And yet, rather than the 4GB ceiling that dogged Radeon R9 Fury X, RX Vega 56 comfortably offers 8GB using 4-hi stacks, similar to the Vega 64. Since Vega 10’s clock rates on the 56-CU card can get up to ~40% higher than Fiji’s, while memory bandwidth actually drops by 102 GB/s, a larger cache should do even more to help prevent bottlenecks here than on the flagship.Īdoption of HBM2 allows AMD to halve the number of memory stacks on its interposer compared to Fiji, cutting an aggregate 4096-bit bus to 2048 bits. Ideally, this means the GPU goes out to HBM2 less often, reducing Vega 10’s reliance on external bandwidth. That L2 is now 4MB in size, whereas Fiji included 2MB of L2 capacity (already a doubling of Hawaii’s 1MB L2). These render back-ends become clients of the L2, as we already know. Using AMD’s peak compute performance figures, that knocks theoretical SP performance down from 13.7 TFLOPS to 10.5 TFLOPS.Įach of Vega 10's Shader Engines sports four render back-ends capable of 16 pixels per clock cycle, yielding 64 ROPs. Radeon RX Vega 56's base is 1156 MHz compared to Vega 64's 1274 MHz, while Vega 56's boost frequency is rated at 1471 MHz versus Vega 64's 1546 MHz. This configuration takes another hit to SP and texturing throughput in the form of lower base and typical boost clock rates. Given 64 Stream processors and four texture units per CU, you get 3584 Stream processors and 224 texture units-~88% of Vega’s NCU resources. You’ll still find four Shader Engines under the hood, each with its own geometry processor and draw stream binning rasterizer.īut rather than 64 active Compute Units across those Shader Engines, AMD turns off two CUs per Shader Engine, leaving 56 enabled across the GPU. It’s a 486 mm² behemoth sporting 12.5 billion transistors manufactured on GlobalFoundries’ 14LPP platform. Radeon RX Vega 56 utilizes the same Vega 10 processor found in Vega 64.