View Full Version : Futuremark 3DMark Next Q&A
Dave Baumann
28-Jul-2004, 08:52
http://www.beyond3d.com/siteimages/b3dsmall.gif (http://www.beyond3d.com/interviews/fm04/)Although 3DMark03 has been with us for some 18 months now, 3DMark updates have traditionally come inline with new versions of the DirectX API but DirextX9 is still set to last for some time to come. In light of this, here we quiz Futuremark on the directions and technologies they are looking at in for the latest version of their 3D benchmark.<blockquote>"The next 3DMark will use a completely new 3D engine, which dynamically builds HLSL shaders. The shaders are dynamically built and runtime compiled using the most optimal target for the hardware in use. We strongly believe that this is a structure future game-engines will be using because of its flexibility. It goes without saying that the compilations produce the exact same rendering, no matter which target. This is an absolute must for enabling objective performance comparisons."</blockquote>We also have an image captured from a work in progress test from the next 3DMark. Read the full interview here (http://www.beyond3d.com/interviews/fm04/).
Temporary Name
28-Jul-2004, 09:12
Doesn't sound like we should expect the '04 version any time soon, despite the apparent name.
It'll be 3DMark05.
bloodbob
28-Jul-2004, 09:14
I would like to know how they are going to automatically multipass shaders on PS2.0/VS2.0 hardware since they are being generatered on the fly. The teaser shot looks like far cry all over again :P.
Evildeus
28-Jul-2004, 09:16
Any expectetion on the launch date? This year i suppose?
Great interview anyway, some good information. No HDR is a pity :(
bloodbob
28-Jul-2004, 09:20
Any expectetion on the launch date? This year i suppose?
Great interview anyway, some good information. No HDR is a pity :(
Well I've yet to hear of a non NV chip that supports float blends I doubt the series 5 chips do ( I could be very wrong ). So its only a one chip feature so it sort of gets binned with the 3dc. ( although 3dc isn't actually a feature of DX where as float blends are an optional feature ).
Evildeus
28-Jul-2004, 09:31
That also means that no new Ati chip will support it in the near future :?
Hanners
28-Jul-2004, 09:41
That also means that no new Ati chip will support it in the near future :?
It depends what you call the 'near future' I guess. I believe ATI inadvertently confirmed that they will supporting floating point blending in their Shader Model 3.0 compliant hardware.
Rodéric
28-Jul-2004, 11:23
Since you use different colors for your questions and their answers, could it be possible to remove the italics ?
I find it annoying and useless....
Martin Eddy
28-Jul-2004, 12:20
Another good read Dave, Thanks.
That preview shot doesn't look very Next Generation to me. :?
Anonymous
28-Jul-2004, 12:28
i like the preview pic looks promising
I could share a small additional info about the shot.. All shadows in the scene are dynamic & rendered real-time, and all objects (even a straw off grass) cast one shadow for each lightsource (in the shot there are 3 light sources). Oh, and everything is of course self-shadowed too.
Anonymous
28-Jul-2004, 13:18
Nice interview, some boring replies, some interesting and some bullshit. as we know from the past from FutureMark. Especially I don't think that games will largely go away from CPU to GPU bound so soon. Simply by the fact as soon as you free some CPU timeslot it will get occupied by something new at once. Especially as games are looking more and more into complicated physics engine and pre-lighting computation like Spherical Harmonics and PRT. Also even with Geforce 6800 series its faster to have a lot of things pre-computed in textures and by doing the math directly in the shader itself.
Evildeus
28-Jul-2004, 13:48
That also means that no new Ati chip will support it in the near future :?
It depends what you call the 'near future' I guess. I believe ATI inadvertently confirmed that they will supporting floating point blending in their Shader Model 3.0 compliant hardware.In the next 6-9 months ;)
Nice interview.
What's with the cow question, though? I don't get it. :oops:
RingWraith
28-Jul-2004, 14:31
The teaser shot looks like far cry all over again :P.
Why, because they both have trees and grass? LOL :roll:
The lighting in that shot is above and beyond anything in games right now.
Hanners
28-Jul-2004, 14:56
In the next 6-9 months ;)
Nine months might just work. ;)
What's with the cow question, though? I don't get it.
Both 3DMark03 and 2001 have featured cows in some capacity. :P
digitalwanderer
28-Jul-2004, 15:14
9 months, hmmmm.... :|
What's with the cow question, though? I don't get it.
Both 3DMark03 and 2001 have featured cows in some capacity. :P
I can't seem to remember any cows in either benchmark..
Could you refresh my memory?
i like the preview pic looks promising
I could share a small additional info about the shot.. All shadows in the scene are dynamic & rendered real-time, and all objects (even a straw off grass) cast one shadow for each lightsource (in the shot there are 3 light sources). Oh, and everything is of course self-shadowed too.
:shock: :shock: :shock:
That can slow down things a bit...
:shock: :shock: :shock:
That can slow down things a bit...
