Here is our most memorable glance at the “AD103” and “AD104” chips controlling the GeForce RTX 4080 16 GB and RTX 4080 12 GB, individually, on account of Ryan Smith from Anandtech. These are the second-and third-biggest executions of the GeForce “Ada” designs engineering, with the “AD102” driving the RTX 4090 being the biggest. The two chips are based on a similar TSMC 4N (4 nm EUV) silicon creation process as the AD102, however, are fundamentally far off from it in determinations. For instance, the AD102 has a stunning 80 percent more calculating hardware than the AD103 and a 50 percent more extensive memory interface. The sheer numbers at play here, empower NVIDIA to cut out many SKUs in light of the three chips alone before we’re shown the mid-range “AD106” later on.

The AD103 bite the dust measures 378.6 mm², fundamentally more modest than the 608 mm² of the AD102, and it reflects in a much lower semiconductor count of 45.9 billion. The chip actually includes 80 streaming multiprocessors (SM), which work out to 10,240 CUDA centers, 320 Tensor centers, 80 RT centers, and 320 TMUs. The chip is blessed with a solid ROP count of 112 and has a 256-digit wide GDDR6X memory interface. The AD104 is more modest still, with a kick the bucket size of 294.5 mm², a semiconductor count of 35.8 billion, 60 SM, 7,680 CUDA centers, 240 Tensor centers, 60 RT centers, 240 TMUs, and 80 ROPs. Ryan Smith says that the RTX 4080 12 GB maximizes the AD104, and that implies its memory point of interaction is actually only 192-digit wide.