New det 50s IQ comparisons at gamersdepot using halo + TRAOD

It no wonder they have sent an NDA out to all the sites saying "Don't talk about features, but mention performance though"

 

All the sites need to respond in one voice....send the drivers BACK to nvidia, saying:

"Don't send us a driver until you're confortable about us talking about any god damned thing we want to."

 

Persoanlly I hadn't intended to look at these until they became official, but I might have to now...

 

You can see why Nvidia went to the trouble of blocking the Anti-cheat tool (would have probably suffered from a buffer overload anyway with these drivers ). It's as though they've decided that these 10's of 1000's customers they refered too are happy playing CS so image quality can't be important.
I think the only way you're going to get a true reflection of the NV35 is going back to pre 3Dmark03 cheat drivers (Was it 42.23??).

 

I'm sure you missed it, but I have already made a deal with Rev....I give you my permission to use the beta det 50s for an article to get to the truth (good or bad) of what they are doing, if other sites use det 50s for benchmark comparisons.

In no other capacity would I condone the use of drivers that the IHV has not released to the public in at least a public beta.

 

I cropped the zone in proximity of Lara's foots, and lighted up by 30%. No need for PNGs, no need for ultra-precise comparisions: the difference is so noticeable to be... funny!

ATI


nVidia


May be they have leaved DOF on for nVidia?

Bye!

 

Yeah, I'll toss you folks my permission for it too. I was pretty much a bit of a prick against the idea with Rev, but with everyone and their mother out there using them anyways I'd really rather have some people I trust and who know what they're doing take a look at them and explain what is going on.

I here by give you my blessing to use the unreleased Det 50s in a review/benchmark situation, I trust ya.

 

Is it like UT2003, though, where getting the in-game settings don't really cover this and you have to edit files outside the game? Which ones actually DO apply trilinear to anything but the first stage? I was under the impression that relatively few actually do, and UT2003 was one of the few examples of games that apply trilinear to more stages in by their own methods.

 

I'm sorry, I thought you were kidding but thought again and thunked I'd best ask...

...you are joking about that aren't you? Or did they really do that?!?

 

Don't bother, Dave: Whatever issue you may or may not find on image quality will be countered by some talk about the drivers being early beta and thus having bugs they are ready well aware of and being fixed 'as we speak' and that super-good-stuff parts were written from ground up and will thus feature high image quality and even better performance at release time. And yada, yada, yada....

I'm not saying that the drivers will be a joke at release time but right now I'm too underwhelmed to think that beyond3d should waste time on them.

 

update from the article of 3dcenter.de

AF on stage 1-7 is only set to 2x.

 

There's a lot of good reasons that Nvidia should be denied any publicity on the Det 5's - they're just trying to distract from the Valve bombshell with the standard "it's a driver bug, everything will be fixed in the next version".

 

Det50 = Xabre TexTurbo

Just Bizzare how Times change.. Wierd.. Just Wierd...

 

Re: New det 50s IQ comparisons at gamersdepot using halo + T

If you look at the picture comparisons in the article, they state between the pictures that the gamma was adjusted, but no matter what the screens always saved that dark.

 

This is true. I have the same NDA. But - talking about IQ is not talking about driver features.

 

Ever since I read about this "not full trilinear"-trick in UT2003 I've had a weird feeling about this particular optimization.
Don't hit me too hard , I know the circumstances are different, but:
Isn't this a similar methodology (slightly lower the filtering quality to improve speed) to what the Radeons do with AF (I mean lowered levels of anisotropy at different angles, not the "only 1st stage trilinear" stuff)?

 

At this point in time, I'd think some insight as to why Newell stated why the Det50s should be used.... this shouldn't be a conclusion to why the Det50s may end up differently when it is official... but... seeing the only PR from NVidia is a new shader compiler, missing fog (for now), and "a commitment" to optimize for HL2... I'd like to see some more analysis on the matter...

Right now.. it seems that the following are being made...

Forced tri-bi mix
Potential force of FP16 (and other work reduction/precision algorithms being done)
Possible sacrifice of lighting

Maybe more...

As stated before.. this would have to be analysis of why Gabe Newell said the Dets50s shouldn't be used... not the end-all statement to why the Det50s should not ever be used (you can do that later, lol)

 

OMFG, they really are in full-blown panic mode!

Thanks for that tidbit [3dc]Leonidas, waaaaary interesting... 8)

 

Jap. And I think, the new nVidia optimization is a reaction on this old ATi optimization (since 03.2). ATi: optimize a little bit more, nVidia: optimize a little bit more, ATi: optimize a little bit more, nVidia: optimize a little bit more ..... 2005 we are on the good old point sampling filter

 

Hang on, ATi's behavior is caused when _forcing_ the anisotropic filtering to be on rather than letting the game select it itself. As I recall their answer when queried on this was that this feature is intended for older games that tend to use texture stage 0 primarily anyway. For newer games that allow you to set anisotropy in their configuration the recommendation has always been to do that and leave the driver set to 'application preference'.

Indeed if you do so the testing with UT2003 confirms that filtering on all stages is correctly enabled exactly as the application asks.

What nVidia are doing here is quite different in that there seems to be, still, no way to set the drivers so that what the application requests is what the application gets. It is this forced on behavior that pushes this away from being a 'response' to ATi and into a full blown cheat.