NVIDIA introduces new Ampere-based RTX GPUs tailored for professional applications. Named the RTX A1000 and RTX 400, these models feature a sleek, minimalist design. They debut approximately two and a half years after the release of the RTX A2000 12GB, which marked the latest addition to the Ampere series, and nearly three years following the last significant update for Ampere RTX workstation models.
As depicted in official renders, the GPUs sport a low-profile, single-slot design. Notably, both cards operate without external power connections, as their power consumption is capped at 50W. This marks the lowest TDP among all contemporary NVIDIA GPUs tailored for the ProViz market.
The RTX A1000 seems to utilize the GA106 GPU architecture, boasting 2304 CUDA cores, 72 Tensor Cores, and 18 RT Cores. It achieves a maximum FP32 compute performance of 6.74 TFLOPS with a boost clock of 1463 MHz. The card is outfitted with 8GB of GDDR6 memory on a 128-bit memory bus, offering a maximum bandwidth of 192 GB/s at a memory speed of 12 Gbps according to official specifications.
NVIDIA has yet to officially announce the GPU powering the RTX 400, but indications suggest it may be based on the GA107 architecture. It purportedly features 768 CUDA cores, 24 Tensor Cores, and 6 RT Cores, delivering around 2.7 TFLOPS of FP32 compute performance with a clock speed of approximately 1758 MHz. This model is equipped with 4GB of GDDR6 memory on a 64-bit memory bus, offering a bandwidth of 96 GB/s.
In contrast to ADA-based models, these cards do not include AV1 encoding support; nevertheless, they do offer decoding and encoding capabilities for other codecs such as H264 or H265. Although NVIDIA hasn’t disclosed pricing for these cards, it’s confirmed that two board partners (PNY and Ryoyo Electric) will be offering them for sale. The RTX A1000 is already available, while the RTX 400 is set to commence shipping next month.
Source: NVIDIA