AMD unveils new flash-based memory to dramatically boost graphics performance

Advanced Micro Devices today unveiled its new Radeon Solid State Graphics Technology to help speed the performance of workstation graphics.

Sunnyvale, Calif.-based AMD said it is developing a Radeon Pro solution with a terabyte of memory to operate large dataset applications.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":2013012,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,games,","session":"D"}']

The company is offering a $10,000 developer kit for delivery in 2017. The Radeon Solid State Graphics developer kit will have a terabyte of memory, using flash-based non-volatile memory. The company made the announcement at the Siggraph graphics technology conference in Anaheim, Calif.

The new solution is ideal for the next wave of demanding graphics, including real-time post-production of 8K video, high-resolution rendering, VR content creation, oil and gas exploration, computational engineering, medical imaging, and life sciences.

Raja Koduri, senior vice president and chief architect of the Radeon Technologies Group at AMD, said in an interview with VentureBeat that for state-of-the art content creation, scientific and engineering visualization applications require the processing of big datasets, far larger than can be contained within the capacities of existing GPU memory. He said that the non-volatile memory and graphics combination could be “disruptive.”

“This is fast memory next to the GPU,” he said.

Current limitations require slices of data to be processed individually and later merged by software and often incur significant slowdowns for fetching additional data from system memory. These big data problems discourage developers in these domains from leveraging the advantages of the GPU. Radeon Pro SSG memory addresses the big data problem for GPUs, paving the way for improved performance and dramatically increased user productivity.

When data isn’t available in graphics memory, the computer has to fetch data through the slow process of contacting the central processing unit, which then retrieves the content from CPU memory, or from a hard drive. This slow process limits how quickly a GPU can perform. With the new technology, a terabyte of data in a new kind of memory system is available to support the GPU. So the GPU can fetch a lot more data without going to the CPU.

Applications for developer kits are now being accepted. The $10,000 kits will ship in 2017.