Radeon HD4800 Series Supports a 100% Ray-Traced Pipeline

[btarunr]

During AMD's recent Cinema 2.0 event, TG Daily caught up with representatives of JulesWorld. They are associated with the video production process of an unnamed movie studio and adopted components that are standard in the FX industry (AMD Opteron + Nvidia Quadro). Ray-tracing was an early focus of the company and they were told that JulesWorld started experimenting with a 100% ray-traced pipeline on a GPU with the arrival of ATI's R600 (2900XT) chip. And the results were impressive.

JulesWorld will be releasing OTOY and LightScape, two distinctive technologies that could shape the future of movie and games production. The company developed a ray-tracer that uses elements of the DirectX 9 API as well as its own high-level code that uses Tessellation and anti-aliasing algorithms.

Urbach further added that ray-tracing in real time became a reality with the Radeon 2900XT - which was used for a series of trailers for last summer's hit-move Transformers. All those Transformers teaser trailers were rendered on a GPU and - more importantly - directed in real-time. The producer of these trailers had complete freedom to play around with a "virtual lens" and direct the trailer in his own way.

In terms of performance, the Radeon 2900XT 1GB rendered Transformers scenes in 20-30 frames per second, in 720p resolution and no Anti-Aliasing. With the Radeon 3870, the test scene jumped to 60 fps, with a drop to 20 fps when the proprietary Anti-Aliasing algorithm was applied. Urbach mentioned that the Radeon 4870 hits the same 60 fps - and stays at that level with Anti-Aliasing (a ray-tracer is not expecting more than 60 fps.) JulesWorld's technology also works on Nvidia GeForce 8800 cards and above, but the lack of a tessellation unit causes a bit more work on the ray-tracer side.

A ray-tracer is limited by the amount of local memory on your video card, so if you are able to get a Radeon 4850 or 4870 with 1 GB or even more of on-board memory, you have very capable hardware. JulesWorld's LightStage technology can take wireframes consisting of an insane 32 million vectors to enable real world characters and their expressions. That means: Extra memory on your graphics card doesn't hurt.



View at TechPowerUp Main Site

 

[DaJMasta]

Let's get some 2GB HD4970s and some extreme-resolution detail packs for games.



Let the era of the playable movie begin!

 

[tkpenalty]

Holy crap. Ray tracing!?!?!?!? lol looks like intel is screwed when they release the larabee That was one of the touted features of it, now the HD4850 already has it!

 

[InnocentCriminal]

Wonder if they'd used a 4870X2. I'm wondering a lot this morning and another thing I'm wondering is how well this would all scale in Crossfire(X).

 

[tkpenalty]

lol, ray tracing is usually done with a CPU, remember how a CPU doesnt have anything near the same computational power that a GPU has. Consider how ray tracing with CPUs is slow... now speed that up 100 times or so . Thats the 4850 .

 

[DarkMatter]

Cool. When are we going to see this on our games?

lol, ray tracing is usually done with a CPU, remember how a CPU doesnt have anything near the same computational power that a GPU has. Consider how ray tracing with CPUs is slow... now speed that up 100 times or so . Thats the 4850 .

Yeah that's true. Intel tried it hard saying ray-tracing was only possible on CPUs. Now that the much more powerful graphics cards can do it, it's a more viable solution performance wise. They have a long way to go though, as a demo at 720p is still far from any resolution we use for gaming nowadays, and one of the cons of raytracing is that, unlike rasterization, it's performance decreases linearly as the resolution increases. But it's an impressive start nonetheless.

 

[InnocentCriminal]

Not for a while I wouldn't have thought, it'll take dev's a lil while to get their head around it probably. Saying that, I wouldn't be surprised if iD bring something out before anyone else.

 

[btarunr]

And now, for the million dollar question, folks: What would have caused the HD3870 to churn out 2x the frame-rates over a HD2900XT...and that too a 1024 MB model, considering memory size plays a key role in ray-tracing?

 

[tkpenalty]

would be nice if they used ray tracing instead, honestly, shaders FTL. Shaders dont bring a perfect image, unlike ray tracing. Ray tracing is just based off algorithms...

