US residents who purchased an Nvidia GeForce GTX 970 graphics processing unit between Sept. 1, 2014, and Aug. 24, 2016, can now submit a claim for $30 cash for each qualifying device.
A class-action lawsuit, settled in late July, claimed Nvidia sold GPUs to consumers who expected 4GB of RAM but could access only 3.5GB.
As PCMag's sister site Geek.com explains, "With the GTX 970 and other Maxwell architecture cards, the company adopted a new way of disabling bad L2 cache blocks, and the division of memory is a logical extension of this. The total L2 cache was lower than originally reported by Nvidia, too (1.75MB vs. 2MB). There are eight memory controllers in the 970, but the GPU has access to just seven of them at a time. Nvidia's solution to this was to tell the GPU to only use 3.5GB of memory and reserve that last half a gigabyte."
Claims must be postmarked by Nov. 30 at the latest. You'll need to provide a verification code or proof of purchase. Verification codes should be listed on the notices of the settlement, if you received one. If not, you can submit a sales receipt, credit card receipt or statement, a shipping manifest, a purchase order, an email confirmation, or other similar documentation.
You can also submit a Product Identification Number found on the box, back, or bottom of each GPU or in the settings for the GPU via software.
On a more positive note for Nvidia, the company this week released a single-processor version of its Drive PX 2 computing platform for self-driving cars (above), which will first be deployed by China's Baidu for its "HD map-based self-driving solution for car manufacturers."
The Drive PX 2 we saw earlier this year combined two mobile processors and two discrete GPUs. The one announced today uses just 10 watts of power and is palm-sized.
"Bringing an AI computer to the car in a small, efficient form factor is the goal of many automakers," said Rob Csongor, vice president and general manager of Automotive at Nvidia. "Nvidia Drive PX 2 in the car solves this challenge for our OEM and tier 1 partners, and complements our data center solution for mapping and training."
Want to build your own autonomous vehicle? Maybe train your computer to automatically detect cancer? Sign up for courses with the Nvidia's newly announced Deep Learning Institute. Instructors include Nvidia's own deep learning experts, as well as pros from partners Coursera and Udacity; the company is also teaming up with Microsoft to teach deep learning for robotics.
Over the next four months, developers can take more than 80 workshops and hands-on labs around the world. Initial sessions, however, will be offered at the GPU Technology Conferences scheduled in eight cities, starting this week in China.