Horde3D

I've read that the team wants to add ms d3d9/10 support in future programs.By adding d3d support Horde can using windows d3d9 and windows vista d3d10 + supporting xbox consols.Currently ogl supports most of platforms such as winxp/vista linux based os and ... + PS3 gaming console.
By adding d3d support we can only add support for xbox consols.Why don't focussing on new Intels Larrabee GPUs.You know that Intel is going to support opensource projects.

I think you have heard about their new gpus, but I will write a little for n00bs.
Many years ago the military has bought the Intels Pentium MMX series for their military projects (radiation and ...).They have worked on those pentium mmx cpus and improved the cpus architecture.Yes they have made multicore (higher than currently dual core or quad core models) pentium mmx and vectorised them because of its small architecture and promised the intel to give the research results to them.Now intel has improved them and corrected the problems.The result is Larrabee GPGPUs.They are very similar to both cpus and gpus.The Larrabee couldn't compete the current cpus in single threaded programs but in multithreaded programs they can beat their fast core2duo and quadcore series (Because of small and 18~24 cores).Intel showed that Larrabee GPUs can beat well known ati and nvidea gpus in games.
Intel announsed that their gpus will support both ogl and d3d apis but they will release Larrabee api for native supporting their gpus.

I say isn't it better to team currently focus on ogl and improving horde engine, then going to adding support for larrabee api and then be the first next-generation engine that supports larrabee api

Excuse me for my poor english, if there is any problem or I'm wrong about something please post and edit them.

wikipedia
http://arstechnica.com

OpenGL is probably going to die in the near future, for obvious reasons(OGL3?), and Direct3D is probably the most popular/powerful API for games anyway. Which is sad because I've always liked OpenGL.



Well, not really.

Although Larrabee sounds pretty cool, I don't think anyone would have to develop for it specifically. It appears both current API's will run on it. I mean what are the chances that all this Larrabee business will just blow over? I remember hearing about the Cell and how it was going to replace all of the chips in our boxes by now, it hasn't happened yet.

Aside from the fact that the multi-billion dollar CAD/CAM industry (far larger than the games industry), most of Academia, Apple, and a majority of embedded devices use OpenGL exclusively... I don't see OpenGL dying anytime soon. Sure, we were all dissapointed at not getting a shiny new object model in GL 3.0, but it isn't any worse than GL 2, and it does add pretty much all the needed extensions to core.

Good point - despite all the hype, Larrabee is basically vapourware at this stage.

_________________
Tristam MacDonald - [swiftcoding]

Then by going to D3D will be hardware limitations problem solved (OGL2 GPUs) ? Can I use my old Radeon7000 and Geforce MX400 to load Horde3D or hardware limitations will continue and I must obtain a new graphic card that supports D3D9 at least?

Whats the features that OGL don't supports ? The only difference is geometry shaders but the problem has been solved by extensions provided to the OGL2 and OGL3 supports it at core.
I heard that OGL have ~20% performance drops in comparisons with D3D, is this true ?

I think that OGL is much faster. I have a Radeon7000 and Geforce MX400 and I have seen that the OGL gives better performance and fps than D3D (I have tested it on my both graphic cards + latest drivers).
You can test this by using Counter-Strike 1.6 (It supports both D3D & OGL) or any other games

I couldn't find any information on the size of the CAD industry, but I can't imagine it being too much larger than the interactive entertainment industry. Especially with the last few booms it's had. And just to clarify my comment, I was talking solely about OpenGL dying in high-performance real time rendering. The only reason I can see for them to continue using OpenGL is to maintain portability across multiple platforms. I can see the GPU manufacturers "feature freeze" the OpenGL drivers, making it basically exclusive for that market. CAD programs have no need for HDR, or parallax occlusion mapping.



Well, to be perfectly honest, I wasn't excited about the new object model. To me, the attraction to OpenGL is the same as in Horde3D, the procedural C interface. I just think that it works better.

1) LRB is a long way from hitting the street. 2010 best estimates, and software that will take full advantage will not be out until 2012-2013, there is no sdk available to the public and there probably will not be until late 2010. Let's keep things in perspective i.e. let's talk 2 years from now. Plus it will support OGL and D3D so Horde3D will run on LRB out of the box.

2) Minimum card specs is SM2.0, D3D9/OGL2x hardware support. So no.

3) Speed differences from both APIs are related to drivers, and the feature set used in the applications, there is no point in comparing performance. If you can't make your game/demo run at 30/60fps, you can't blame the API for it (since it is using the same piece of hardware). The API is just a tool, like an open wrench or a ring wrench they both can be used to turn/"screw" bults and nuts. To put it in a cliche "it's not the tool, it's how you use it that makes the difference".

Dude, it is a fact that OpenGL is much slower than Direct3D. To use your example, OpenGL would be like a regular old hammer, where d3d is the same hammer handled by the Hulk. You're right that it's not the actual API that is causing the speed difference, it's the drivers. But even so, OpenGL is slower across all of the GPU's/drivers which is what matters, otherwise there would be AAA games using OpenGL.

