FutureMark will not be accepting NON "WHQL" driver

Have you looked in a mirror lately?

 

Nvidia is suffering from a radically new chip design, which they had when moving from the GF2 to the GF3 - And unfortunately they have hit some very back luck along the way with various factors - However the point alot of the nvidia doom zealots seem to completely overlook is that we're talking about a billion dollar company here...

The way some people talk about the NV30, you would think that the dozens (if not hundreds) of engineers were playing ping pong while the eager beavers at ATI were perfecting the second coming of christ.... and unfortunately - 700 million dollars of R&D was somehow completely wasted and the technology is somehow destined for the scrap heap....

Come on guys, guess again - it was nvidia who shook up the industry with their TNT, and back in the day - they were the saviour of the industry, taking on 3dfx and the like..... now granted the NV3x is having some teething troubles and no doubt this will be worked out - but if you honestly think the card wont be tweaked and pushed to the max in months to come - then you might aswell be in the 3dmark2003 troll demo for your crimes to troll-kind.

Ive owned 3 r300 cards recently, and ill say they were all brilliant cards - and i had very little bad things to say about them.... But this ATI fanboy nonsense is horrendus....

 

Isn't "cinematic" rendering supposed to run at 25~30 fps ?

I support FutureMark in doing this. Making the benchmark more reliable and a fairer game to everyone

 

I have a question.

WHQL means conforming to DX standards and no FP16 right? I read on one of the threads here (sorry, can't remember which one) that WHQL is only tested at 640x480. Does that mean that nVidia still can run FP16 on any other resolutions (particularly 1024x768) except for 640x480, thereby still qualifying for both WHQL and 3DMark03?

 

Hi guys,

Just wanted to drop a note to clarify what actually happened yesterday (a few news sites are giving this a bit melodramatic spin):

First what we've done is following:

1) We've disabled unofficial or 'leaked' driver results based to detonator 42.67/42.68/42.69 from our Online Result Browser (ORB).
2) The results are still available to anyone who's submitted their scores i.e. we haven't deleted anyone's scores and anyone can keep them as reference or compare their own results from different drivers. The result based on these drivers have been excluded them from the project search and compare functionalities.

This is an exception and does not mean that we will be disabling all non-WHQL results in the future.

Second, the reasons for doing such an exception are following:

1) Our policy has always been to encourage use of WHQL certified drivers. If WHQL certified drivers are not available we recommend using shipping drivers i.e. drivers that are publicly available and officially included with IHV's commercial versions of their products.
2) The reason we've made an exception and disabled above mentioned drivers is two fold; These drivers have been given to review web sites as a curiosity in order to prove that 3DMark03 optimizations can be done at driver level. This point has now been made, but results based on these drivers do not belong to our ORB. Result based on shipping drivers or with WHQL certificate are more reflective to the performance that people who purchase hardware will experience.
3) We've taken this action because of constant requests from our users and beta members. We hope that this situation is an exception and we do not need to disable any scores in the future.

Cheers,

AJ

 

I think that sums everything

 

Yup. Very hypocritical of non-beta members to state how "bad" it is for the industry to have synthetic benchmarks, and then produce drivers that they publically admit are optimized for it, and "request" that web-sites limit testing to use those drivers.

 

What, you mean that members who actually participate in the scheme and attempt to abide by the rules that Futuremark request should be happy that a non-beta member promotes drivers that increase scores by potentially dropping rendered elements and lowering IQ?

 

No, the members who PAY futuremark for access to 3dM prior to release so they can have their engineers optimise their drivers for best 3dM performance. All Nvidia have done is do the optimisation after the release!

The latest OFFICIAL developer drivers (43.40) give even more performance in 3dM03 than the baned drivers, and I doubt they would ship drivers to developers that drop elements or lower IQ.

If Futuremark are really independent than they should state a policy of no beta, no developer, no unreleased driversm, and apply that to ATI as well.

 

Joe didn't Ati say the same thing?

whql, i'm just saying for someone who claims to be out of any influence, that just proves the contrary. It's their software and database, they can do as they want, and i don't really care

I have the response to my first question "why now" Thx.

