What the story regarding RTX A6000?
NVIDIA declared it would dispatch two executive items dependent on Ampere GPUs: A40 and RTX A6000 were to include GA104 and GA102 GPUs, respectively. The real shock was that neither of those graphics cards was to feature Quadro or Tesla branding.
NVIDIA has ended its Quadro and Tesla line-up by presenting a lone-series of products, namely NVIDIA RTX Axx or NVIDIA Axx. The graphics cards must carry the RTX brand-name in light of the fact that these cards will supplant the Quadro series, while the non-RTX A40 card is fundamentally a Tesla replacement that will go with the all-around dispatched GA100 A100 accelerator.
The RTX A6000 is the solitary graphics card dependent on GA102 GPU to contain all active CUDA-cores. The GPU has a Single-Precision computation performance of up to 38.7 TFLOPs, due to the presence of 10752 CUDA cores, enabling it to execute 3.1 TLFOPs more than GeForce RTX 3090.
The RTX A6000 includes double the memory limit of 48 GB, different from the RTX 3090. These modules are presently accessible with GDDR6 memory innovation; thus, memory bandwidth is lower on the previous chief model.
NVIDIA RTX A6000 sports four DisplayPort 1.4 connectors but doesn’t have any HDMI 2.1 port. The cooler’s design resembles the blower, and it is not similar to a data center card such as A40 that has a mildly cooled heatsink. A new feature, NVLink bridge, lets you connect only two cards. The A6000 is meant for workstations, supporting NVIDIA vGPU Virtualization Technologies.
Sources: Twitter, NVIDIA Design, 15th December 2020.