AMD has unveiled two workstation GPUs, the Radeon Pro W7900 and Radeon Pro W7800, based on the advanced RDNA 3 architecture. These high-performance GPUs, designed for workflows such as visualization, real-time 3D, ray trace rendering, and more, are equipped with unified RDNA 3 compute units, dual-issue stream processors, AI accelerators, and second-generation ray tracing accelerators. According to AMD, RDNA 3 delivers up to a 50% increase in raytracing performance per compute unit compared to the previous generation.

The Radeon Pro W7900 is a triple (2.5) slot GPU, packing 48 GB of GDDR6 memory and 61 TFLOPs of peak single precision performance with a total board power of 295W, priced at $3,999. On the other hand, the Radeon Pro W7800 is a dual slot GPU, featuring 32 GB of GDDR6 memory and 45 TFLOPs of peak single precision performance with a total board power of 260W, available for $2,499.

Both GPUs are engineered to rival Nvidia’s RTX 6000 Ada Generation (48 GB). While AMD’s Radeon Pro W7900 outperforms Nvidia’s flagship pro GPU in certain benchmarks, AMD emphasizes the competitive price-performance ratio of both GPUs. For instance, in the SPECviewperf 2020 GeoMean benchmark, AMD claims the Radeon Pro W7900 comes within 7% of the Nvidia RTX A6000 Ada Generation’s performance, but at less than half the cost ($3,999 vs $8,615), delivering more than double the price-performance.

The newly launched GPUs also feature support for DisplayPort 2.1, a significant upgrade from the previous generation that allows for increased refresh rate, pixel resolution, and color bit-depth. This ensures these GPUs are future-proofed for next-gen displays. Both the Radeon Pro W7800 and W7900 offer three DisplayPort 2.1 and one Mini DisplayPort 2.1 connectors, contrasting with the previous Radeon Pro W6800’s six Mini DisplayPort 1.4 connectors.

The Radeon Pro W7900, with its 48 GB memory, offers a 50% increase from its predecessor, the Radeon Pro W6800, matching the Nvidia RTX 6000 Ada’s memory capacity. As workflows become increasingly complex, larger memory capacity is crucial for handling high-polygon datasets and multitasking. Both GPUs are expected to be available in Q2 2023, with OEM and SI system availability expected in 2H 2023.

#AMD #GPU #red_team