 

[DarkMatter]

And now, for the million dollar question, folks: What would have caused the HD3870 to churn out 2x the frame-rates over a HD2900XT...and that too a 1024 MB model, considering memory size plays a key role in ray-tracing?

Since based on what we know R600 had pretty much all the features that RV670 had, despite one being DX10.1 aproved. I mean the features are supported, but since not ALL of them are supported R600 can't wear the DX10.1 sticker. Same with Nvidia cards. It must be IMO one of these:

- Depth buffer readback, even though this feature seems to be less precise than the common Z depth tests. You want ray-tracers as precise as you can but...

- PCIe 2.0: A ray-tracer will surely benefit from the increased bandwidth, just as it benefits from the increased frame buffer, but as to make such a difference?

Those two are the only ones that I could think of now. I just woke up so if I think of another one I will post it later.

 

[JRMBelgium]

Does anyone know how they did the CGI in Juressic Park back in 1992-1993. Because I still don't understand how they managed to create such huge and realistic animals with the computers from that time.

 

[InnocentCriminal]

Juressic Park the porn version?

Jurassic Park was a mixture a both Animatronics and CGI...

 

[AsRock]

Does anyone know how they did the CGI in Juressic Park back in 1992-1993. Because I still don't understand how they managed to create such huge and realistic animals with the computers from that time.

As i know of they did not, they used models like with the T Rex had 12 people operating it. The T Rex was made out of fiber glass and 1000's of lb's of clay.

EDIT: http://www.lost-world.com/Lost_World02/Jurassic_Park.Site/Jurassic_Park.html

Apparently 3000 lb's of clay lol

 

[DarkMatter]

Does anyone know how they did the CGI in Juressic Park back in 1992-1993. Because I still don't understand how they managed to create such huge and realistic animals with the computers from that time.

As they already have said they used primarily animatronics. They even used static figures in some of the close up shots. And they did use some CGI. CG was mostly used to give animatronics a more life-like appearance though.

Anyway there were some powerful computers back then. They didn't exactly make films on PCs you know , they don't make them on PCs today neither. Maybe they start doing it in the future with 4 x HD4870 or 3x GTX280 though...

 

[btarunr]

Since based on what we know R600 had pretty much all the features that RV670 had, despite one being DX10.1 aproved. I mean the features are supported, but since not ALL of them are supported R600 can't wear the DX10.1 sticker. Same with Nvidia cards. It must be IMO one of these:

- Depth buffer readback, even though this feature seems to be less precise than the common Z depth tests. You want ray-tracers as precise as you can but...

- PCIe 2.0: A ray-tracer will surely benefit from the increased bandwidth, just as it benefits from the increased frame buffer, but as to make such a difference?

Those two are the only ones that I could think of now. I just woke up so if I think of another one I will post it later.

These renders are on a DX 9.0c level

 

[DarkMatter]

These renders are on a DX 9.0c level

Depth buffer readback is not a feature of DX10.1. It's a feature required for DX10.1 compliance which is not the same.

Graphics cards can be used totally outside DirectX you know, they can be programed to do whatever you want if you know how to do it. You don't need CUDA or CTM either.

"The company developed a ray-tracer that uses elements of the DirectX 9 API as well as its own high-level code that uses Tessellation and anti-aliasing algorithms."

 

[btarunr]

No wonder ATI kept the tessellation unit in the RV770 intact, what it inherited from its uncle, the ATI Xenos.

 

[DarkMatter]

No wonder ATI kept the tessellation unit in the RV770 intact, what it inherited from its uncle, the ATI Xenos.

Anyway I think that at this point we should talk more about AMD than Ati. Sure they retain some independence, but not all. Nevermind, I say this not because of the (in)dependency but because of the new contacts that now Ati has thanks to AMD. AMD has always had a very good relationship with FX companies, so probably they knew this was coming.

Anyway the tesselator is a very valuable feature, not only for ray-tracing, so taking out of RV770 wuldn't be a good choice.

