After a disappointing no-show from NVIDIA concerning the 1080 Ti, I was holding out hope that AMD might finally drop their new Vega GPU architecture. That didn’t happen, but we did get more info on what features AMD’s Vega will have.
Raja Koduri, chief architect and senior vice president of the Radeon Technologies Group, took to the stage at the AMD Tech Summit and detailed some of the upgrades coming with Vega’s architecture. One of the biggest assets the AMD Vega platform will bring is the use of flexible memory hierarchy using HBM 2. AMD is replacing the frame buffer concept with what they’re calling a “high-bandwidth cache.”
By keeping the fastest memory closer to the GPU, AMD wants to use HBM2 as a high-bandwidth cache to help improve performance when the GPU is utilizing system memory and storage. HBM2 offers an up-to-8x capacity increase per vertical stack compared to first-gen HBM. If AMD Vega is utilizing the same 4-hi 700 MHz HBM2 stacks NVIDIA used in their Tesla P100 accelerators, Vega would have 16 GBs of RAM capable of 720GB/s throughput.
Vega also uses an improved geometry pipeline that has a throughput of 11 polygons per clock cycle, up from four in AMD’s previous generations of GPU. A new primitive shader stage in Vega’s geometry pipeline allows this 2.75x boost. AMD Vega’s Next-Generation Compute Unit brings 64 shaders which allow for 128 32-bit operations per clock cycle. The Vega’s new rasterizer enables a fetch-once, shade-once approach which culls pixels that are invisible to the final scene. AMD is also changing cache hierarchy with the Vega platform by making the render back-ends clients of the L2 cache which allows for shared point synchronization of pixel and texture memory access.
What does all this mean for gaming? An on-stage demo of DOOM running at 4K on ultra on a PC using an AMD Vega architecture GPU clocked around 70 FPS using the Vulkan API. That’s roughly comparable to between the performance of a GTX 1080 and Titan X. No specific models or pricing have been released for any GPUs using AMD’s Vega architecture, but with what we’ve seen so far of Vega, NVIDIA’s 10-series GPUs will have a rival when AMD Vega products hit the market sometime in the first half of 2017.
AMD has also opened a site where you can watch some promo videos for the AMD Radeon Vega and sign up to win one of the Vega GPUs when they debut.
Jason Faulkner posted a new article, CES 2017: AMD Says Vega GPUs Not Ready Yet