NVIDIA announces A2 Tensor Core GPU: GA107 GPU + 16GB GDDR6 memory

NVIDIA has just announced its new A2 Tensor Core, its latest entry-level Ampere-based accelerator using the GA107 GPU which packs 1280 CUDA cores, with 16GB of GDDR6 memory.

4

The new NVIDIA A2 Tensor Core is quite a bit cut down compared to the A16 Tensor Core, which has 4 x GA107 GPUs compared to the 1 x GA107 GPU on the new A2 Tensor Core. That huge cut down means the TBP (total board power) drops to just 60W, but you can also support it at 40W, meaning you don't need external power.

NVIDIA is clocking the GA107 GPU at 1440MHz base, 1770MHz boost, offering up to 4.5 TFLOPs of single-precision compute power. This means the low-end GeForce RTX 3050 Ti Mobile GPU beats the A2 Tensor Core, with the RTX 3050 Ti Mobile GPU packing up to 7.1 TFLOPs of compute performance.

4

4

NVIDIA compares the new A2 Tensor Core up against its entry-level Tensor T4, the previous-gen Turing-based low-end accelerator. NVIDIA says the new A2 Tensor Core will be up to 20-30% faster in T4 benchmarks, while offering 60% better price-to-performance ratio.