NVIDIA has unveiled two new professional graphics cards — the RTX A5000 and RTX A4000 – at today’s GTC 2021 event. Both new products are built on the latest Ampere architecture and are aimed at professional users engaged in the field of digital information visualization. Recall that last fall, NVIDIA introduced the older model of the series, the RTX A6000.
The RTX A5000 uses 8192 CUDA cores, as well as 256 tensor and 64 RT cores. The graphics card received 24 GB of GDDR6 video memory with ECC error correction and 384-bit bus support. The theoretical peak memory bandwidth is stated at 768 GB / s.
According to NVIDIA, the RTX A5000 is ready to provide performance in single-precision operations at 27.8 Tflops. The performance of RT cores in ray tracing tasks can reach 54.2 Tflops, and the performance of tensor cores — 222.2 Tflops.

The stated power consumption of the RTX A5000 is 230 watts. The video card is equipped with one 8-pin additional power connector. As part of the cooling system, a tangential fan is used. The external interfaces include four DisplayPort 1.4 connectors. The manufacturer indicates that the card will take up two expansion slots in the system.
For the new product, support for GPU virtualization is announced through tools such as NVIDIA Virtual Workstation and NVIDIA Virtual Compute Server. The card offers support for NVLink technology, allowing you to organize workstations with two similar professional graphics accelerators. The cooling system of the new product is of the tangential type. In the PC case, the video card will take up two expansion slots.

The RTX A4000 model received 6144 CUDA cores, 192 tensor and 48 RT cores. The peak performance of the RTX A4000 model in single-precision operations reaches a value of 19.2 Tflops, in ray tracing tasks – up to 37.4 Tflops, and the performance of tensor cores can reach a value of 153.4 Tflops.
The graphics card is equipped with 16 GB of GDDR6 video memory with a 256-bit bus and an ECC error correction function. The theoretical memory bandwidth is 448 GB/s.

The stated level of energy consumption of the new product is 140 watts. The video card is equipped with one 6-pin additional power connector. The external interfaces of the graphics accelerator include four DisplayPort 1.4 connectors. The graphics card is compact and requires one free expansion slot to install. For the younger podele, support for NVLink technology is not declared.