Openness for Academia

I am here to alert of the urgency for nVidia to drastically change direction regarding their licensing policy.

If the vast majority of nVidia GPU users are players using GeForce products, it is otherwise on other platforms, especially Linux and FreeBSD with which Quadro, Tesla and Titan are mainly used for computing, visualization, and science. In this second group, the vast majority are from the academic world (Berkeley, Harvard…), major research institutes (CalTech, MIT, NOAA…), space agencies (NASA, ESA…) and public services (town hall, schools etc.). These are the biggest players in the field of HPC, fudge of GPGPU solutions.

The nVidia’s Academic Research Support program, and the various tools (OpenACC, acceleration under MatLab or Mathematica softwares) that nVidia offers proves that this is a profitable market for the company, and that is a good thing. Nevertheless, the company’s current policy can no longer continue, and here is why.

A few years ago, nVidia offered open source “nv” drivers for Windows and Linux systems. Since nVidia has focused on offering only closed-source solutions. If this policy can be understood for gaming uses intended for the general public, it can not answer the stakes to which are confronted the actors, greedy of GPGPU cited previously (scholars, research institutes both private and public). This policy goes in the opposite direction by slowing down the development of powerful collaborative tools based on nVidia technologies. The situation has even worsened, since from now on, the recent GPUs (Maxwell and Pascal) require firmware, totally opaque and whose license of redistribution de facto limits, the use, in the academic world in particular, of nVidia technologies.

Also, as a bottle in the sea, I solemnly call on nVidia to review its policy concerning the licenses of drivers, firmware and GPGPU solution (i.e., CUDA, OpenACC and OpenCL support) in order to provide the academic world with the means to continue to use nVidia GPUs without infringing the necessary, collaborative and open advance of scientific research. There is an urgent need for nVidia to acknowledge the need to open these technologies at least to the academic world (QT licensing is something that comes to my mind as a good policy). Otherwise, I predict that many of its players will turn (and the movement is already on the march) to competing solutions that are open-source and based on industry standards (e.g., OpenCL 2.0).

This would be a tremendous showcase for the company and a considerable gain of confidence from a high value-added customer base.

NVidia, if you can hear me.