It will debut this quarter with the ability to offload PhysX and other tasks from the GPU

Jan 13, 2014 08:20 GMT  ·  By

The Maxwell has been the subject of rumors for months, but as the launch date nears, the speculations grow, and information leaks allow us to get a rough picture of what we can expect. Apparently, an ARM CPU will be involved.

For years, the tune that NVIDIA has been playing has said that its GPUs are good for parallel computing in addition to graphics processing because of the CUDA cores and technology.

Now, the company's architecture is making a vertical loop, doing basically the opposite of what AMD has done with its APUs.

Instead of integrating GPU cores in CPUs, NVIDIA is including a CPU into its GPU.

More specifically, the Maxwell GM117 graphics processing unit will have a Denver ARM central processing unit built into it.

A 64-bit capable, dual-core chip, Denver will offload some of the tasks from the GPU, probably PhysX calculations.

Already some Maxwell GeForce chips have been spotted, but not the GM117. Only the GM108 and GM107, which only go in notebooks.

The GM117 will be used for both mobile and desktop graphics. Still, that only the GM100 series appeared could suggest that the mobile lineup is coming out first, and that the desktop range will make us wait for a while.

Not too shocking, knowing that the GeForce GTX 700 series is fairly young still, and that a 6 GB model of the GeForce GTX 780 Ti is in the works.

We can only hope that the NVIDIA GPU Technology Conference, set to be held on March 24, 2014 (GTG 2014) will bring with it some clarifications.

Then again, even if NVIDIA proves tight-lipped, we can still count on the occasional information leak to continue to help form an idea of what is in store for us. Who knows, maybe AMD will end up doing something similar for its GPU.

Now we can only wonder how long it will be before there is no difference left between GPUs and CPUs/APUs. If nothing else, everything seems to be leading to an age of single-chip computers.