THE WORLD'S MOST ADVANCED GPU MEMORY ARCHITECTURE
This represents an entirely new memory hierarchy for Vega GPUs featuring a new high-bandwidth cache and a controller. The cache makes use of performance HBM2 (High Bandwidth Memory 2) technology which allows double pin transfer rates up to 2 GT/s (gigatransfers per second), and as much as 8 gigabytes of RAM per package.
Another advantage of HBM2 is that it halves the amount of physical space needed on a board compared to an equivalent amount of GDDR5 memory.
Vega architecture is optimized for streaming very large datasets and can work with a variety of memory types with up to 512 terabytes of virtual address space.
NEXT-GENERATION GEOMETRY PIPELINE
Vega's next-generation geometry pipeline not only brings higher levels of efficiencies for processing complex geometry, but at the same time allows for more than 200 percent of the throughput-per-clock compared to previous Radeon architectures.
Vega also brings with it better load-balancing with an intelligent workload distributor to deliver consistent performance.
NEXT-GENERATION COMPUTE ENGINE
Vega's next-generation compute engine can natively process 8-bit, 16-bit, 32-bit, or 64-bit operations in each clock cycle, and these compute units are optimized for significantly higher frequencies than previous generations and their support of variable datatypes makes the architecture highly versatile across workloads.
ADVANCED PIXEL ENGINE
The new Vega pixel engine employs a Draw Stream Binning Rasterizer, designed to improve performance and power efficiency, which allows for "fetch once, shade once" of pixels through the use of a smart on-chip bin cache and what AMD describes as "early culling of pixels invisible in a final scene."
Vega's pixel engine is now a client of the onboard L2 cache, which offers sizeable reduction in overhead for graphics workloads that make use of heavy read-after-write operations.