900-2G183-6300-T00

• Turing Architecture GPU at 12nm with 2560 CUDA cores, 320 Tensor Cores and 40 RT Cores
• 16GB GDDR6 with ECC support, 256-Bit bus, 300GB/S Memory Bandwidth
• Peak Throughput 65.1 TFLOPs FP16, 8.1 TFLOPs FP32, 254.4 GFLOPs FP64
• 585M Hz Core clock, 1590M Hz Boost Clock
• Single-Slot PCIe 3.0 x16 Form factor, 10.5in X 4.4in, Passively Cooled
• 70W TDP with No Auxiliary Power Connector Required

The NVIDIA 900-2G183-6300-T00 is a Tesla T4 professional accelerator card with 16GB of GDDR6 memory on a 256-bit bus. It is built on the Turing GPU architecture at 12nm with 2560 CUDA cores, 320 tensor cores and 40 RT cores, running at a 585M Hz core clock and 1590M Hz boost clock. The 900-2G183-6300-T00 is a passively cooled single-slot PCIe 3.0 x16 card with a 70W TDP and no auxiliary power connector.

The Tesla T4 is an inference and mixed-workload accelerator sized for dense server deployment rather than training the largest models. Its 70W passive single-slot design means a 1U or 2U node can host several cards on chassis airflow alone with no extra power cabling, which is why it became common in virtual desktop infrastructure, video transcoding, and machine learning inference fleets. ECC memory support suits production workloads where silent data corruption is unacceptable. Buyers are typically data centre operators, cloud and hosting providers, and server OEMs and integrators.