Interview with Futuremark, Pt. 2 - About The Next 3DMark

That sounds like the main score is derived from rendering a "world" and the camera flys through it, with different rendering techniques dependent on draw distances, interior/exterior, nature/urban etc.

And guessing that only SM3 cards can get valid scores, with all the effects turned on.

Jawed

 

Thanks Nick(FM). So this will be a pretty significant change. Since there aren't any game tests (that are effectively the score), can you tell us how the scoring (I assume there will still be a score, otherwise it wouldn't be a "3DMark" ! ) will be like?

 

It depends on how you interpret what I said. All I can say at this moment is that we don't call any of the tests "game tests" in the next 3DMark.

 

there are no "game tests" anymore. :wink
"game" -> "experience" ?
Or instead of separate tests you have one big scene (with increasing complexity maybe?) - the games of the future ought to have dynamic streamed loading and 3DMark being a game benchmark...

 

...

 

Most of the article went over my head but the screenshots at the end sure brought back some overclocking/benching memories.

 

So, it's just a matter of naming?

Hopefully you'll take the following as no more than my humble observation (and hopefully constructive ones) :

I think there are 2 main reasons why all the "game tests" in all past 3DMarks have failed to gain the credit (they deserve, purely from the amount of work you guys put into them) that {should have been forthcoming) from a site like B3D (by that, I mean a technology-focussed site) :

1) Not enough is known about "3D checklist" performance hits
While the 3DMark whitepapers are good attempts at telling us what each "game test" uses in terms of 3D features, it is very difficult to know what the costs are in terms of individual "important" 3D feature imlemented. Yes, this becomes a bit like benchmarking game timedemos (where there are a variety of 3D features used but reviewers finds it difficult to check various feature costs). But 3DMark is anot a game . Personally, I have always loved the graphics in all the game tests but I just don't know how to make use of them in reviews.

2) "Game tests" are tech demos or cutscenes
An obvious statement perhaps, but for a benchmark that attempts to tell us how "future games" will likely perform, we need something from a gamer's perspective. It could be a first-person view (3DMark99 was the best, if not the only thus far) or a RPG (3DMark03's Troll came very close). 3DMark05 is/was a complete failure in this aspect.

Hope Next3DMark is almost all that Dave expects (and can use!).

 

Reading your post, the first thing that comes to my mind is how people seem to look at the same things so differently. I know that naming certain things can make a huge difference, so that's why we have reworked the whole benchmark. I'm still not ready to spill the beans on what the next 3DMark will have different, but I think it will please as well as the users, also the media professionals. So far the feedback has been good about the changes, but of course we still wait for much more feedback as soon as we are ready to send out copies to the media for review.

I am not only talking about how we name certain things, but the basics of the benchmark. It was a big decision we had to make, since it is something we have never done in any of the previous 3DMark's.

Oh well, since I can't reveal the "mystery" behind the next 3DMark, I'll stop before I slip something.

 

...

 

Matasar.

I posted that pic earlier but then I decided to edit my post because it's unofficial material, as told by Nick[FM] on Futuremark forums.

 

What pic ?

 

What is Cascaded Shadow Maps (CSM)? Is any information available about it? I can't found anything useful in the net.

 

I have few questions?

Question1
R520/G70 perform about 30-35FPS Game Test 2 - Firefly Forest 3DMArk05.
The next revision 3DMark [(SM3.0 which is/are faster then SM2.0)]
--->Will the FPS in New 3Dmark be similar with 3Dmark05 since it will use Full hardware/feature capabilities?

Question2
We now have (Duel Core CPU's) which is only useful in some applications but NOT in games, is FutureMark is about how Future Games will look like or I shall say should. Will they have a game test that will remove bottleneck from CPU by utilizing Duel core tech?

 

Shadow mapping method in 3DMark05


Shadow mapping method in the next 3DMark


Links worth reading
ftp://download.nvidia.com/developer/presentations/2004/GPU_Jackpot/Shadow_Mapping.pdf
http://download.nvidia.com/developer/presentations/GDC_2004/GDC2004_PerspectiveShadowMap.pdf
http://developer.nvidia.com/object/shadow_considerations.html
http://developer.nvidia.com/object/hwshadowmap_paper.html

 

Neeyik is right. We comment a FM interview here, not a D Baumann interview .

Still, I agree with Reverend, Dave does use 3DMark results very carefully. FM does its best to make accurate benchmarks, and I never understood why B3D does not use 3DM scores more. If the 3DMark score is considered politically incorrect (which I don't understand why it would be that), why not use scores from separate game tests. Some years back game test 2 scores from '03 were used in B3D reviews, but I've seen very few '05 numbers

An average score of a number of tests is generally less biased than a measurement of a single test. '05 doesn't even weight these in any way. One could thereby argue that such an average would thereby be actually more valuable than a single test score. This is why I put such an average in my 3DMs.

So what is wrong with the 3DMark score? Is it just the fact that tens of thousands of young dudez are comparing their comp speed using that number? It's thereby teenie, not seriously scientific enough or something? Game bms are definitely a must, but adding the 3DMark score should only make a hw review more complete. I could imagine that many B3D readers would very much like to know the 3DM score of each new gfx card, since they may not be able to run these scores themselves.

- Then again, nowadays I don't need to understand such things anymore (but these days I can post comments like this publicly )

ex-FM

 

As for the rest of your post Patric, Nick's earlier posts in response to mine basically tells us the Next3DMark will not be perceived by the public the same way as all past 3DMarks, and that is crucial and vitally important. If 3DMarkXX wants to allow users to compare their hardware with others by only using 3DMarkXX, then fine. As a "games of the future" predictor, as it has been marketed as by FM, it doesn't have a lot of relevance. I'm guessing that's what Nick meant regarding the naming convention (regarding the "game tests") wrt the Next3DMark.

 

Reverend: Pleased now?

The guys here really can be sensitive sometime. Whatever you post, some negative interpretation will be found.

... just like that other forum where the FM Sales & Marketing VP likes to post
Hard...something

/me grabs some of digitalwanderer's popcorn

 

Thanks. Ditto.

Really ?