monkeys at an approximately equal number of typewriters.
I'm pretty sure it's not entirely your fault, and that you normally probably wouldn't think about doing this. I'm sure a number of you wish you didn't have to do this. However, that only makes me empathize with people who were unwillingly forced into being hacks.
In addition, most of you are probably stand-up individuals, who would be fun to get to know and invite to some kind of social function, but--and I give my condolences--you have been reduced to digital hacks, with whatever skills and capability you have debased by actions seemingly more appropriate to a middle-schooler peering over someone's shoulder to get an answer to a math quiz.
Regretfully, even should the latest drivers be the best Nvidia ever made (which is somewhat questionable, since they were written by hacks), there will always be the asterisk over your achievement, since you are part of a team of hacks. Hopefuly, there will come a time where you can simply be one of the premier driver development teams anywhere, as opposed to ridiculously overqualified hacks. Here's to hoping your employer gets its act together soon.

edited to change some wording
 
digitalwanderer said:
I'm just wondering if these "fix" that 13-15% loss FX owners have been getting since the 3dm2k3 3.40 patch came out...
Off course they do!

As confirmed by bkswaney here: http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=20892&pagenumber=3

The IQ is "wonderful" even on mother nature.
I got back 700 of 850 to 860 points I lost with the patch from
looking at all my past scores.
So it did not come all the way back up from my figures.
But still got a nice jump with " my eyes " no iq lose.
If anything the IQ is better.

Like I said before: wash - rinse - repeat :rolleyes:

You gotta love those "leaked" drivers..
 
http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=21177&pagenumber=3

Yup. 3DMark scores are back up, PS2.0 scores do not increase according to Shadermark, but there are also some oddities:

1. Ragtroll increase 1.7 FPS (over 10%) between 52.16 and 53.03 while CPU scores remain the same?

2. PS2.0 test *drops* 7 FPS?

3. Vertex shader up 4 FPS (from 13.1 to 17.1--almost 30%)?

Haven't had time to try anything else yet.
 
Just remember:
<center>An optimization must produce the correct image
An optimization must accelerate more than just a benchmark
An optimization must not contain pre-computed state
</center>
 
ahh but the general principle of "sticking our own shader in" will optimize other games ;)

i was thinking about how other companies could do this.. i reckon VIA should release new drivers which recognise that sisoft sandra is being run.. Now since the only memory write which really matters is the last one , the drivers might as well ignore all others.. Just THINK how fast the memory will be reported !!!

-dave-
 
Dave B.:
is that from FM's or NV's guidelines? I honestly can't remember.

other Dave:
don't give them any ideas.. :D
 
DaveBaumann said:
Just remember:
<center>An optimization must produce the correct image
An optimization must accelerate more than just a benchmark
An optimization must not contain pre-computed state
</center>
Yeah, what's your point? Aren't you just glad that nVidia "fixed" their problem? ;)

EDITED BITS: Unchecked "Disable HTML in this post" so Dave's quote worked better. ;)
 
RTCW: Enemy Territory

Anyone have a chance to try ET with these drivers? 52.xx seem to cause a hard lockup.

thx
 
An optimization must produce the correct image
An optimization must accelerate more than just a benchmark
An optimization must not contain pre-computed state

It sounds more like...

An optimization must produce the correct image, whether or not our hardware can produce it normally with the optimization.
An optimization must accelerate more than just a benchmark, especially if its a popular one.
An optimization must not contain pre-computed state, except if it doesn't use our proprietary extensions or special coding to produce a similar effect.
 
Best drivers are 2.08 because it had AA on my TNT 2.

Yeah people are getting like 800 points back.

On the futuremark board someone was saying supposedly the only app specific code that is in the drivers is for vertex shaders as they currently have no vertex shader dynamic recompilation module in the driver.
 
Back
Top