On April 17th, NVIDIA unveiled two new professional GPUs, the RTX A1000 and RTX A400. As indicated by their model numbers, these cards are positioned as entry-level options and are based on the previous-generation Ampere architecture, rather than the latest Ada Lovelace architecture.
It's worth noting that the last time NVIDIA introduced new professional GPUs based on the Ampere architecture was over two and a half years ago with the RTX A2000.
Both the RTX A1000 and A400 feature a single-slot, half-height design, incredibly compact due to their modest 50W power consumption. With just a single fan, they require no additional external power supply.
Both cards integrate the second-generation RT cores, third-generation Tensor cores, seventh-generation NVENC encoder, and fifth-generation NVDEC decoder (supporting AV1).
The RTX A1000 is likely based on the GA106 chip, similar to the RTX A2000, featuring 2304 CUDA cores, 72 Tensor cores, and 18 RT cores. With a core boost clock of 1463MHz, it delivers FP32 floating-point performance of 6.74 TFLOPs, which falls short even compared to the RTX 3050.
The memory remains at an 8GB GDDR6 with a 128-bit bus, offering a bandwidth of 192GB/s.
The core of the RTX A400 is likely the GA107, featuring only 768 CUDA cores, 24 Tensor cores, and 6 ray tracing cores. It operates at a boost frequency of 1758MHz, delivering floating-point performance of 207TFlops. Its memory configuration includes a modest 64-bit 4GB GDDR6 with a bandwidth of 96GB/s.
The RTX A1000 is now shipping, with the RTX 400 expected to follow suit next month.
