Optimization Guidelines QA

Mysterious. Good job. Keep it up.
You must feel like a major burden has been taken off your shoulders now. Getting so many positive responses after the umm.. not so good ones recently.

Your guidelines and especially the good QA is good reason to regain trust in Futuremark even for the most critical voices. I just hope you will stick to your guns and don't disappoint.

 

I really look forward to seeing how this develops.

Good luck.

 

I second the sentiment. I like both the new groundrules and the Q&A, they're clear and precise. And though I actually think more than a month between setting the rules and enforcing them is a stretch, I still can't wait for Oct. 31st.

cu

incurable

 

Worm, has the role of non IHV members changed/expanded with regards to the new rules (and there implementation)?

 

Definitely another step in the right direction. Thumbs up !

 

I'll exapand on this.

I think a one month "grace period" for IHVs getting their ducks in a row is reasonable, and with now "official" guidelines posted, I completely understand the need to give the IHVs some time to do their own internal work (and perhaps soul searching) to comply, before possibly getting put in the situation of having no legitimate drivers to use.

Not that I don't think that some companies deserve that right now, but the facts are, these guidelines were just published.

HOWEVER, I am a bit concerned that the Oct 31 deadline isn't set in stone. I think FutureMark needs to really emphasize that Halloween is in fact "drop dead" date. No exceptions.

It's unfortunate enough that ATI and nVidia (and perhaps XGI as well) will both likely go through another product launch without worrying about "enforced" guidelines for their drivers. Accept no excuses for non-compliance by the 31st. (And make it known to the IHVs now that no excuses will be accepted.)

Come Nov 1, I expect FutureMark to have posted its list of FM Qualified drivers for each chip. Even if that means any particular IHV has none.

 

Yeah, that's sort of my watermark too for how well this is going to work. If the Oct 31st deadline isn't enforced I'll start up with some unhappy noises.

 

You won't be the only one.

 

Well done Worm (and the gang)! Checking all the drivers (even if just WHQL) will probably double your workload, but it's worth the effort! (AFAIK no-one else is doing that)

The Q&A explains a lot and gives a way better understanding about what the guidelines actually mean.

Is there anything we can do to help you enforce those rules i.e. once we get info about more funny business with *cough* *cough* drivers of certain IHV is there any way to bounce that back & forth with Futuremark and ask (request, demand) for action?

- Tom

 

Thank you, Worm, this all sounds pretty well.

One thing is very interesting:

The Det50 drivers are currently using "quasi trilinear" filtering in all Direct3D programs. This will be a major problem for NVidia. Because with that quasi trilinear filtering Det50 will not be approved by FutureMark. Now if NVidia tries to do real trilinear filtering only for FutureMark but not for the rest of the games, Det50 will *STILL* not be approved by FutureMark, because that would only be possible by detecting 3D Mark!!

So as far as I understand it, FutureMark's new rules might enforce NVidia to drop the "quasi trilinear" filtering from Det50 for all Direct3D applications. Otherwise I see no way how the Det50 can be approved by FutureMark.

 

Ahh see now thats the kinda stuff I wanted to see. Nice work and quick too.

Glad to see you answering questions on here too Worm. That makes me feel much better about futuremarks public relations. Moving in the right direction keep it up. Wish I had questions but it looks like they've all been answered.

 

This is interesting. At least the way I understand that you're absolutely right.

Worm, care to comment on this? Is this the reason why you're giving time for the IHV's to err... 'adjust' to the guidelines before enforcing them? (sounds like a certain IHV will have a lot of driver work ahead of them )

- Tom

 

I think the thing that impresses me the most about this Q&A is the wording. I love the diffrence in wording between this and the guidelines. It shows if nothing else you definitely know what your doing.

 

AFAIK, they could still do "quasi" trilinear in the quality/performance settings. Just not in the application setting.

 

For immediate release

Recently, FutureMark Corporation released optimization guidelines. Here at Nvidia Corp, we are dedicated to giving our 54 487 156 237 745 125 756 109 632 451 users the best gaming and benchmarking experience possible with our cutting-edge Detonator drivers, and we fully support FutureMark's guidelines which are totally consistent with our internal recommandations. In fact, slides leaked from Nvidia mention similar guidelines, while slides leaked from our direct competitor mention them being awful cheating scumbags, which should prove that gamers interested in quality (not to mention heavily encrypted) drivers should really go with Nvidia hardware.

Regarding optimizations guidelines, we have asked Ray Charles to comment on the visual quality provided by our upcoming Detonator 50 drivers, and he couldn't see the difference. We at Nvidia Corporation believe this unbiased and professional testimony by a great artist should convince FutureMark that our Detonators don't contain any quality degrading code.

Nvidia Corp is a global cheater in the communication age, and our goal is to "deface every pixel of the planet".

 

Correct. But right now Det50 uses the "quasi" mode even in the application setting.

 

Driver validation process. Keeping consumers informed. Exposing information on validation process to consumers for their own evaluation. In short, focusing on veracity, and letting "political" impact be determined by users.

I love it. You even bit off more than I expected in terms of ongoing validation (the evaluating by 3dmark score range and card clock speed idea was intended to allow some time between "audits").

Good luck in executing all this in a timely and effective fashion.

...

Concerning "shader replacement":

The act of detecting a shader, as would be required for "replacement", is precluded by the rules. I.e., the "how" of ATI's past optimization is still "wrong" for 3dmark 03, as I think it should be.

The act of optimizing a shader based on general rules is not, but this is compilation/scheduling, not "replacement". ATI's stated intent for achieving what they did in 3dmark 03 by improving compilation principles is "right" for 3dmark 03...again, as I think it should be.

I am also still interested in finding out whether nVidia has any optimizations done the "right" way buried amongst everything else. I guess we now know the latest date by which we'll have a good start on determining that.

And future numbers, such as Volari results, as one example, might tell us a lot more about new contenders.