Futuremark: 3DMark06

[JasonCross]

Everyone, stay tuned. I've been talking with Futuremark and ATI and have cleared up a couple of things on the whole 24-bit depth stencil texture issue. There should be an update to the article (second page) soon.

I'm still a little concerned that the benchmark doesn't use some form of parallax mapping technique - it's already in FEAR, it's a major feature of UE3, it's going to be in a lot of SM3.0 games, and it's not in 3DMark06. It seems like one of the defining shader effects of 2006-era 3D engines to me.

 

[digitalwanderer]

Can you define "fine" for me then? I would presume everyone has an own definition of "fine", agree?

Well, mine run fine too and I was happy as hell with my rig....up until I ran this new benchmark.

Now I can see that not only does my rig have limits, but if games start to appear that are as demanding as this benchmark I'm gonna need to upgrade a bit.

Then again, that's half the reason I like 3dMark....I kind of missed the upgrade jones.

 

[Razor1]

Everyone, stay tuned. I've been talking with Futuremark and ATI and have cleared up a couple of things on the whole 24-bit depth stencil texture issue. There should be an update to the article (second page) soon.

I'm still a little concerned that the benchmark doesn't use some form of parallax mapping technique - it's already in FEAR, it's a major feature of UE3, it's going to be in a lot of SM3.0 games, and it's not in 3DMark06. It seems like one of the defining shader effects of 2006-era 3D engines to me.

Parrellex mapping usually isn't used on all surfaces. Mostly used for areas like rocky (mountain) areas and bricks or tiled surfaces its very difficult to use parrellex on unwrapped objects, the shader if not made right gets some very very nasty results (pretty much you have to turn down the parrallex effect down to the point where its barely even noticable, (like around .05 pixels). Also Parrellex bump mapping isn't much more expensive over regular normal mapping think maybe 1 or 2 more instructions at most 4 more.

 

[Nick[FM]]

I'm still a little concerned that the benchmark doesn't use some form of parallax mapping technique - it's already in FEAR, it's a major feature of UE3, it's going to be in a lot of SM3.0 games, and it's not in 3DMark06. It seems like one of the defining shader effects of 2006-era 3D engines to me.

POM might be useful in certain situations/games, but we didn't see any use of it in 3DMark06's graphics tests. Though we try to add all known (and unknown) effects into 3DMark, we can't do them all. We didn't find any use of POM in 3DMark06 and thus we dropped the idea. We found other effects which are much more "in-your-face" than POM would have been, and we wanted to focus on them instead. The heterogeneous fog is a good example. Though engines (UE3 as an example) support POM, it doesn't automatically mean every game developer will use it. Supporting is one thing, using it is another. I don't downplay POM here, so don't get me wrong. I am just saying that we didn't find use for it in 3DMark06 though it can be very well used in other applications/games.

 

[Ragemare]

169 3D Marks

The only benchmark that scored above 1fps was return to proxycon.

Radeon 9800 pro 128mb
2800xp

If the size of your 3d mark has such a profound effect, as to make your ePenis smaller then I think I now have a eVagina.

 

[Nick[FM]]

If the size of your 3d mark has such a profound effect, as to make your ePenis smaller then I think I now have a eVagina.

A good one!

 

[Apple740]

/me runs to install the 6.1 drivers!

no difference for me vs Cat5.13. (X1800XT)

 

[Rys]

So because your art direction was already set, you couldn't implement the technique because no scene required it? While deep freeze is graphically very impressive, that might have been your candidate for a different look and set piece that included some of the techniques people are asking for. You can't undo what's done, though.

I get the very distinct impression that this release of 06 was very much an in-house thing, with little to no consultation and discussion with your major industry (non-IHV) and media BDP partners.

Is that correct? I know I didn't do anything this time around, but it seems like nobody else in the same boat as me did much either.

Jason, did you push for parallax mapping's inclusion during the 06 development process, as part of ET's participation in the BDP? Wavey, did you do anything for 06? Hanners?

Seems like a waste of resources to push out a (knowingly, I bet) controversial update to 3DMark this close to Vista and D3D10, and it smells IHV-led, on the face of it.

 

[Kombatant]

Now I can see that not only does my rig have limits, but if games start to appear that are as demanding as this benchmark I'm gonna need to upgrade a bit.

Trust me Dig; I can't think of a single system out there that can claim that it doesn't need to upgrade after seeing the new 3dmark in (slow) motion

 

[Entropy]

Thanks! I take that as a compliment.

Have you tried all our official mirrors yet? I know most of them are getting hammered, but some seem to work pretty ok (Guru3D and MajorGeeks).

Since I bought it, I could download it from your own site. Nice and quick.

It was a compliment.
I do believe you help raise expectations for the future among many people, and thus help drive the industry forward. It could be said that you are advertising for the hardware manufacturers, and there's a lot of truth in that. But you make some damn fine-looking and polished benchmarks, and if your benchmarks make people desire more powerful hardware, well, it reflects on your product too. It's the best at what it does.

And I still haven't seen trees in any game that move as nicely as even the Nature test in 3DMark01.

 

[Dave Baumann]

Jason, did you push for parallax mapping's including during the 06 development process, as part of ET's participation in the BDP? Wavey, did you do anything for 06? Hanners?

