Support Matrix#

These support matrices list the supported hardware and software requirements of Riva.

Riva 2.15.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.15.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per bi-lingual model

    • ~5500 Mb per megatron 500m model

    • ~8500 Mb per megatron 1b model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.15.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 6.0

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.15.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.15.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.14.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.14.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per bi-lingual model

    • ~5500 Mb per megatron 500m model

    • ~8500 Mb per megatron 1b model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.14.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.1.1

  • JetPack 5.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.14.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.14.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.13.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.13.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.13.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.1.1

  • JetPack 5.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.13.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.13.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.12.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.12.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.12.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.1.1

  • JetPack 5.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.12.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.12.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.11.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.11.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.11.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.1.1

  • JetPack 5.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.11.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.11.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.10.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.10.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.10.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • NMT Models

    • ~5000 MB

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~9500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.10.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.10.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.9.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.9.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • NMT Models

    • ~4500 Mb per model

  • Total

    • 20500 MB

Embedded#

The following table shows the supported hardware for Riva 2.9.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0.2

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.9.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.9.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.8.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.8.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.8.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0.2

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.8.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.8.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.7.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.7.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.7.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0.2

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.7.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.7.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.6.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.6.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.6.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0.2

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.6.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.6.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.5.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.5.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.5.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0.2

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.5.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.5.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.4.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.4.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.4.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0 Developer Preview

  • JetPack 5.0.1 Developer Preview

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.4.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.4.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.3.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.3.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.3.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0 Developer Preview

  • JetPack 5.0.1 Developer Preview

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.3.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.3.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.2.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.2.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.2.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 5.0 Developer Preview

  • JetPack 5.0.1 Developer Preview

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.2.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.2.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.1.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.1.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.1.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 4.6.1

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.1.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.1.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 2.0.0#

Server Hardware#

Data Center#

The following table shows the supported hardware for Riva 2.0.0 on data center platforms.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Embedded#

The following table shows the supported hardware for Riva 2.0.0 on embedded platforms.

Hardware Compatibility

Operating System

Riva server requires Linux AArch64.

GPU Model

Deployment Platforms:

Jetson SDK version

  • JetPack 4.6

Mic, Camera, and Headset

Microphone:

  • Linux AArch64 with a USB microphone (for example, a Logitech H390 USB computer headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

RAM requirement

  • ASR Models

    • ~2300 MB

  • NLP Models

    • ~1800 MB (with punctuation)

  • TTS Models

    • ~2100 MB

  • All Models combined

    • ~4500 MB

Server Software#

Data Center#

The following table shows the supported software for Riva 2.0.0 on data center platforms.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Embedded#

The following table shows the supported software for Riva 2.0.0 on embedded platforms.

Software Compatibility

Container

Docker

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.10.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.10.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.10.0 Beta.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.9.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.9.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any NVIDIA Volta or later GPU (NVIDIA Turing and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.9.0 Beta.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

TAO Toolkit

NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.8.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.8.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500 MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.8.0 Beta.

Software Compatibility

Container

MIG (Multi Instance GPU)

  • On A100 and A30 platforms, MIG is supported provided enough vRam is available for the selected models.

Note

For further information on Ampere and MIG, refer to Ampere Architecture In-Depth:.

Docker

Note

For DGX users, refer to Preparing to use NVIDIA Containers.

NVIDIA TAO Toolkit

NVIDIA NeMo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.7.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.7.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.7.0 Beta.

Software Compatibility

Container

MiG (Multi Instance GPU)

  • On A100 and A30 platforms MiG is supported provided enough vRam is available for the selected Models.

Note

For further information on Ampere and MiG, see Ampere Architecture In-Depth:.

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

TAO

Nemo

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.6.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.6.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.6.0 Beta.

Software Compatibility

Container

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.5.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.5.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.5.0 Beta.

Software Compatibility

Container

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Riva 1.4.0 Beta#

Server Hardware#

The following table shows the supported hardware for Riva 1.4.0 Beta.

Hardware Compatibility

Operating System

Riva server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Riva is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Server Software#

The following table shows the supported software for Riva 1.4.0 Beta.

Software Compatibility

Container

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Skills Clients#

Riva Speech AI Skills clients do not require a local GPU and have minimal hardware requirements. Refer to Clients in a New Programming Language for creating client bindings in your programming language. The Python section describes the prebuilt Python bindings included in the Riva Quick Start package, which are also based on gRPC.

Jarvis 1.3.0 Beta

Hardware#

The following table shows the supported hardware for Jarvis 1.3.0 Beta.

Hardware Compatibility

Operating System

Jarvis server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Jarvis is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy. 16+ GB VRAM is recommended.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Software#

The following table shows the supported software for Jarvis 1.3.0 Beta.

Software Compatibility

Container

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Jarvis 1.2.x Beta

Hardware#

The following table shows the supported hardware for Jarvis 1.2.0 Beta.

Hardware Compatibility

Operating System

Jarvis server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Jarvis is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Software#

The following table shows the supported software for Jarvis 1.2.0 Beta.

Software Compatibility

Container

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Jarvis 1.1.0 Beta

Hardware#

The following table shows the supported hardware for Jarvis 1.1.0 Beta.

Hardware Compatibility

Operating System

Jarvis server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Jarvis is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Software#

The following table shows the supported software for Jarvis 1.1.0 Beta.

Software Compatibility

CUDA

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.

Jarvis 1.0.x Beta

Hardware#

The following table shows the supported hardware for Jarvis 1.0.0 Beta.

Hardware Compatibility

Operating System

Jarvis server requires Linux x86_64.

GPU Model

Preferred Deployment Platforms:

Note

Jarvis is supported on any Volta or later NVIDIA GPU (Volta, Turing, and NVIDIA Ampere GPU architecture) for development purposes. Care must be taken to not exceed the memory available when selecting models to deploy.

Mic, Camera, and Headset

Microphone:

  • Linux x86 with USB microphone (for example, a Logitech H390 USB Computer Headset)

Headphones:

  • Logitech H340

  • Logitech H390

  • Microsoft LX3000

GPU Memory

  • ASR Models

    • Streaming Models: ~5600 MB

    • Non-streaming models: ~3100 MB

  • NLP Models

    • ~500MB per BERT model

  • TTS Models

    • ~3500 MB

  • Total

    • 16000 MB

Software#

The following table shows the supported software for Jarvis 1.0.0 Beta.

Software Compatibility

CUDA

Docker

Note

For DGX users, see Preparing to use NVIDIA Containers.

Helm

NVIDIA Driver

Note

For earlier driver versions, refer to the NVIDIA Driver section in Deep Learning Frameworks Support Matrix.