NVIDIA has slightly confused the offer of its next-generation accelerators, looking at the difference between the A100 and A40. This time, however, the matter is easier, because the NVIDIA A2 accelerator, which ranks at the lowest level in the hierarchy, has been released.
The premiere of NVIDIA A2, the weakest accelerator of the company, has just taken place
A2 Tensor Core is the least efficient basic graphics accelerator, which reached for the 8 nm Ampere GA107 graphics processor with 1280 CUDA cores, which is 2304 less than the weakest accelerator (A30) so far. A2 has a lot in common with the A16, which is based on four GA107 GPUs, because it shares 16 GB of GDDR6 memory on a 128-bit interface, although of course for a single, not all four cores.
Also read: NVIDIA DLSS in future games. Where will you experience higher liquidity?
This core has a base clock of 1440 MHz and a boost clock of 1770 MHz, the GPU offers up to 4.5 TFLOPS of single-precision processing power. It is not one of the most efficient, but it introduces to NVIDIA’s offer a very, very energy-efficient solution, because it devours 60 W in performance mode or 40 W when properly set, which frees it from the need to provide external power.
Also read: Ray Tracing in Battlefield 2042 for PC only. We know the NVIDIA and DICE plan
The NVIDIA A2 is compared to the entry-level Turing T4 accelerator, which was the weakest model of the previous generation. According to NVIDIA, it will be 20 to 30% faster than the T4, offering 60% better price-performance ratio and 10% better energy efficiency compared to the T4.