On April 17th, NVIDIA announced the release of two new professional graphics cards, the RTX A1000 and RTX A400. Their numbering indicates they're positioned as entry-level, and they are based on the previous-generation Ampere architecture, not the latest Ada Lovelace.
It's worth noting that the last time NVIDIA introduced a new professional card based on the Ampere architecture was two and a half years ago with the RTX A2000.
The RTX A1000/A400 both feature a single-slot, half-height design, making them incredibly compact at just 50W power consumption, easily cooled by a single fan without the need for external power.
They both integrate the 2nd generation RT core, 3rd generation Tensor core, 7th generation NVENC encoder, and 5th generation NVDEC decoder (supporting AV1).
The core of the RTX A1000, like the RTX A2000, is likely GA106-based, with 2304 CUDA cores, 72 Tensor cores, and 18 RT cores, running at a core boost frequency of 1463MHz, delivering FP32 floating-point performance of 6.74 TFlops, which is even lower than that of the RTX 3050.
Memory remains at 8GB GDDR6 with a 128-bit bus width, offering a bandwidth of 192GB/s.
The core of the RTX A400 should be the GA107, featuring only 768 CUDA cores, 24 Tensor cores, and 6 ray tracing cores. Its boost clock is 1758MHz, with floating-point performance of 207 TFlops. It also comes with a modest 64-bit 4GB GDDR6 memory, offering a bandwidth of 96GB/s.
The RTX A1000 has already shipped, with the RTX 400 following suit next month.
