Geforce 400/500 series video cards - OpenGL blocked/crippled !?

[A.Mitov]

I came across quite some articles (one of which on Wikipedia) and forum posts saying that there are some serious problems with OpenGL performance on the 400 and 500 generation Geforce cards. Is this true? As far as I know, quite a few apps have OpenGL viewports. Does that mean that if I use apps like Maya, Modo or Rhino 3D which are OpenGL based I would suffer bad viewport performance with a GeForce 400/500 series card?

 

Quite a bit people are complaining with their slow OpenGL performance on 400/500 series GeForce cards. I currently own GTX460 and tried it out on Maya, 1 million poly scene, no materials and textures. The fastest FPS I could get was 30 FPS. The same scene in 3ds Max (which is DirectX) gives me about 100-110 FPS. It looks like there really is an issue with 400/500 series GeForce cards in OpenGL apps. Some people report superior Maya viewport performance with 200 series Geforces over 400/500 series since they were the last to work OK in OpenGL.

 

Please take a look at these links to see what other people say too:

http://www.polycount.com/forum/showthread.php?t=92072

http://www.tomshardware.co.uk/forum...l-issues-series

 

Any thoughts? Can anybody confirm?

Would you recommend changing my GTX460 for a Radeon 6850? How are Radeons in Max and other DirectX apps? I hope they are not DirectX crippled .

[Dimitris Tolios]

You have to admit that 30fps is not that bad...actually it is good and I would call it a non-issue, especially when you think how cheap is a GTX 460 in comparison with Pro CG cards...

 

Yes, Radeons are usually better with OpenGL apps, and the really hi-end models scream through OpenCL - too bad they screw up the drivers and you cannot use it with VRay RT GPU (or it works with one version and breaks with the other...79xx, the best OpenCL raw performing card today and announced with up to 6GB of Ram does not work at all for example

- that could be a game changer for AMD which struggles to remain ontop of stuff, and actually is thought to be one of the top contributors to the OpenCL platform...WAKE UP!

 

Unless you use MARI or other CUDA only accelerated apps, Radeon was and is a good value (just like FireGL). Adobe is making the change to OpenCL instead of CUDA in their creative suite, taking advantage of discreet GPUs and APUs alike. Hopefully that will leave us with more options in the future, but, realistically before 2014 or so , nVidia will seem to enjoy the lion's share in all fields.

-----

I cannot "accuse" nVidia of "conspiring" in favor of their Quadro line etc...it might be the case, but I would bet that their architecture is far more game oriented than not. In most cases Quadro's and Tesla's architecture is also 99-100% similar to the GF, and the differences are mostly into BIOS and driver optimizations. Generation after generation, the GFLOPs produced by GPUs is raised by a considerable margin, but unfortunately just like with CPUs, the software development is lagging seriously into harnessing the available power...more than 7 years after dual cores became the norm, even entry level systems have 4-6 cores or more, yet very few applications are really multithreaded.

 

I don't see why we have to "accuse" just nVidia or AMD for lack of driver optimizations, any less than why have to accuse Autodesk or Chaosgroup for not looking into optimizing themselves their software, having nVidia / AMD modes etc. Their heart is set at rest, and let us troubleshoot their own shortcomings: the difference is that 3DS or Maya costs 7-10 times the best or second best gaming GPU. Paired with VRay we are talking adding 2-4 cards more...I would accuse nVidia those many times that products bearing the "professional" title are pwned by gaming cards 1/2 the cost or much less, but I cannot see why the 3DS/Maya developer has to serve the established hardware capabilities users have any less than the hardware developers have to serve the software companies. The fat and lazy guy in the story is almost always the code, not the silicon.