NVIDIA A10 GPU

Performance

The NVIDIA A10 GPU leverages Ampere architecture which is built on Samsung’s 8nm process and based on the GA102 graphics processor. Unlike the GeForce RTX 3090 Ti, which uses the same GPU, the A10 features 9216 shading units, 288 third-generation tensor cores, 96 ROPs, and 72 second-generation ray tracing cores, which is a little less than the 10752 cores and GDDR6X memory offered on the RTX 3090 TI. The single-slot GPU also supports a base clock of 885 MHz and boosts up to 1695 MHz. A10 is suited for high-performance virtual workstations running professional visualization applications when combined with NVIDIA RTXTM Virtual Workstation (vWS) software, or for multimedia-rich virtual desktops when combined with NVIDIA Virtual PC (vPC) software. A variety of other applications can all be supported including deep learning, machine learning, cloud gaming, and more. The card supports all AI frameworks and network models because it is part of the NVIDIA AI Platform, resulting in substantial performance and efficiency increases.

Memory

The full-length, single-slot GPU features 24 GB of GDDR6 VRAM and provides a bandwidth of 600 GB/s. It runs at 12.5 Gbps via a 384-bit wide bus interface. It’s compatible with PCIe 4.0 and uses a x16 link slot to maximize throughput.

Features

The A10 is powered by an 8-pin EPS connector with a maximum TDP power of up to 150 W, a 450W system PSU is recommended. As a PCIe card, it is connected directly to the servers that support NVIDIA’s RTX Virtual Workstation software. To operate within its thermal envelope, this card is passively cooled and requires system airflow. Although it’s considered a workhouse, it does not have any display connectivity as this card is geared towards server deployments and virtualized workloads.

Summary

Combining 2nd generation NVIDIA Tensor RT cores and 24 GB of GDDR6 memory in a single slot, the NVIDIA A10 GPU is ideal for multi-application workflows and remote collaborations. Its PCIe 4.0 form factor is compatible with a variety of servers from worldwide OEMs and includes secure hardware root- of-trust technology, ensuring that firmware isn’t tampered with or corrupted.