I think that is the point - 3Dmark is suposed to run like crap on all graphics cards when it is released. Otherwise it will be CPU limited way to soon.
Nick[FM]
28-Jul-2004, 16:27
I can't seem to remember any cows in either benchmark..
Could you refresh my memory?
In 2001 you can enable an easter-egg where you can drive around and shoot down hoovering evil cows.. :wink: In 03 we had some cows in the ending of the first part of the demo.
Headstone
28-Jul-2004, 16:30
Good interview. Even answered a couple of my questions:D.
That preview scene looks as though it will either have fireflies or possibly wind to show off the 3 light sources and their new shadowing.
Ragemare
28-Jul-2004, 16:38
I like the sound of perspective shadow maps :D
:shock: :shock: :shock:
That can slow down things a bit...
I think that is the point - 3Dmark is suposed to run like crap on all graphics cards when it is released. Otherwise it will be CPU limited way to soon.
Yes it is but this is massive
:shock: :shock: :shock:
That can slow down things a bit...
I think that is the point - 3Dmark is suposed to run like crap on all graphics cards when it is released. Otherwise it will be CPU limited way to soon.
Yes it is but this is massive.
BTW this means that everything is per pixel, like next gen and UE3?
]I can't seem to remember any cows in either benchmark..
Could you refresh my memory?
In 2001 you can enable an easter-egg where you can drive around and shoot down hoovering evil cows.. :wink: In 03 we had some cows in the ending of the first part of the demo.
Cool. I will have to google around to see if I can find a way to find that easter egg. Unless you want to tell me about it, that is. :)
Edit: Found the easter egg. Gonna try it now. :)
GraphixViolence
28-Jul-2004, 22:01
So it looks like 3Dc is out because it's only supported by one IHV, but partial precision shaders are in. How many IHV's support that optimization again? It could be argued that each of these features can provide significant performance improvements for one vendor's hardware only, while having negligible impact on image quality.
If Futuremark is going to take a position on this issue, should'nt they at least try to be consistent? I personally agree with their policy of only using features supported by multiple vendors, but I don't see how _pp hints fall into that category.
If Futuremark is going to take a position on this issue, should'nt they at least try to be consistent? I personally agree with their policy of only using features supported by multiple vendors, but I don't see how _pp hints fall into that category.
As I posted on [H]
....Because _pp hints is in the DX9 and OpenGL spec.
3dc ain't. It's strictly a proprietary format of texture compression, nobody CAN use it but ATI (hell, I think they even have a patent on it).
Partial precision is in the spec, and anybody who wants to *can* use it if they want. 3dc isn't, and can't be.
3dc ain't. It's strictly a proprietary format of texture compression, nobody CAN use it but ATI (hell, I think they even have a patent on it).
Partial precision is in the spec, and anybody who wants to *can* use it if they want. 3dc isn't, and can't be.
People really need to read some of my posts more closely... :wink:
GraphixViolence
28-Jul-2004, 22:37
....Because _pp hints is in the DX9 and OpenGL spec.
All of the new SM3.0 capabilities are in the DX9 spec. FP blending is in the DX9 spec. Yet they are not in 3DMark04 because they are currently supported by only one vendor. I still don't see why _pp hints should be any different. OpenGL is irrelevant because this is a D3D benchmark.
3dc ain't. It's strictly a proprietary format of texture compression, nobody CAN use it but ATI (hell, I think they even have a patent on it).
3Dc is not in the DX9 spec, but it can be supported using DX9's custom texture format facility. Also, ATI has publicly stated that they are openly sharing information on 3Dc to whoever wants to use it.
Regardless, I'm not necessarily saying they should support 3Dc, I'm saying I think it would make more sense to do away with partial precision shaders.
termikk
29-Jul-2004, 00:55
not much impressed by the screenshot
if it's supposed to be realistic...
the ground textures and the shape of the rocks is cartoony
for something future looking better looks can be achieved
Good interview, Dave and FM.
I share Ingenu's annoyance with the italics, though. The meat of an interview is the answer, not the question, so making the answer less legible than the questions seems backwards to me. (I'm sure this has been requested before, too. I hope we're not waiting no a site redesign for greater legibility.) I think you should bold and italicize the Q, and leave the A plain. Keep the coloring, too. And adding a "B3D" to the front of each Q wouldn't hurt, either.
I'm with GfxViolence in disagreeing with FM's decision to exclude 3Dc. A "gamer's benchmark" would make use of whatever features a gamer's video card supports, no?
A related Q as to it not being in the D3D spec: were DXTC texture compression and partial precision added to the spec because of MS, or because of an IHV's pressure? Isn't ATi's hardware advantage with 3Dc similar to nV's advantage with doubled z-ops or partial precision performance? Heck, trylinear and brilinear seem fairly proprietary, but will FM protect against those?
Or is the fact that it's uniquely lossy a strike against it? Wouldn't this be countered by the fact that it offers better quality than DXTC, which FM seems to be amenable to using?
