Support Matrix for NVIDIA NIM for Object Detection#

This documentation describes the software and hardware that NVIDIA NIM for Object Detection supports.

CPU#

Release 2.0.0 supports amd64 and arm64 container architectures. The host CPU architecture must match the container architecture. For a list of supported systems, refer to the NVIDIA Certified Systems Catalog.

Models#

Object Detection NIM supports the following models.

Publisher

Model Name

Model ID

Model Card

Demo / Explore

NVIDIA

Nemotron Page Elements v3

nvidia/nemotron-page-elements-v3

Model Card

Demo

NVIDIA

Nemotron Table Structure v1

nvidia/nemotron-table-structure-v1

Model Card

Demo

Supported Hardware and Memory Footprint#

Release 2.0.0 supports FP16 inference on the following GPU SKUs. The approximate GPU memory sizes are the startup measurements from the latency configuration after forward buffers are allocated. Runtime GPU memory use can increase with larger inputs and input-dependent working buffers.

GPU SKU

GPU

Nemotron Page Elements v3 approximate GPU memory size (GiB)

Nemotron Table Structure v1 approximate GPU memory size (GiB)

Precision

NVIDIA-A100-SXM4-80GB

NVIDIA A100 SXM4 80GB

1.05

1.05

FP16

NVIDIA-A10G

NVIDIA A10G

0.88

0.88

FP16

NVIDIA-B200

NVIDIA B200

1.26

1.26

FP16

NVIDIA-GB10

NVIDIA GB10 Grace Blackwell Superchip, such as NVIDIA DGX Spark systems

N/A

N/A

FP16

NVIDIA-GB200

NVIDIA GB200

1.32

1.32

FP16

NVIDIA-H100-80GB-HBM3

NVIDIA H100 80GB HBM3

1.15

1.15

FP16

NVIDIA-H100-NVL

NVIDIA H100 NVL

1.15

1.15

FP16

NVIDIA-H200

NVIDIA H200

1.15

1.15

FP16

NVIDIA-L4

NVIDIA L4

0.82

0.82

FP16

NVIDIA-L40S

NVIDIA L40S

1.06

1.06

FP16

NVIDIA-RTX-PRO-6000-Blackwell-Server-Edition

NVIDIA RTX PRO 6000 Blackwell Server Edition

1.19

1.19

FP16

Note

GB10 uses unified system memory. GPU memory usage is reported as N/A because the usual VRAM measurements are not accurate for this architecture. For more information, refer to Guidance for Reporting Memory Resources with Unified Memory Architecture.

Runtime Selection#

Starting in version 2.0.0, Object Detection NIM does not expose user-selectable model profiles. The runtime selects the supported FP16 pipeline internally for the loaded object detection model and detected GPU. You cannot override the selection by using NIM_MODEL_PROFILE, and the list-model-profiles command is not available.

For details, refer to Optimization for NVIDIA NIM for Object Detection.

Software#

NVIDIA Driver#

Ensure that the latest compatible NVIDIA driver is installed on your system before launching NIM containers. If you experience issues starting the containers, verify that your driver is up to date.

NVIDIA Container Toolkit#

Your Docker environment must support NVIDIA GPUs. For more information, refer to NVIDIA Container Toolkit.