GPU Based Processing

[roskirko]

GPU based processing is far faster than CPU based processing. 

Infraworks must use CPU based processing currently (my CPU readout is maxed to 4500 MHz,has temperatures reaching 212° F and shows nearly 100% CPU use when processing Infraworks models).

The code should be modified to optimize GPU based processing.

An article can be read here showing details: www.nvidia.ca/object/what-is-gpu-computing.html

There are 14 Autodesk products using this technology and Infraworks is not currently.

https://www.nvidia.com/content/gpu-applications/PDF/gpu-applications-catalog.pdf

[dgorsman]

There are limits with GPU processing.  While it is subjectively "faster" that's only on certain specific tasks, as they run on a far more limited instruction set than the CPU.  Anything other than those tasks simply cannot be done.  Those instruction sets are also have vendor limitations e.g. there are some functions only available on NVidia GPUs.

They are also limited to the onboard VRAM, which makes it nearly impossible for users with average video cards (~ 4 - 6 GB) to work with large projects.

The gaming-line video cards work in single-precision math, which isn't necessarily a good thing when dealing with technical designs and large coordinate values.  The workstation Quadro/Tesla and AMD-equivalent lines are capable of doing double-precision math however they are more expensive for equivalent VRAM and other capabilities.

And you really should get your system checked out - assuming that temperature is accurate, that is by *far* way to hot for running, even when maxed out on all cores.  If you're overclocking stock cooling just isn't going to cut it.