The RTX 4000 series and Radeon RX 7000 family, which are expected to be introduced this year, will take the competition in the graphics card market to the next level. The RDNA 3 architecture to be used in the AMD Radeon RX 7000 series may surpass Nvidia.
Both AMD and Nvidia will use chips produced by TSMC with 5nm technology in the graphics cards they will introduce this year. According to the leaked information, the GPU area of AD102 will be less than its rival Navi 31. It is estimated that the RDNA 3 architecture will consume less power than its competitor. According to the leaked information, power consumption will be 500W in Navi 31 and 600W in AD102.
AMD RDNA 3 will leave Nvidia Ada Lovelace architecture behind
Nvidia AD102 will be presented to users with 18,432 CUDA cores. The FP32 core value of AMD Navi 31 will be 15,360 Stream Processor. Navi 31 will surpass AD102 with a GPU clock speed of 3.0GHz.
It is stated that AMD will be one step ahead of its competitor in terms of GPU memory. While the GPU memory in Navi 31 is 32 GB, AD102 will have only 24 GB of memory.
AD102 100 TFLOP, built on Ada Lovelace architecture, and Navi 31, which will be introduced with RDNA 3 architecture, will have 92 TFLOPs of FP32 performance. It is estimated that both companies will launch their new graphics cards in the third or fourth quarter of 2022.
AD102 | NAVI 31 | |
---|---|---|
Fabrication Process | TSMC N5 | TSMC N5/N6 |
GPU Architecture | Nvidia Ada Lovelace | AMD RDNA3 |
Estimated GPU area | 600mm² | 800mm² |
GPU Cluster | 12 Graphics Processing Clusters (GPC) | 2×3 Shader Engine |
GPU Texture Set | 72 Texture Processing Clusters (TPC) | 2×30 RDNA Workgroups (WGP) |
FP32 Kernel | 18,432 CUDA | 15360 Stream Processors |
GPU Clock Speed | 2.7GHz | 3.0GHz | GPU Memory | 24GB GDDR6X | 32GB GDDR6 |
Memory Bandwidth | 21 Gbps 384- bit | 18 Gbps 256-bit |
Cache | 96 MB (L2) | 512 MB Infinity Cache |
Power Consumption | 600W | 500W |
Introduced | 2022 | 2022 |
FP32 Performance | 100 TFLOPs | 92 TFLOPs |