Yesterday, NVIDIA unveiled RTX A6000 and A40 ‘Ampere’ GPUs for high-end workstations and data centers without the familiar branding in a move that many believe could be the end of the road for its ‘Quadro‘ branded professional graphics cards. It must be noted that it is a follow-up move of the company’s recent decision to replace the Tesla V100 with the NVIDIA A100. Here, the company might possibly discontinue the ‘Titan’ branding for its top-end consumer cards with this strategic move.
According to the company, the new GPUs bring heavy-duty computing power will accelerate rendering, AI, VR / AR, and compute workloads for artists, designers, and engineers.
The GPUs are built on the NVIDIA Ampere architecture and they come with RT Cores, Tensor Cores, and CUDA cores. The company claims that the second generation RT cores offer up to 2x the throughput of the previous generation, along with concurrent ray tracing, shading, and computational ability. On the other hand, the third generation Tensor cores provide up to 5x the throughput of the previous generation, up to 10x with sparsity. It also offers support for TF32 and BF16 data formats.
Last, but not the least, the new CUDA cores deliver up to 2x the FP32 throughput of the previous generation. This provides significant increases in graphics and computational ability.
As far as hard numbers are concerned, both the graphics cards come with NVIDIA’s Ampere architecture. They are powered by 10,572 CUDA cores alongside 48 GB VRAM. More or less, other specifications are similar including 336 tensor cores, 384-bit memory bus width and 300 W TDP.
NVIDIA RTX A6000 is meant for standard workstations while NVidia A40 ‘Ampere is intended for use in commercial data centers and server farms. Both the GPUs come with GDDR6 RAM. Nvidia RTX A600 is clocked at 16 Gbps while NVidia A40 is clocked at 14.5 Gbps. Both the cards will only use DisplayPort for video output and will connect to computers through PCIe Gen 4.
It should be noted that though the company has unveiled two GPUs, it has not yet announced the release date or price of any of the GPUs. The company only said that the RTX A6000 will be available from a number of channel partners, including PNY, Leadtek, Ingram Micro, Ryoyo, and on nvidia.com starting from mid-December. The company also said that notable systems manufacturer stores such as stores of brands like Dell, HP, Lenovo, and Boxx will also sell these GPUs from next year onwards.
On the other hand, the company said the A40 GPU will be available in OEM workstation and server vendor stores like that of Cisco, Dell, Fujitsu, Hewlett Packard Enterprise, and Lenovo from early next year.