Why in hell does neither IHV offer a complete OpenGL version/extension and hardware feature matrix so that the thousands of developers around the planet save valuable time and nerves?
This is pissing me off!
Do you target cards older than GF8x00 and RHD2x00?
There’s http://www.realtech-vr.com/glview/index.html, contains a database.
Thanks, that’s at least something.
However this doesn’t look very reliable. Every IHV should have a feature matrix for their latest drivers for all their GPUs. The database is also missing hardware capabilites as far as I can see.
It is a matter of course to deliver comprehensive specification for every major kind of programmable hardware device. Why in hell is this not the case for GPUs?
Oh and yes, I unfortunately have to support GPUs that do not support DX10.
Every IHV should have a feature matrix for their latest drivers for all their GPUs.
Every IHV should also support OpenGL perfectly and in its entirety without errors. And every IHV’s drivers should be free from memory leaks and crashes.
A lot of things should be. But they aren’t. And odds are, will not be in the near future. OpenGL View is about as good as it gets currently.
At the page
http://developer.nvidia.com/object/nvidia_opengl_specs.html
click on “All NVIDIA OpenGL Extensions (.pdf)”, you will download
http://developer.download.nvidia.com/opengl/specs/nvOpenGLspecs.pdf
The nice thing is “Table of NVIDIA OpenGL Extension” from page 6 to 10. It gives you which extensions are supported per chipset, it even gives you if a given extension is supported in software fallback (sw) or hardware.
The sad thing is it from February 2008.
They also have a table for supported texture formats:
http://developer.nvidia.com/object/nv_ogl_texture_formats.html
but again the last update is from March 2009.
nVidia, can you please update your documents?
In addition, Apple provides this matrix:
http://developer.apple.com/graphicsimaging/opengl/capabilities/
A bug free and feature complete API support is a “little bit” more demanding than providing a comprehensive specification for products.
Right, I remember to have seen this document beforehand but couln’t find it anymore. But yeah, it’s totally outdated anyway as you have already said.
nVidia, can you please update your documents?
I think the demand should be that a detailed specifcation has to be provided for every GPU product of any GPU developer.
The amount of documentation and specification relative to the complexity of programmability is a total disgrace in comparison to other product types!
I can’t vouch for how reliable or up to date it is, but the site here: http://zp.amsnet.pl/cdragan/d3dcaps.html has more or less what you’re looking for.
The amount of documentation and specification relative to the complexity of programmability is a total disgrace in comparison to other product types!
That’s probably because most GPU users (the people who buy them. AKA: customers) don’t care. If it says, “Direct3D 10” on the box, that’s all they need to know.
I’m curious: what “other product types” are you referring to? I don’t know of any mice, scanners, or even video capture cards that have detailed API specifications.
It’s us, the developers who gives the end costumer a reason to buy their product. Now they even intend to activate a huge GPGPU market, which makes it even more important to document a comprehensive collection of specs.
I’m curious: what “other product types” are you referring to? I don’t know of any mice, scanners, or even video capture cards that have detailed API specifications.
CPUs or PLDs in general for example.
Be happy you’re not developing whatever software for mobile phones :P.
It’s us, the developers who gives the end costumer a reason to buy their product.
To be fair, it doesn’t matter much to us beyond “Supports DX10” or “OpenGL v3.3” or whatever. Both D3D and OpenGL are well-documented as to what they provide. Sure, any particular card might support more than the base specifications of these APIs. But if you’re in D3D land, you’ll never be able to use it, so it doesn’t matter. And even if you’re in OpenGL land, they have to publish a specification for you to be able to use it, and you’ll still need to query for the extension.
So you’ve pretty much got 95% of the relevant information just from the OpenGL version number.
Now they even intend to activate a huge GPGPU market, which makes it even more important to document a comprehensive collection of specs.
GPGPU is implemented via proxy, through a controlled API (DXCompute, OpenCL, CUDA, whatever) like OpenGL and D3D. So there is no need for “a comprehensive collection of specs” because you can’t get at any of the stuff you might need such a collection for anyway. You’ll only get at what the proxy (DXCompute, OpenCL, etc) allows you, and that is well-defined.