Interview with Futuremark Corporation

[Neeyik]

<a href="http://www.beyond3d.com/interviews/fm05/"><img border="1" src="http://www.beyond3d.com/siteimages/b3dsmall.gif" align="right" width="100" height="66"></a>Over the years, Futuremark Corporation has launched a new version of their well-known application, 3DMark, to coincide with either a new version of DirectX or when there is a big enough range of products that offers support for specific key features within D3D. With three SM3.0 implementations currently on the market (NVIDIA's 6 and 7 series, as well as ATI's X1000 series), there is sufficient hardware support available to warrant a new benchmark that tests these capabilities.

On this basis, we approached Futuremark with a series of questions; the first part of this exclusive interview aspects of their past and present status, as well as goals for the future. <a href="http://www.beyond3d.com/interviews/fm05/">Click here to read the first part of this interview</a>.

[dizietsma]

The next 3dmark seems to be a very interesting product from the hints that Nick dropped. I'm slightly concerned when Nick says

" but the minimum requirement in order to get a valid result is 256MB (i.e. without having to get swapping from system RAM <-> video RAM). The next 3DMark is targeted at high-end hardware. "

I assume you can still run the demo with 128MB, but does this mean you cannot post a score without 256MB or that you can but it will suffer due to lack of onboard vga memory ?


I'm interested to know how Futuremark manages to make it very graphics biased but at the same time make the cpu / ram have some effect as well, I cannot mentally sqaure that circle at present !

Looking forward to it's release, which in which quarter is it ? ;)

[Tokelil]

Nice interview and a good read!

The first part (introduction) ends with an unfinished sentence:
" In the first part of our interview"

[Bouncing Zabaglione Bros.]

I'd like to comment on the use of italics for all the interviewee's replies. It's really annoying and makes it harder to read. You've already got the questions in blue, so do all the answers need to be in italics? At the very least, the questions (being so much shorter than the answers) would have worked better in italics, while the large blocks of answers would have been easier to read in non-italics.

[Neeyik]

For me it's actually the other way round - I find the italics easier to read than the blue text! This is on a 19" LCD @ 1280 x 1024 with ClearType enabled; what are you using BBZ?

[pocketmoon66]

I'm suprised that relief mapping has been rejected. I'm sure we'll see it more and more as developers look for new ways to push hardware and there are implementations that work with curved surfaced...

http://fabio.policarpo.nom.br/relief/index.htm

[Bouncing Zabaglione Bros.]

For me it's actually the other way round - I find the italics easier to read than the blue text! This is on a 19" LCD @ 1280 x 1024 with ClearType enabled; what are you using BBZ?

19 inch Flatron Plus. It's a flat CRT with .24 dot pitch. I'm using Firefox 1.07 at 1280x1024

[Pete]

Great interview, Nick. I'm looking forward to part two.

As for the font, I also prefer italics for the Q and regular fact for the A, for the same reasons BZB mentions. This is on both an LCD and CRT (both w/Cleartype). I think my preference would remain even if you switched to a serif font (which I doubt). In fact, I'm pretty sure this criticism has been leveled before, but obviously nothing's come of it. Maybe it's too much of a hassle to change the site template?

To reinforce the rightness of our position (:)), I submit that most print interviews do as we suggested (from boot to the New York Times Magazine, IIRC). At the very least, you could do away with italics entirely and simply use color to distinguish b/w Q & A.

Oh, and the intro left us dangling, too. Will it be completed with the second half of the interview? :)

Sorry if I'm spending too much text OT. Maybe we could start (and link to) a thread in Site Feedback for every article, just to keep the discussion focused?

[Geo]

Sweet job, Neeyik. I'm impressed that's just a "part one".

