GPU-Accelerated Air Traffic Management

[BD231]

http://developer.nvidia.com/content/cuda-spotlight-gpu-accelerated-air-traffic-management

Using CUDA and GPU computing we were able to achieve a 250X speedup over NASA’s baseline software called Future ATM Concepts Evaluation Tool (FACET). While FACET takes around 10 minutes to perform a typical 24-hour trajectory prediction for 35,000 aircraft in the NAS, the CARPAT prototype performs the same prediction in under 2.5 seconds

 

[T_Yamamoto]

 

[at80eighty]

nice

 

[-Slacker-]

2.5 seconds x 250 = 625 seconds, or it took air traffic personal over 10 minutes to do the same operations before. Wow is right

 

[Joseph F]

I can see the massively parallel architecture working well here.
I wish nVidia would license CUDA to AMD, but that'll never happen.

 

[Lotheron]

I can see the massively parallel architecture working well here.
I wish nVidia would license CUDA to AMD, but that'll never happen.

Intel is getting behind OpenCL and with their weight that might change.

It's going to be a very interesting battle between CUDA and OpenCL.

 

[blckgrffn]

Intel is getting behind OpenCL and with their weight that might change.

It's going to be a very interesting battle between CUDA and OpenCL.

I think this performance increase is awesome, first off.

Secondly, what does nvidia plan to do when every Intel and AMD CPU carries a OpenCL capable GPU? That has to be a logical part of the onboard GPU war between AMD and Intel. A year or two after that the install base for OpenCL will be staggering.

By keeping CUDA to themselves is nvidia making it the next Glide?

 

[MrK6]

This is pretty awesome. I have friends who are theoretical physics grads and they swear by CUDA to run their simulations. It's amazing what kind of horsepower GPU's have when running the correct software.

 

[BD231]

The mere fact that AMD/Intel chose to acknowledge GP-GPU computing points to an inevitable trend if you ask me, GPU's that perform tasks just as fast or many times FASTER than top of the line x86 cpu's? I doubt they're willing to believe they can ignore a threat that if proven versatile, which by all accounts is certainly shaping up to be, won't possibly pose some competition in unexpected places.

No clue if it'll take on big or not but I definitely don't see it fading away though as there are just to many different ways to repackage GP-GPU (h.264) and label it a new feature.