XLNX - Xilinx

[Nell-e]

I see that Amazon is using FPGA's for machine learning: https://aws.amazon.com/ec2/instance-types/f1/

 

I hear that FPGA's have advantages over GPU's for more complicated algorithms.  Anyone have an opinion on how big a market this could grow into?

[JRM]

FPGAs and CPLDs are more flexible and can be easily modified if needed without hardware modifications.  Unless I missed something, I don't think they will outperform an optimized GPU for a specific task.  FPGAs are great for prototyping and for applications where cutting edge speed isn't required.  They are still plenty capable for most applications and provide a cost advantage for smaller scales where it doesn't make sense for a one-off hardware design.

 

Not sure where the future is headed for these devices.  Maybe there is a use for these in Elon Musk's "machines building machines" dystopian future.

[Jurgis]

Yeah, I think FPGAs are niche in terms of AI/ML. Maybe less niche for robotics? - but I don't really know.

 

In terms of XLNX is also depends a lot of who drives the company and in which direction. NVDA was just a GPU company at one time... it might be fun to compare its trajectory to ATI/AMD. Was that predictable though? Not sure.

 

Disclaimer: I have not looked at XLNX for quite some time...