In the past I mentioned that Intel, Nvidia and AMD/ATI are trying to expand to each others application domains (CPU und GPU fusion, a demo with OGRE3D and Nvidia's expansions). One may think that it's about market share but if you consider that parallelism is the key for future performance increases, then it looks more like a natural evolution.

GPUs no longer separate Pixel- and Vertex-Shader processing units on hardware. With DX10 it got unified - pixel and vertex (+geometry/tesslation) shaders are executed by the same hardware units. Therefore if a scene uses few vertices but a lot of pixels the hardware is less idle as it adapts accordingly - which is equal to higher performance. Therefore GPUs become more programmable while CPUs become more parallel. 

At Siggraph 2008 there have been a couple of talks about CPU/GPU architecture. Their course notes are now online: http://s08.idav.ucdavis.edu/  

Interestingly Kayvon Fatahalian thinks that Intel's chips needs ~ 32 Larabee units @ 1GHZ to reach the processing power of a current GPU which he says is about 1 TFLOPS. But today a game needs GPU AND CPU processing power. 

No matter how fast the CPU gets, CPU+GPU will always be even faster! Of course money matters and if the ratio moves heavily towards one favor, it might create a change. And that's maybe the point. INTEL's Larabee introduces a much higher flexibility, i.e. easy access to render-targets, so this could reduce costs. For this aspect also checkout the notes from the ID software guy where he mentions Voxels engines and other examples. On the other hand it brings even more choices and that doesn't make life always easier!

Until a shift happens (and if ever) we will probably see Larabee as add-on chip. Here's an interesting Intel statement from one of the sources of my earlier blog articles i mentioned at the beginning:

"Intel is not predicting the end of the discrete graphics business. Moore's Law has allowed Intel to innovate and integrate. As a result, we expect that we and others will integrate graphics and visual computing capabilities directly into our CPUs in the future much like floating point coprocessors and other multimedia functions have in the past. However, we don't expect that this integration will eliminate the market for higher-end discrete graphics cards and the value they provide."  

Like this it sounds a bit like Intel's Next-Gen GMA chip. That's probably the short-term perspective. It is said that Larabee is available in 2009 with 24 to 32 cores. But if they release those at 2 GHZ per core it sounds more like a GPU replacement unless you consider the too high power consumption (sources: german , english).

If you want to read a bit more about NVIDIA's new GTX 200 architecture and it's performance, Tom's hardware has a detailed review.
And here a very detailed article about Larabee from early august.

Happy reading ;-) 

It's already old news but I just came across it. Boy, there are many acquisitions lately!!

Some RealViz products will be discontinued. One of them is Matchmover of which we have a license…
I guess if it gets integrated into 3ds max, we won't mind at the end.

(…) 

Autodesk intends to develop and sell REALVIZ’s Stitcher Unlimited, Stitcher Express, ImageModeler and Movimento software as standalone products. Matchmover, Retimer and VTour will no longer be available as standalone products; core technology from these REALVIZ products will be integrated into future versions of Autodesk’s existing products, enabling customers to bring the real world into design environments.

The following REALVIZ offerings have been discontinued: Stitcher Pro, Stitcher Unlimited DS, StoryViz, and hardware and software product bundles. Education versions of ImageModeler and Stitcher continue to be available.  Student versions of ImageModeler and Stitcher are no longer available.

(…) 

Source: Autodesk Press Release  

Also take note of the part I highlighted. This probably means image-based modeling/lightning and pathfinding+traffic modules for Max, Maya etc. 

(…)

where it's easy to build 3D environments or virtual sets by stitching pictures together or derive 3D models very quickly from set photos using image-based modeling. Games now require a lot of facial animation. Animators can rough out facial animations using a webcam and optical motion capture.

(…)

We have plans to use these technologies across a wide spectrum of industries… Every country in the world is starting to model their cities in 3D, and the Kynapse technology offers some very relevant potential for these applications. The path-finding and AI technology can be used to do simulations of crowds inside of buildings or stadiums or to model and simulate car traffic in digital models of cities.

