Nvidia on Monday introduced the RTX A6000 and A40 ‘Ampere’ GPUs for high-end workstations and data centres without the familiar branding, in a development that many believe may be the end of the road for its ‘Quadro’-branded professional graphics cards. The move follows the recent decision by the company to replace the Tesla V100 with the Nvidia A100 and likely discontinue the branding of ‘Titan’ for its top-end market cards.
Nvidia RTX A6000 and Nvidia A40: Features and Specifications
The new GPUs carry heavy-duty computing power that the company claims will speed up artists, designers and engineers’ rendering, AI, VR/AR and computing workloads. They include new RT Cores, Tensor Cores and CUDA cores, based on the NVIDIA Ampere architecture.
The second-gen RT cores deliver up to 2x the performance of the previous version, plus simultaneous ray tracing, shading and computing, according to Nvidia.
Meanwhile, the third-generation tensor cores provide up to 5x the previous-generation throughput, up to 10X with sparsity, with support for new TF32 and BF16 data formats. As for the latest CUDA cores, for major graphics and computation increases, they offer up to 2x the FP32 throughput of the previous generation.
Both cards come with Nvidia’s Ampere architecture in terms of hard numbers and carry 10,572 CUDA cores alongside 48 GB of VRAM. Other specifications, including 336 tensor cores, 384-bit memory bus width and 300W TDP, are also relatively similar.
Nevertheless, although the former is intended for traditional workstations,the latter is intended for use in commercial data centres and server farms.Also, the A600 is clocked at 16 Gbps while both come with GDDR6 RAM, while the A40 comes with a 14.5 Gbps memory clock. For video output, the cards will only use DisplayPort and will connect to computers via PCIe Gen 4.
Price and Availability for Nvidia RTX A6000 and Nvidia A40
Nvidia has yet to reveal either an exact release date or pricing for its latest GPUs, but says the RTX A6000 will be available from a variety of channel partners beginning in mid-December, including PNY, Leadtek, Ingram Micro, Ryoyo and nvidia.com. They will also be available beginning next year from leading manufacturers of systems, such as Dell, HP, Lenovo and Boxx.
As for the A40, beginning early next year, it will also be available from OEM workstation and server vendors, including Cisco, Dell, Fujitsu, Hewlett Packard Enterprise and Lenovo.