On a side note, after a second thought, I don't know why they are using DX9 and not DX10. In reality none of the two are really usable for ray-tracing, as they are focused on rasterization, so maybe the part they are using is almost the same in DX9 and 10 and in 9 they have less overhead or something.

But what really interests me in all this, is that it could give us a hint about why Nvidia didn't bothered to support DX10.1, even though they pretty much support all the features except a few ones. If they knew that ray-tracing on GPUs is possible and that it was not too far away from being feasible to implement it on games, don't you think it makes sense to forget about spending the money on DirectX and use the money in a "better" way like say... buying a ray-tracing company like RayScale? Just a thought.

 

[lemonadesoda]

The company developed a ray-tracer that uses elements of the DirectX 9 API as well as its own high-level code that uses Tessellation and anti-aliasing algorithms

I've no clue as to why the raytracing performance doubled from 2900XT to 3870. All specs and synthestic benchmarks give no significant advantage to RV670 http://www.digit-life.com/articles3/video/rv670-part2-page1.html except ONE: exposure of tessellator under DX10.1

So my guess is one or more of the following:

1./ Proprietary tessellation routines couldnt access the R600 tessellator for acceleration, due to hardware tessellator not being exposed under DX9 API/compiler
2./ Under DX10.1 the hardware tessellator is exposured, resulting in double performance
3./ There may have been other changes to the proprietary algorithms they coded for AA and didnt notice they were testing different code on the different hardware.

Dark... could you give us a 3 sentence summary of how the tesselator works.

 

[tkpenalty]

I've no clue as to why the raytracing performance doubled from 2900XT to 3870. All specs and synthestic benchmarks give no significant advantage to RV670 http://www.digit-life.com/articles3/video/rv670-part2-page1.html except ONE: exposure of tessellator under DX10.1

So my guess is one or more of the following:

1./ Proprietary tessellation routines couldnt access the R600 tessellator for acceleration, due to hardware tessellator not being exposed under DX9 API/compiler
2./ Under DX10.1 the hardware tessellator is exposured, resulting in double performance
3./ There may have been other changes to the proprietary algorithms they coded for AA and didnt notice they were testing different code on the different hardware.

Dark... could you give us a 3 sentence summary of how the tesselator works.

No need for him to explain. Just think, Tesselators basically are like AA on edges of objects, except on a 3D perspective; adding more triangles and smoothing out the shape of an object. This simplifies coding for models, and moreover allows very realistic models.

 

[Weer]

Gosh, everyone keeps talking about the HD4000 series cards and running benchmarks on them and dressing them up in finely knit skirts to show off the fact that AMD stuck 'em with a crapload of useless superscaler ALU's that are worthless in games but somehow breath magic fireballs out of their arses when no one is looking.

But what about the GTX 280? Surely it can do this Ray-tracing bullshit better. It's more than twice as powerful.

 

[InnocentCriminal]

You've got a point, but the company in this thread were probably happy with ATi and continued with ATi. Maybe the tessellation units in the 280 aren't as efficient as ATi's.

 

[lemonadesoda]

So R6xx and R7xx "tessellators" are the same as ATi "truform"? (old terminology). But how would that help raytracing? That's just geometry instancing/scaling. Cant see how that would boost raytrace performance. And thats the question.

 

[Rurouni Strife]

would be nice if they used ray tracing instead, honestly, shaders FTL. Shaders dont bring a perfect image, unlike ray tracing. Ray tracing is just based off algorithms...

Shaders are independent of ray tracing. Shaders are surface materials. They are what give a wet rock it's look. It also tells how the material interacts with light. Raytracing gives light a nearly perfect interaction with the material. What ray tracing would replace is the current imperfect light setups, allow perfect reflections and refractions, and better shadows. Current shadows are bit-map shadows, and the softer they are just eats more memory. Ray traced shadows apperear more realistic and use a little less for roughly the same IQ.

As for this news, I'm impressed. Even if this isnt real time, 1 GPU doing a raytraced scene as complex as their trailers is definately going to be faster than a dual core or quad core CPU of today.

 

[Rurouni Strife]

280's dont have tesselators.