900-2G171-0000-100

• NVIDIA Tesla A16, Ampere architecture, 8 Nm Lithography
• 64 GB GDDR6 memory, 128-Bit bus, 200 GB/S bandwidth, ECC Enabled by Default
• 1280 CUDA cores, 40 Tensor Cores, 10 RT Cores
• 1312 MHz Core clock, 1755 MHz Boost Clock
• Single-Slot PCIe 4.0 x16 Form factor, 10.5 in X 4.4 in, Passive Cooling
• 250 W TDP, Single 8-Pin Power connector, 600 W Recommended PSU
• Certified for NVIDIA vPC, V Apps, RTX vWS, AI Enterprise and Virtual Compute Server vGPU Software

The NVIDIA 900-2G171-0000-100 is a Tesla A16 accelerator card built on the Ampere architecture at 8 nm, carrying 64 GB of GDDR6 memory on a 128-bit bus with 200 GB/s of bandwidth and ECC enabled by default. It provides 1280 CUDA cores, 40 Tensor Cores and 10 RT cores, with clocks of 1312 MHz base and 1755 MHz boost. The card is a single-slot PCIe 4.0 x16 board with passive cooling and a 250 W TDP, and this SKU is supplied without CEC. Rated throughput is 4.5 TFLOPS FP32, 17.9 TFLOPS FP16 and 135.6 GFLOPS FP64.

The Tesla A16 is positioned for virtual desktop infrastructure and remote graphics: its 64 GB of frame buffer and broad vGPU software certification let one card serve a high number of concurrent virtual desktop or virtual application users. The passive single-slot design means it depends on chassis airflow and is intended for rack servers rather than workstations, and the 250 W TDP with a single 8-pin connector keeps it inside standard server power budgets. Typical buyers are enterprise IT teams running VDI at scale, managed service and cloud desktop providers, and education, healthcare and public-sector organizations consolidating graphics workloads into the data center.