GPUEmbeddedReleased —

NVIDIA RTX A2000 Embedded

8 GB GDDR6 · 128-bit · 2,560 cores

NVIDIA RTX A2000 Embedded is an Ampere-generation professional GPU delivered as an MXM 3.1 Type A module for industrial and embedded systems. The GA107 die provides 2,560 CUDA cores, 80 third-generation Tensor Cores and 20 second-generation RT Cores on the Samsung 8 nm process. It carries 8 GB of ECC-protected GDDR6 on a 128-bit bus for 224 GB/s of bandwidth, connects over PCIe 4.0 x8 and fits a 60 W total graphics power budget, with long-life availability for designs that must stay in production for years. The similarly named RTX A2000 Laptop GPU is a different SKU.

General

Manufacturer: NVIDIA
Model: RTX A2000 Embedded
Segment: Embedded
GPU chip: GA107
Architecture: Ampere
Release date: —
Price: Search on Amazon

Memory

Memory size: 8 GB
Memory type: GDDR6
Bus width: 128-bit
Bandwidth: 224 GB/s
Memory clock: 1,750 MHz

Clocks & silicon

GPU clock: —
Boost clock: —
Shader units: 2,560 CUDA Cores
Tensor / AI cores: 80
RT cores: 20
Transistors: 8.7 billion
Technology (nm): 8 nm
Samsung 8N (8 nm)
TDP / TBP: 60 W

Interface & APIs

Bus interface: PCIe 4.0 x8
SLI / multi-GPU: None
DirectX: 12 Ultimate (12_2)
OpenGL: 4.6
Vulkan: 1.4
Other APIs: OpenCL 3.0, CUDA compute capability 8.6

AI support

Precisions: FP8FP16BF16INT8INT4TF32FP64
AI accelerator: 3rd-gen Tensor Cores (80)
Peak AI TOPS: —
Notes: The third-generation Tensor Cores in Ampere add TF32 and BF16 alongside FP16 and support structural sparsity for up to double the effective throughput. INT8 and INT4 are available for quantised inference. The module also carries ECC memory protection, which is uncommon in this power class. FP64 is present but runs at one sixty-fourth of the FP32 rate, so this is an inference and graphics part rather than a double-precision compute part.

All NVIDIA GPUs →

Embedded gpus from NVIDIA announced around the same time as the NVIDIA RTX A2000 Embedded. Add any of them to your compare list to see the specs side by side.

gpu