Source:  VFX World Interview

As previously mentioned, INTEL is trying to bring GFX calculations back to the CPU system. Their argumentation for this is

(…) that Intel processors offered “more bang for the buck” and that it was more economical to go from single to multiple core processors versus popping multiple graphics cards into a machine.  “The fact of the matter is that you’re going to have one graphics card, you may have a dual graphics card, but you’re not going to have a four graphics card or eight graphics card system,” said Fosner.

(…) 

This quote comes from this tgdaily article about an interview they had with INTEL on the Intel Developer Forum in Shanghai. The real-time demo shows 8 hardware thread on 4 cores in action. It's based on a game-engine framework called "Smoke Framework" and currently it's using OGRE3D for rendering.

INTEL is considering to release the sourcecode to the developer community as stated on the OGRE3D forum. As the OGRE3D community is currently working on DX10 support, the Smoke Framework probably will continue using OGRE.

There are two videos available about the demo and the interview - one is the article mentioned above but I'll list them both here again:

• People “probably won’t” need discrete graphics cards anymore
• Intel shows off Nehalem’s power with fire and smoke

Somehow related to the topic:

• SwiftShader - a software rasterizer fpr Dx9 and Shader 2.0

Maybe you already saw the plenty of ads everywhere about GarageGame's new gaming portal called InstantAction. I think it's very well done. Most of the site is HTML - only some areas do require flash and of course the games are using an extra plugin. Installation of that plugin in firefox was very smooth - very nice with pictures for each step. The website design is trendy and makes it already stand apart, but also the fact that you can quickly play multiplayer 3d games in your browser - after registration and installation of the plug-in.

On the negative side I found, that after downloading a new game, the process does often hang. Leaving and rejoining the game fixes it. I also had to give internet-connection rights to each game separately. In addition to that, in the lobby you don't see ping times thus lag or disconnection might occur often in some matches. If you want to try it yourself, I'll suggest "Marble Blast" or "Think Thanks". The others are less interesting. A Tribes-Like title is announced called: Fallen Empire Legions.

As we are talking of GarageGames, there is maybe something else interesting especially in regards to the topic of one of my recent postings. They have a 2D-Game Authoring environment called TorqueX Builder. It's layered upon their for their new XNA based engine called TorqueX. They now announced that they added 3D support on the engine level - the editor will be 3D enabled in the future. TorqueX allows deployment on Windows and Xbox360.

As I mentioned the potential merge of GPU and CPU previously, here's a recently published interview with Jon Carmack who gives his thoughts on the whole subject:

John Carmack on id Tech 6, Ray Tracing, Consoles, Physics and more

Well Nvidia recently bought MentalImages and now Ageia. Some people already think about if it would make sense for NVIDIA to buy AMD/ATI, too.

For many years now I always favored NVIDIA Gfx cards. ATI was said to have good hardware too, but for many years their drivers simply didn't work well enough for me. Acceleration for 3ds max wasn't working well and dual-screen support for accelerated 3D fullscreen content wasn't available for some years neither.

I was a bit disappointed about the usability when Nvidia switched their control panel GUI/System. It improved though over time, but recently I was surprised how bad dual screen management has become. I had a hard time to make the driver do what I wanted and sometimes it totally failed until I reinstalled the driver. Somehow the second output popped back to HD TV and wrong resolution very frequently. Then it somehow worked for a while and sometimes it stopped working again. Not very transparent…

In addition to that we sometimes hit randomly blue screens in a over 50 nodes cluster. I don't know if it's a Virtools issue, because I know that sometimes in the past years Virtools was able to create blue screens. Recently Virtools told me, that they have more people reporting them about unstable Nvidia hard-/software. I also found quite a large number of hits in google with the bluescreen keyword that is referring to the Nvidia driver. Many of them from 2007. This makes me think the following:

Vista and DX10 was important for Microsoft and Nvidia. It was two big and new things: DX10 driver/hardware plus new OS system with significant changes. This combo probably consumed a lot of manpower. I wonder if all this may had an effect on quality/stability? If so, it may mean that if Nvidia goes more and more into other markets (partly forced by INTELs efforts) and grows and grows … quality may suffer (even more)?

I hope not! We will see, it's certainly interesting times ahead …

If you are using Nvidia cards, do also feel that they become less solid/stable - or is still everything for you like always?

