Rx vega 646/14/2023 The bottom line is, ranging from the beastly 345 Watt liquid-cooled RX Vega 64, to the stout 295 Watt air-cooled RX Vega 64 and 210 Watt Vega 56, AMD's Radeon RX Vega line-up promises up to 13.7 TFLOPS of single precision and 27.5 TFLOPS half precision performance. We've covered AMD's Radeon Vega architecture in depth in the past, so if you'd like to dive deeper, we'd highly suggest hitting our Vega GPU reveal from earlier this year, as well as our more recent Radeon RX Vega architecture details here. Since GPUs are addressing a wide range of workloads these days, from machine learning to data analytics and of course gaming, having the flexibility to accelerate certain operations that don't require full 32-bit precision - while maintaining the ability to deliver full precision when needed - is a feature that AMD feels confident will be a key advantage for the architecture moving forward. ![]() ![]() In short, Vega's compute units support both 32-bit and 16-bit packed math, the latter of which halves the precision but also doubles peak floating point and integer processing throughput. Vega also offers a new mixed-precision support AMD is calling Rapid Packed Math. ![]() Radeon RX Vega is an all-new GPU architecture for AMD, built from the ground-up to be optimized for DirectX 12 and asynchronous compute, with a revised pixel engine and higher clock speeds that the previous generation Fiji architecture could not achieve.
0 Comments
Leave a Reply. |