I'm also starting to get a wee bit excited about all the Signs and Portents re the next 3dm --if part two ends with something like "so see you in 2007 when we actually release it!" I'm going to be majorly bummed. :razz:

[Geo]

Thinking about this some more, the answer that made me the most curious was this one:


Well, the next 3DMark will require 256MB of VRAM for all the graphics tests, so I don't think we will encounter that problem. Of course users with 128MB cards will be able to run it, but the minimum requirement in order to get a valid result is 256MB (i.e. without having to get swapping from system RAM <-> video RAM). The next 3DMark is targeted at high-end hardware.



So you can get a score with a 128mb board, but it is not "valid" because you start relying on the pcie bus and system memory at that point? I'm struggling with that a bit. Why is that not "valid"? Isn't it the case that those folks are likely to have the same problem with the games they play and therefore the result of that degradation is also a valid predictor of the suitability of their current card for those games?

Also, I found myself wondering if the FM guys have thot of using the typical gaming solution to this problem --scalable IQ settings. Lastly, to be forward looking, which seems to be the goal, shouldn't we want to be stressing 512mb cards at this point, rather than 256mb ones?

[Jawed]

I think the whole point with the "validity of a 128MB card's score" is that, for example, a 128MB X1600Pro will suffer quite a hit in this test as compared with a 256MB X1600Pro.

Surely, going back through previous versions (05, 03) there was a minima for memory, below which a card was just asking for trouble. 64MB? 128MB?

There's no point in having scalable IQ options in a graphics test - the bloody shadowing "option" in 05 is enough to invalidate the entire test in many people's eyes anyway.

Jawed

[Neeyik]

So you can get a score with a 128mb board, but it is not "valid" because you start relying on the pcie bus and system memory at that point? I'm struggling with that a bit. Why is that not "valid"? Isn't it the case that those folks are likely to have the same problem with the games they play and therefore the result of that degradation is also a valid predictor of the suitability of their current card for those games?

By "not valid", I should imagine they mean the performance measurement would now include factors, such as system memory settings, etc outside what 3DMark is supposed to be testing (ie. the graphics card) and since all of these variables are difficult to record and quantify, it makes comparisons of scores on the ORB less reliable.


Also, I found myself wondering if the FM guys have thot of using the typical gaming solution to this problem --scalable IQ settings.

They used to, in some ways, with the low detail and high details tests in 3DMark2001 - the reason, IIRC, for dropping the low detail tests is their contribution to the overall score was always less significant than the high detail ones.


Lastly, to be forward looking, which seems to be the goal, shouldn't we want to be stressing 512mb cards at this point, rather than 256mb ones?
What would that do to the download though? It's likely to be huge as it is!

[Geo]

Well, I always figured the entire point of a new 3dm was to make me personally go "jaysus, I need a new card!" anyway. . .:lol:

[Neeyik]

Having just noticed my name splattered everywhere with this, I would like to point out that Richard (aka Mordenkainen) contributed significantly to this interview so he should get an equal share of the compliments!

Edit: I should also like to point out that the date on the article is a bit borked! It's 100%, absolutely not 13 October 2004 :wink:

[Reverend]

Patric left? Didn't know that.

BTW, how many games (shipped, to be out) use PSM, like it's used in '05?

[Neeyik]

Patric works for Nokia now, can't remember the exact date he left FM but it's been a good few months (perhaps longer).

[Fdooch]

Anyone have an idea what the new shadow mapping algorithms is?

[Rolphus]

This is a totally off-the-wall prediction without a single piece of knowledge or fact behind it, but I'd be moderately unsurprised if they went with oversampled or otherwise-blurred stencil-buffer shadows, seeing as, with flow control, you could optimise away a fairly good proportion of the cost by only blurring or fading the penumbra (edge) of the shadow.

That would be my guess, as it could look bloody amazing, and fits with their general tactic of taxing the hell out of the highest-end cards. Plus, it's something that I believe is very much a >SM2.0 feature... perhaps there are ways to make it work on SM2.0, but I wouldn't like to bet on it...

[SlmDnk]

Second part coming up soon?

[Neeyik]

Yes, should be soon - just waiting for some bits and bobs to come back.

[Geo]

Status update?