AMD Announces Radeon Vega Frontier Edition; Video Card for AI and Enterprise Revealed
Advanced Micro Devices has been holding back their new Vega architecture for too long — at least, according to AMD fans. Finally, the wait is over as details about the first graphics card to feature the new Vega generation has been revealed.
AMD's next generation of high-end Graphics Processing Units (GPU), which is called Vega, finally takes shape this week as an actual device with specs and more importantly, a release date. As part of their financial meeting, the company has revealed details on their next step for the Radeon line of GPUs, according to AnandTech.
The first Vega-based GPU from AMD will be called the Radeon Vega Frontier Edition, a high-end card meant to be a powerhouse for the most demanding graphics task today. The company has stated that the new GPU can run up to 25 TeraFLOPs for 16-bit floating point operations, which is more than twice of what their recent Radeon Pro Duo card can handle, according to Tom's Hardware.
The card will produce 13 teraFLOPs of 32-bit computing performance as well, giving it a clear advantage over NVidia's top-end cards. The Nvidia GTX 1080 can offer 11.3 TFLOPs in comparison. Nvidia's fastest offering, the Nvidia Titan Xp, can run 12 TFLOPs, which is short of the performance that the Radeon Vega Frontier Edition is said to be capable of, according to Digital Trends.
The graphics memory equipped with the upcoming GPU will also be unconventional. Its new High Bandwidth Memory 2 technology is capable of hauling data in and out of the card at an impressive rate of 480 gigabytes per second, which is a speed difference so huge the company is marketing the memory as "Cache" rather than the usual "Memory."
AMD says that the new memory engine of the Radeon Vega Frontier Edition can access data terabytes at a time — a design that makes it well suited for scientific and enterprise work.
The new card is expected to be released this June, but pricing information is not available as of this time.
