Intel Arc A770M die size confirmed with TSMC 6 nm node helping the new GPU to offer 21.7 billion transistors and 13.5 TFLOPs of performance
Intel has finally announced its Arc Alchemist GPUs, albeit only for laptops at the moment. While Intel has shared many details about its five new GPUs, it left a few details out of its presentation. For example, the company did not list any die sizes, nor FP32 theoretical performance, for example.
However, Intel supplied some additional information to Hardware Unboxed about the Arc A770M, which it bases on the ACM-G10 GPU. According to Hardware Unboxed, Intel has built the Arc A770M on TSMC N6 nodes, billed as delivering 18% higher logic density than TSMC's N7 process. In comparison, NVIDIA uses Samsung's 8 nm nodes for the RTX 30 series, but Apple has adopted TSMC's N5P process instead.
As the table below shows, the Arc A770M has a larger die than the Radeon RX 6800M and the GeForce RTX 3070 Ti, along with more transistors than them both too. In comparison, the M1 Max features more than double the transistors that the Arc A770M has and with a larger die to boot. The Arc A770M should deliver up to 13.5 TFLOPs of performance though, only second to the GeForce RTX 3070 Ti of the four represented below.
In other words, the Arc A770M could compete with current upper mid-range laptop GPUs. Mind you, gaming performance also relies on drivers, so there is no guarantee that the Arc A770M will have the edge over the Radeon RX 6800M in all workloads. Still, the Arc A770M seems interesting on paper, even if Intel does not plan to release it until the summer.
|Intel Arc A770M||AMD Radeon RX 6800M (Navi 22)||Apple M1 Max||NVIDIA GeForce RTX 3070 Ti (GA104)|
|Die size||406 mm²||336 mm²||475 mm²||392 mm²|
|GPU||Intel ACM-G10||AMD Navi 22||M1 Max||NVIDIA GA104|
|FP32 performance||13.5 TFLOPs||12.24 TFLOPs||10.6 TFLOPs||15.88 TFLOPs|
|Manufacturing process||TSMC N6||TSMC N7||TSMC N5P||Samsung 8N|
|Transistors||21.7 billion||17.2 billion||57 billion||17.4 billion|