Introduction

The AMD Vega graphics architecture has been grabbing headlines for well over a year now and finally makes its debut in the consumer-segment with the Radeon RX Vega series. When AMD launched its "Polaris" architecture with the Radeon RX 480, which could at best compete with upper-mainstream graphics cards from NVIDIA, such as the GeForce GTX 1060, consumers were half-expecting Vega to be just around the corner, but were put through another year's wait. It's finally here, as the $499 Radeon RX Vega 64, and its little, more exciting sibling, the $399 RX Vega 56.

It appears as if the launch of the Radeon RX Vega series took longer than it did because it's been over two years since AMD's last high-end graphics card launch. The company launched the Radeon R9 Fury series in mid-2015, but has since left the market segment unattended. NVIDIA dominated several key price points, including the $399 point with its GTX 1070, $499 with its GTX 1080, and $649 with its GTX 1080 Ti, and its TITAN series SKUs well above the $1000 mark. The $399 performance-segment price point has been particularly important for both NVIDIA and AMD, and how AMD has left it unattended has been quite surprising. We can't imagine that it would have been difficult to build a slightly bigger "Polaris" based GPU with conventional technologies, such as GDDR5 memory, to take a swing at the $399 price-point; at least now, AMD has something to compete with in the Radeon RX Vega 56. AMD has greatly simplified its nomenclature with the RX Vega series. The term "Vega" signifies the architecture on which the GPU is based on, and the number after that signifies the compute unit (CU) count.



"Vega 10" is a multi-chip module, much like "Fiji," which powered the R9 Fury series. It combines a large GPU die based on the Vega architecture with two HBM2 memory stacks. On the RX Vega series, there is 8 GB of total memory; while on some of the Pro Vega SKUs, the chip has 16 GB of it. AMD is referring to this memory as "High Bandwidth Cache," with the operative term being "cache" as opposed to "memory." This is because of sweeping changes the company brought about in the way the GPU addresses video memory. The GPU features a large virtual address-space spanning several terabytes, where a small portion of it is physical and the GPU juggles "hot data" (most frequently accessed data) in and out of it. With this, AMD claims to overcome many of the fundamental memory-management issues that had been inherent to its Graphics CoreNext architecture, which it had been working around by throwing brute memory bus width and data compression at it.

The Radeon RX Vega 56 features 56 of the 64 next-generation Compute Units (NGCUs) present on the "Vega 10" silicon, which works out to a stream processor count of 3,584. This proportionately lowers the TMU count to 224. The ROP count is left untouched, at 64. The card features 8 GB of HBM2 memory across a 2048-bit wide memory interface. AMD has also tempered down clock speeds a bit, with its core running at up to 1156 MHz and memory at 800 MHz, yielding a memory bandwidth of 409 GB/s, less than the 484 GB/s of the RX Vega 64. This dialing down of clock speeds may have had a profound impact on energy efficiency, and if the performance lives up, AMD could finally have a price-performance sweetspot product to take on the GTX 1070, given its $399 price.

We put the RX Vega 56 through our vast and modern test bench and had it compete against key NVIDIA SKUs, such as the GTX 1060 6 GB, GTX 1070, GTX 1080, and GTX 1080 Ti; at resolutions as high as 4K Ultra HD. If this SKU can tame 4K, it will be the first AMD GPU to do so.