If you read Beyond3D or Tomf's blog then you probably already heard of INTEL's project 'larrabee'. If not: the basic idea is to mix CPU and GPU functionality into one 'multi mini-core' system. One could call that "CGPU". The target for 2009 is up to 16 cores (maybe less) with each capable of 4 hardware threads.

The source article ends with: 

Now do you understand why AMD had to buy ATI to survive? 

In the context of the mixed architecture that might be more flexible to to program, raytracing (as alternatives to rasterasation) for real-time rendering is discussed. Another citation from the above article is

Now do you see why Nvidia is dead?

Maybe you know about Mental Images, the creator of the famous MentalRay offline renderer that ships with 3dsMax, Maya and XSI … … they have been bought by NVIDIA! Nvidia does provide with Gelato another rendering product for the DCC market but maybe their acquisition was also interesting for them in regards to the Intel's Larrabee project where raytracing for real-time (RTRT) is researched.

So what will the future bring us? Maybe instead of 2 pairs of separate competitors 'AMD<->INTEL and NVIDIA<->ATI' we will have a triangle 'INTEL<->NVIDIA<->AMD/ATI' ?

( P.s. Actually, as I am right now discovering, Larrabeee details are already public since April 2006 )

Last week I saw a presentation from ATI about their new FireGL cards. The presenter said, that the new cards will come with new drivers that have been completely rewritten from scratch. We all know (and they know it too) that their drivers have been a weak point. That's also why I always did vote for Nvidia cards. My experience with ATI cards under 3ds max (i.e. in 2004) have been very bad. Also when I needed 3d acceleration on multi-monitor setups, ATI didn't offer that feature at all. Last time I checked (i think dec 2005/ jan. 2007) they had just release a very new driver having this feature and it didn't work very well for me, which only confirmed my choice of preferring NVidia cards.

An interesting aspect was, that he mentioned a collaboration with Dassault Systemes: as they provide one of the most widely used CAD application in the industry (CATIA) they worked with them to improve performance. He said, that once DS reported of being able to display and handle 90 million faces (? or was it vertices - sorry I don't remember anymore exactly) inside CATIA. This is done via Vertex Buffer Objects.

The cards will have up to 320 unified shader pipeline and up to 2 GB of RAM - these numbers are of course from the high-end version "V8650".

The short answer is: no!

The long answer is:  3ds max hides complexity. Before per-pixel handling via programmable pixel-shaders became available on consumer gfx cards, shading components across the triangle were mainly calculated by interpolating the value between the three vertices. A triangle is flat and thus you can interpolate in a linear/straight way. For example the vertex colors and normals.

Now, if you have a mesh inside 3ds max - let's say a cube in it's default state. It has 6 smoothing groups, because there are hard edges between each face. How many normals do you need for each corner vertex? Each corner has 3 faces - each of them pointing to another direction and having "hard" seams at it's borders.

We need 3 normals for all 8 corners … that's 24 vertices. Ok, now convert it to an 'editable poly', go into the face sub-level and hit CTRL+a to select all faces. Hit "clear all" and then smoothing group "1" in order to assign a single smoothing group to all faces. Export this one and load it into Virtools. And?

Oh, it ain't 24 anymore but not 8 neither? Yes! That's because there are not only normals stored in the vertices but also UV coordinates and color/material information. In 3ds max do then the following:

• add the "UV Unwrap" modifier to the cube
• go into face sub-level mode
• select all
• open the "Edit…" dialog
• select Mapping -> Unfold Mapping ('walk to closest face') -> Ok
• You now have a cross-like UV layout
• go into vertex sub-level mode
• select through each vertex and notice that sometimes another one is highlighted in blue
• right click and select "target weld" form the context menu
• now weld each vertex to it's "blue partner" in case it has one

Import the smoothed cube that has no UV seams into Virtools … … … what does the mesh setup say? 8 vertices, finally !!!

I hope this little tutorial has shed some light on the topic therefore you should now be able to understand better why there might be differences in vertex count between Virtools and 3ds max.

3ds max 2008 is announced. Something interesting is the GPU shadowing system for the viewport. Checkout the video about the "review rendering" feature. It does support up to 64 lights.

I also like the new working-pivot and the new LOD options. The ability to UV unwrap multiple objects manually at the same time also sounds very useful.

And … finally … a better maxscript editor!! Yay :)