Dude AAA games that use OpenGL??? DOOM3, KOTOR, RAGE, etc...

Slower than what for what? If the game runs at 250fps vs 200fps you will not notice any diff, if the game runs at 30fps vs 60fps you will notice a difference... both APIs can be used to achieve that level of graphic efficiency (speed).

D3D is a more evolved, refined, smaller and simpler to use API that does not have 4 ways to do exactly the same thing. I love how MS makes clean breaks in functionality and moves much faster than the OpenGL ARB, giving developers access to features in a much more efficient and quicker way .

However, OpenGL is not dead and it's demise is greatly exaggerated! It is used by the vast majority of DCC makers: Avid, Autodesk, Adobe, Luxology, etc... it has a huge amount of code base and users. So even if you don't use OGL in your game engine, guess what!? the content for your game (3d assets and video composition) was probably created in a OGL powered pipeline

Real Time OGL holds the same position it always had. It is a great way to do some research work, it can be used to create multiplatform games and interactive applications, but in the MS/windows market OGL has always and will forever be in a minority position, XBox leverages the windows programming model and gives us DirectX as an SDK, Sony PS, and Nintendo don't use OGL because it is a HUGE messy library. Being huge and messy is actually good if you are doing research, but not in a production environment, that needs to be quick and efficient.

Anyway, I am all for supporting both backends, instead of arguing we should really be working on implementing a d3d9 backend in H3D. My d3d9 experience is very limited, and i never worked with d3d10 ( i don't have any hardware that runs it), i am very interested in studying the d3d9 backend code as a way to learn more about it, i hope that someone volunteers to do this work.

I can tell you right away that i will not be doing a D3D backend, but if one is introduced in a future version of H3D and provides benefits on the windows platform, i see no reason not to use it on that platform and fallback to OGL in other platforms.

OpenGL is said to have less overhead than D3D9, so draw calls are cheaper. D3D10 was designed to fix that shortcoming. Basically D3D and GL can do the same but there are a few things where D3D is superior. For example you can query hardware caps with D3D and have support for instancing (which is not that urgently required in GL because of the cheaper draw calls). Another important thing lacking in GL is binary shaders. That's a problem for huge AAA games which have tens or even hundreds of thousands of shader combinations that are usually stored in a precompiled shader cache. The GL community was hoping for that to be introduced in GL3, but you know the story...

So you're telling me if one was to write the simplest/most efficient scene in both API's, like drawing a simple sphere, OpenGL would be faster? If so, then that's awesome:)



Yeah I'm definitely not going to defend my comment. I truly forgot about Doom 3 (which is probable as I was thoroughly disappointed). However, I think most of those games support (default always seemed to be d3d on the ones I recognized) OpenGL for portability. And a lot of their sequels aren't on the list, which means they probably dropped OpenGL for DirectX. Most modern AAA games don't use OpenGL for previously mentioned reasons:)



That's all true, but it definitely seems like it's dying when it comes to commercial interactive entertainment.

No it's not, GL is fine.
It's true that GL3.0 should have just been called GL2.2, but the fact is that GL3 has created the process required to rejuvenate GL by stripping out all the old crap and allowing direct state access just like DX10 does.

The PS3 doesn't use OpenGL.
It has a low-level graphics API of it's own, plus it has a high-level API which is similar to OpenGL (but not many people use the high-level API on PS3 anyway, because the low-level one is much faster).
GL3 has specified a direct state access extension, which gives you DX10 style (i.e. fast) calls.

There's a GL extension for hardware instancing now too

Remember that Blizzard Entertainment is on the GL ARB now

Guys lets not turn it into a ogl/dx dispute like it happened with the topic in witch I asked about a dx renderer.

Imho there is no need to create a version of horde just for a particular version of GPU, even it being revolutionary and all.
If it had 60% of the market and this way it delivered more performance/support than horde's normal renderer, sure, i would agree.
But it isn't out yet, and it won't for some years, and who knows how much it will take for it to be a stable piece of technology
and widely adopted.

_________________
Paul

The high level API for the PS3 is in fact OpenGL ES (albeit with a number of platform-specific extensions). As you said, nobody uses it, as coding for the metal is faster on the PS3 hardware.
OpenGL draw calls are already as fast as D3D10 draw calls. The state model doesn't have much to do with this. DX10 does have a much better designed and better performing buffer/stream binding API though.

True, but from what I can see, instancing is usually still slightly slower than regular draw calls under GL (at least on recent cards) - instancing was created to allow D3D9 to perform as well as GL, after all. Hardware Caps queries are desperately needed in OpenGL though...

And don't forget ID Software - Carmack carries a fair amount of weight, and he hasn't abandoned OpenGL, he was even pretty upbeat about the future.

_________________
Tristam MacDonald - [swiftcoding]