AMD Shows Off Industry First 7nm Vega GPU at Computex
The company promises that the Vega architecture will run faster and more efficiently on the new process.
AMD tooks the wraps off its newest GPU developments today during Computex. CEO Lisa Su announced that the company plans to have the new 7nm parts inside consumer graphics cards in the future. AMD’s Radeon Technology Group (RTG) released cards based on the Vega architecture at the end of last summer following months of speculation and delays. The 16nm Vega cards arrived in two flavors, a professional work card carrying 16GB of HBM2 memory and the consumer variants, known as RX Vega 64 and RX Vega 56. The latter parts failed to compete with NVIDIA’s 1080 Ti despite higher power usage and inconsistent pricing (due to demand and cryptocurrency influence). New Vega cards running on the 7nm process will be available by the end of this year, according to AMD.
The first public demonstration of the 7nm Vega chip, known as Vega Instinct, was on a professional workstation card carrying 32GB of HBM2 memory. AMD demoed the GPU running a Cinema4D rendering workload with the company's open-source Radeon Pro Render ray tracing solution. This card is likely designed to compete with NVIDIA’s Tesla V100 cards, which also have up to 32GB and were demoed at GTC this year.
It would appear that AMD has beaten both Intel and NVIDIA to the punch with its 7nm process. Intel has been struggling for a while to iron out the bugs with its own 7nm process. Rumors continue to swirl about NVIDIA’s plans to follow up its successful lineup of Pascal-based GPUs. It is unknown if the company’s new GPU will use a new 7nm process or stick with the current process for one more generation. It has been more than two years since the GTX 1080 debut in the summer of 2016. There is no concrete information on when gaming cards based on the 7nm Vega process will be available.
-
Shacknews
replyChris Jarrard posted a new article, AMD Shows Off Industry First 7nm Vega GPU at Computex
-
Wouldn't that be something. The company that had to sell off its own fabs ends up on a sub 10nm node faster than both its major competitors.
