5 Things You Should Know About the New Maxwell GPU Architecture

Originally published at: https://developer.nvidia.com/blog/5-things-you-should-know-about-new-maxwell-gpu-architecture/

[Be sure to check out Maxwell: The Most Advanced CUDA GPU Ever Made, a newer post about the second-generation Maxwell GPU architecture.] The introduction this week of NVIDIA’s first-generation “Maxwell” GPUs is a very exciting moment for GPU computing. These first Maxwell products, such as the GeForce GTX 750 Ti, are based on the GM107 GPU…

Thanks for the details! The improved shared memory atomics are a welcome surprise.

The links to the Maxwell tuning and compatibility guides do not work.

Sorry about that, Steve. The web cache for our registered developer site seems a little slow to sync up today. The files should show up any time now; please check again in a bit.

Cliff Woolley

FYI, the links are still not working.

Apologies again for the delay. The download links have now been fixed.

Excuse me for the question here but Nvidia in the future gpus will be use Hevc hardware acceleration? Thanks.

Hevc h.265 ,, nvidia quadro k600 have h.264 encoder there must be a
protoyp h.265 going on...

Hi !
Now its time for Nvidia to explain which of the Gpus will carry hevc/H.265 encoder.
All quadro got the h.265 encoder but but that i s yesterdeay.
You cannot put all your time and Money on the android platforms... and let us
So may I ask for a roadmap including h.265 encoder gpu in desktop platforms

...correction nvidia quadro i ment "all quadro got the h.264 encoder....

I am currently writing a textbook on Parallel Computing to be published soon. Both CPU and GPGPU parallelization are covered. Maxwell the latest GPU architecture will be included.

Though I am happy that size of the shared memory is back to how it was originally for easy one-to-one correspondence with threadblocks, I am worried about the L1 cache. Has access become slower like the Tex cache or has the Tex cache become faster like the shared memory? I have searched everywhere to determine the new latency of L1 cache in Maxwell archi, but in vain. This surely will have an impact on current codes relying on L1 cache. Hope this will have answers, thanks.

Will it always support PCI 2.0