 

Hope you are right on the 43.40. I totally agree with your last part.

 

Whoa.

 

Well, NVIDIA were beta members up until recently - did the code really significantly change in that short a timespan?

However, look at what Futuremark state - they are concerned about lowering IQ to increase performance, that the type of 'optimisations' they are looking for.

Errr, are developers going to be developing 3DMark? Whatever has been done is likely to be application specific, meaning that it wont interfere with other things or work underway.

Bit difficult to police. But, when since 3dm03 has been released have ATI released anything that wasn't WHQL or eventually became WHQL?

 

nVidia had beta access all the up to late last year, certainly a late enough build to do whatever gross "optimizations" wanted.

Um, "official developer" drivers is an oxymoron.

A driver is either officially released to the public, or it's a non-public developer driver build.

And WHQL is another step entirely.

I do agree that ultimately, that should be the goal.

However, this is a very unique case we're looking at here. And it deserves unique action. The drivers in question:

1) Noticably degrade image quality in the game 4 test.
2) Publically admitted by the IHV to be "specially optimized" for the test
3) These drivers are not officially available to for public consumption, particularly to FutureMark, so a thorough evaluation of what's being done cannot be assessed
4) There are later drivers that don't suffer the same image quality issues, and also don't increase performance.

If nVidia can came up with drivers that are at least officially publically available, so that they can be scrutinized to determine if anything "illegal" is going on, I'd be fine with that. Problem is, they haven't.

 

More or less,

nVidia kicked FutureMark in the nuts by putting out lots of PR against them. Many 'average gamers' don't trust FutureMark anymore because of this.

FutureMark is pissed off by nVidia, so they ban nV's silly cheat drivers.


And by the way, I wonder if anyone's tried renaming "3dmark03" to "Quack003". Maybe the 42.68 drivers wouldn't work then!

 

I don't know why everyone is bitching. The real issue at hands is drivers.
Nvidia users are protecting their cards at every oppotunity.

How about this, you E-mail nvidia (the producer of the most ROCK stable drivers) and ask them why they have not released OFFICIAL WHQL drivers for 5 months. Ati hasn't had a problem releasing 6 sets in that time and on a constant and regular basis.

I seriously don't know how nvidia can claim the best driver support when it's been 4 months into a new year and they can only have leaked or beta drivers put out.

 

I'm curious about a few details that I don't think have ever been sufficently answered. At least that I've seen. Can someone confirm?

Has it been confirmed that the Doom3 engine takes advantage of 24bit to give better image quality than 16bit? I know that there are cases where it could make a difference and that Doom3 does use separate paths but it doesn't necessarily mean that Doom3 uses this to any kind of advantage let alone a significant one.

Has any reputable site done a review that actually shows this IQ loss? I've seen links to forums before, but it's hard enough for review sites to make sure the settings are relatively equal and perform accuarte tests.

 

No, and I personally don't think 24/16 bit would make a difference with Doom3 either. It's designed to look "good" with interger math. However, it's other aspects of the ARB2 rendering path that might make some visual difference, as carmack had suggested in his .plan IIRC. Probably minor though.

IMO, any significant difference will likely be between cards that can render in a single pass, vs. multiple passes. I expect there to be a bigger difference between the R200 and NV2x, than there is between the R300 and NV30.

Also, to be clear, we don't even know if the NV30 path in doom 3 is using much floating point at all. It might be more analogous to the R200 than the ARB2 path: R200 path is integer based, but one pass per light.

What do you call "reputable?"

This forum in particular has examined source images, and it's readily obvious. The lighting in game test 4 is not correct. R300, and "non cheat" NV30 drivers produce images very close to the reference. NV30 with cheat drivers don't.

Again, part of the problem is that because the drivers are not "official", I'm sure that prohibits FutureMark from doing their own tests and publically reporting on findings.

For all we know though, FutureMark actually did their own tests with the drivers, and knows exactly what nVidia is doing. However, since the drivers are not public, they'd probably be in legal trouble if they reported on it.