As far as I'm aware press haven't been part of the input process since AJ left.

And coming to that there does appear to have been a number of changes with the outcome of the benchmark since then (i.e.: letting D3D requirements be the centrepoint of the requirements, not using optional features as requirements / highly decisive in the scoring, having single code paths and not multiple paths per vendor, using features that more than one vendor support, etc.). Many of these principles rooted 3DMark in some kind of solid basis with D3D being the arbiter, however now it seems a little confused as to what it is with all kinds of support for this, that and the other, with some seemingly arbitrary decisions. At the moment, with all the different options and requirement and stipulations I'm at a loss to understand what its actually telling anyone.

 

[Jawed]

Seems like a waste of resources to push out a (knowingly, I bet) controversial update to 3DMark this close to Vista and D3D10, and it smells IHV-led, on the face of it.

I think the link to NVidia papers was all the clue we needed:

http://www.beyond3d.com/forum/showpost.php?p=622641&postcount=54

Having said that, the shadowing quality (which isn't entirely down to CSM - but also due to the filtering) really does seem to be "next-gen". In fact, there's an argument for saying it's next-next-gen. I like it a lot (from what I can tell) and NVidia's "gamble" with support for a far-off format (features first, performance later) plays out in an interesting way.

As far as I can tell, ATI cards are doing less filtering on shadows. Is this correct?

Jawed

 

[fallguy]

Can anyone answer me as to why my SLI'd GTX's are getting just about 52xx? I just swapped out a DFi SLI board, for a Asus one, and SLI is working fine in CoD2. Which is all that matters I guess, but Im a little confused as to why its not working in 3dmark06.

The drivers have a built in profile for 06, but the score seems to be well below what it should be.

Worm?

 

[BrynS]

Scott goes to the birds, or rather they go to him. There's a hint of a possible nV physics outlook as well.

 

[Dave Baumann]

Nick, can you find out a little more about the conditions that Fetch4 is used in please. I'm genuinely trying to find a useful Fetch4 test, and was asking about 3DM's use and I got a rather confusing message back that Fetch4 is only used in the Shader Model 2.0 tests - if thats the case, whats the reason for not using it on the SM3.0 tests (especially as the only ATI hardware that has Fetch4 is SM3.0 and ATI's actual SM2.0 parts need to use a different algorithm because they don't support Fetch4!).

 

[Hanners]

Jason, did you push for parallax mapping's inclusion during the 06 development process, as part of ET's participation in the BDP? Wavey, did you do anything for 06? Hanners?

Nope, I only got to see anything of 3DMark06 way into the development (as in around Christmas time), so the feature set was very much finalised.

As for Fetch4 only being used in the two Shader Model 2.0 graphics tests, that's certainly the impression I got of its use from the white paper.

 

[Neeyik]

No "impression" about it at all - it's plain to see in black and white:

The two HDR/SM3.0 graphics tests in 3DMark06 use a 16 sample kernel that is randomly rotated for each pixel. By using the 16 sample rotated kernel, we are able to produce very high quality smooth edges for the shadows. Point sampling is currently used both for hardware shadow mapping and R32F depth maps.

The SM2.0 graphics tests use a 4 point sample rotated grid kernel to produce relatively smooth edges for the shadows, unless the hardware supports D24X8 Percentage Closer Filtering (PCF) or DF24 & FETCH4. For hardware with support for PCF or FETCH4, only one bilinear sample is fetched.

The depth map D24X8 is sampled using PCF in the SM2.0 Graphics Tests only. DF24 is sampled using FETCH4 with bilinear filtering in the shaders. If the hardware supports D24X8 or DF24 and hardware accelerated PCF or FETCH4, a single bilinearly filtered sample is taken. The non-hardware shadow mapped rendering path uses four rotated point samples without bilinear filtering.

 

[Galduta]

Limited by CPU in 7800 GTX 256 MB SLI ? - 1, 2 and 4 game

Venice 3200 , 2000 Mhz




Venice 3200@2500 Mhz



and in 1900 xt / 7900 ?

 

[Kombatant]

Scott goes to the birds, or rather they go to him. There's a hint of a possible nV physics outlook as well.

Excellent piece

 

[KimB]

And coming to that there does appear to have been a number of changes with the outcome of the benchmark since then (i.e.: letting D3D requirements be the centrepoint of the requirements, not using optional features as requirements / highly decisive in the scoring, having single code paths and not multiple paths per vendor, using features that more than one vendor support, etc.). Many of these principles rooted 3DMark in some kind of solid basis with D3D being the arbiter, however now it seems a little confused as to what it is with all kinds of support for this, that and the other, with some seemingly arbitrary decisions. At the moment, with all the different options and requirement and stipulations I'm at a loss to understand what its actually telling anyone.

Allowing D3D requirements to be the centerpoint is every bit as arbitrary as pretty much any other decision Futuremark could make.

Consider, for example, that there are some features for which both hardware has similar, but slightly different support. PCF vs. Fetch4 is an obvious example. Sticking to only cross-vendor or D3D standard would be both arbitrary and silly in this instance, because both IHV's support very similar functionality. So it's good that this time around, Futuremark has chosen to support both.

Of course, it'll take a hell of a lot more convincing to get me to believe that 3DMark is anywhere close to as good a benchmark as real games are, but we'll have to see.