Evildeus
29-Jul-2004, 07:37
In the next 6-9 months ;)
Nine months might just work. ;)Really? I thought Ati decided to increase the time between each launch. So april 2005 like the R420 ;)
I'm with GfxViolence in disagreeing with FM's decision to exclude 3Dc. A "gamer's benchmark" would make use of whatever features a gamer's video card supports, no?
That would be a monumental task for FM to do though. Cause then they need to support every extra feature that the NV40 has also, FP blending, vertex texturing... Would make it rather difficult to get the exact same rendering on all cards then. And adding loads of feature tests doesn't seem to be a good idea imo.
All of the new SM3.0 capabilities are in the DX9 spec. FP blending is in the DX9 spec. Yet they are not in 3DMark04 because they are currently supported by only one vendor. I still don't see why _pp hints should be any different. OpenGL is irrelevant because this is a D3D benchmark.
PP hints are supported by both vendors. It's just that one vendor doesn't gain anything by using it. And the difference is also that you can use PP hints and get the same quality, as FM says, some materials will look the same with FP 16 vs FP 24/32.
Patric Ojala
29-Jul-2004, 12:34
]In 2001 you can enable an easter-egg where you can drive around and shoot down hoovering evil cows.. :wink:
LOL @ hoovering cows :lol:
- Nicke, it's 'hover' with one 'o'
Nick[FM]
29-Jul-2004, 12:52
]In 2001 you can enable an easter-egg where you can drive around and shoot down hoovering evil cows.. :wink:
LOL @ hoovering cows :lol:
- Nicke, it's 'hover' with one 'o'
Naah, nitpicker! :lol:
Anonymous
29-Jul-2004, 13:14
Actually I thought the idea of hoovering cows was far more amusing and merely one that hover!
Simon F
29-Jul-2004, 13:34
Actually I thought the idea of hoovering cows was far more amusing and merely one that hover!
Especially if you had a giant Dyson (http://www.bekkoame.ne.jp/~yukio-m/books/img/10/dyson.jpg) vacuum cleaner and could watch them spinning around in the cyclone :-)
Nick[FM]
29-Jul-2004, 13:40
Actually I thought the idea of hoovering cows was far more amusing and merely one that hover!
Especially if you had a giant Dyson (http://www.bekkoame.ne.jp/~yukio-m/books/img/10/dyson.jpg) vacuum cleaner and could watch them spinning around in the cyclone :-)
:lol: Not a bad idea.. :wink:
]In 2001 you can enable an easter-egg where you can drive around and shoot down hoovering evil cows.. :wink:
LOL @ hoovering cows :lol:
- Nicke, it's 'hover' with one 'o'
LOL! Isn't hoover another word for vacuum cleaner? Now vacuum cleaning cows, there's an easter egg! :lol:
GraphixViolence
29-Jul-2004, 15:15
PP hints are supported by both vendors. It's just that one vendor doesn't gain anything by using it. And the difference is also that you can use PP hints and get the same quality, as FM says, some materials will look the same with FP 16 vs FP 24/32.
I think you'd be hard pressed to argue that simply ignoring _pp hints comprises "support", but this is exactly how all non-Nvidia hardware will handle them. As for FP16 having the same image quality, that's no different from arguing that 3Dc compressed normal maps will often look indistinguishable from uncompressed normal maps.
Once again, I' m not arguing that they should support 3Dc, since if they did that they'd have to support every other single-vendor feature to be fair. I'm saying they should be consistent in their position about not using any, and that _pp hints should fall into this category.
Anonymous
29-Jul-2004, 20:38
Seems I'm a lit late to the party after all this discussion about it, but still I feel I need to thank Mr. Baumann for adding that q about the cows as I humbly requested in the other thread (although shortened nicely) .
Really made my day, totally unexpected in the end, I'm still smiling about it and it's been like 10 minutes since I read it :)
About the pic. Someone else already mentioned it so it can't be just me... but I'm still asking... Is it just me and the other guy or isn't it a bit cell-shadistic? :shock:
the post above this one
my post, forgot to log in :oops:
As for FP16 having the same image quality, that's no different from arguing that 3Dc compressed normal maps will often look indistinguishable from uncompressed normal maps.
The difference is that often is not always.
GraphixViolence
30-Jul-2004, 14:47
Sure, and FP16 doesn't always provide the same IQ as FP24. Someone has to go through all the shaders and decide on a case-by-case basis to be sure. You could do the same with 3Dc, i.e. only compress the textures that don't artifact. If you really wanted to ensure you had an optimal code path for each card, you'd have to do that to be fair.
Otherwise they should just produce neutral code that doesn't benefit one architecture over another. This may not be as representative of real games, but it's easier and fairer than implementing optimizations for one vendor but not for another.
vBulletin® v3.8.6, Copyright ©2000-2013, Jelsoft Enterprises Ltd.