AMD announced its new
The Navi graphics cards and their succeeding generations won’t be handicapped by the limitations of GCN. Till now Radeon GPUs had a maximum shader count of 4,096, with up to 4 Shader Engines per die. This also limits the ROPs to 64 per chip, reducing the efficiency of AMD’s streaming processors, compared to NVIDIA’s CUDA Cores. According to the Navi die shared a while back on Twitter, the RDNA macro-architecture won’t be held back by any of these issues.
- 7nm Zen 2 Based Ryzen 5 3600 +63% Faster than Intel Core i5-9400
- AMD Radeon RX 5700 GPU (Navi) Faster than NVIDIA GeForce RTX 2070 While Being Almost Half as Large
This is probably where the improved 1.25x performance per clock comes from, while the 7nm node helps with keeping the TDP low. The higher IPC should make AMD cards more suitable for gaming and future Radeon GPUs will be much more efficient than the Vega and Polaris products.
This is not all. The RDNA design also incorporates a revamped graphics pipeline aimed at delivering the maximum performance per clock cycle and higher clock speeds. The new shader engine has also resulted in a redesigned compute unit which further bolsters the gaming capability of the Navi cards.
Lastly, the memory and cache have been upgraded as well. The RDNA architecture features a multi-level cache hierarchy, reducing the latency of predictive or repetitive workloads. The Navi graphics cards will also be AMD’s first gaming product to leverage the GDD6 memory just like NVIDIA’s Turing cards. All these factors will level the playing field between the two long-time rivals. Be sure to catch AMD at E3 for more details and possibly the launch of the Radeon RX 5700 lineup.