Unified Virtual Memory (UVM)#

Unified Virtual Memory (UVM) provides a single, cohesive memory address space accessible by both CPUs and GPUs within a system. This feature creates a managed memory pool, allowing data to be allocated and accessed by code executing on either processor. The primary benefit is simplified programming and enhanced performance for GPU-accelerated workloads, as it eliminates the need for applications to explicitly manage data transfers between CPU and GPU memory. For additional information about this feature, refer to the Unified Virtual Memory documentation.

Unified Virtual Memory (UVM)

UVM Known Issues and Limitations#

  • Unified Virtual Memory (UVM) is restricted to 1:1 time-sliced and MIG vGPU for Compute profiles that allocate the entire framebuffer of a compatible physical GPU or GPU Instance. Fractional time-sliced vGPUs do not support UVM.

  • UVM is only supported on Linux Guest OS distros. Windows Guest OS is not supported. For broader Windows guest limits, refer to Microsoft Windows Server limitations.

  • UVM is disabled by default and must be explicitly enabled for each vGPU that requires it by setting a specific vGPU plugin parameter for the VM.

  • On NVIDIA A100 multi-vGPU VMs, enabling UVM can disable NVLink P2P. Refer to A100 UVM P2P constraint.

Caution

Enabling UVM disables vGPU migration for the VM, which may reduce operational flexibility in environments reliant on live migration.

Tip

When deploying NVIDIA NIM, if UVM is enabled and an optimized engine is available, the model will run on the TensorRT-LLM (TRT-LLM) backend. Otherwise, it will typically run on the vLLM backend.

Hypervisor Platform Support for UVM#

Note

Unified Virtual Memory (UVM) is disabled by default. If used, you must enable unified memory individually for each vGPU for Compute VM that requires it by setting a vGPU plugin parameter. How to enable UVM for a vGPU VM depends on the hypervisor that you are using.

vGPU Support for UVM#

UVM is supported on 1:1 MIG-backed and time sliced vGPUs. These vGPUs have the entire framebuffer of a MIG GPU Instance or physical GPU assigned to a single vGPU.

Table 151 vGPU Support for UVM on the NVIDIA Blackwell GPU Architecture#

Board

vGPU

NVIDIA HGX B200 SXM

  • B200X-7-180C

  • MIG-backed 1:1 vGPUs

NVIDIA RTX PRO 6000 Blackwell SE

  • DC-4-96C

  • MIG-backed 1:1 vGPUs

Table 152 vGPU Support for UVM on the NVIDIA Hopper GPU Architecture#

Board

vGPU

NVIDIA H800 PCIe 94GB (H800 NVL)

  • H800L-94C

  • All MIG-backed vGPUs

NVIDIA H800 PCIe 80GB

  • H800-80C

  • All MIG-backed vGPUs

NVIDIA H800 SXM5 80GB

  • H800XM-80C

  • All MIG-backed vGPUs

NVIDIA H200 SXM5

  • H200X-141C

  • All MIG-backed vGPUs

NVIDIA H200 NVL

  • H200-141C

  • All MIG-backed vGPUs

NVIDIA H100 PCIe 94GB (H100 NVL)

  • H100L-94C

  • All MIG-backed vGPUs

NVIDIA H100 SXM5 94GB

  • H100XL-94C

  • All MIG-backed vGPUs

NVIDIA H100 PCIe 80GB

  • H100-80C

  • All MIG-backed vGPUs

NVIDIA H100 SXM5 80GB

  • H100XM-80C

  • All MIG-backed vGPUs

NVIDIA H100 SXM5 64GB

  • H100XS-64C

  • All MIG-backed vGPUs

NVIDIA H20 SXM5 141GB

  • H20X-141C

  • All MIG-backed vGPUs

NVIDIA H20 SXM5 96GB

  • H20-96C

  • All MIG-backed vGPUs

Table 153 vGPU Support for UVM on the NVIDIA Ada Lovelace GPU Architecture#

Board

vGPU

NVIDIA L40

L40-48C

NVIDIA L40S

L40S-48C

  • NVIDIA L20

  • NVIDIA L20 liquid-cooled

L20-48C

NVIDIA L4

L4-24C

NVIDIA L2

L2-24C

NVIDIA RTX 6000 Ada

RTX 6000 Ada-48C

NVIDIA RTX 5880 Ada

RTX 5880 Ada-48C

NVIDIA RTX 5000 Ada

RTX 6000 Ada-32C

Table 154 vGPU Support for UVM on the NVIDIA Ampere GPU Architecture#

Board

vGPU

  • NVIDIA A800 PCIe 80GB

  • NVIDIA A800 PCIe 80GB liquid-cooled

  • NVIDIA AX800

  • A800D-80C

  • All MIG-backed vGPUs

NVIDIA A800 PCIe 40GB active-cooled

  • A800-40C

  • All MIG-backed vGPUs

NVIDIA A800 HGX 80GB

  • A800DX-80C

  • All MIG-backed vGPUs

  • NVIDIA A100 PCIe 80GB

  • NVIDIA A100 PCIe 80GB liquid-cooled

  • NVIDIA A100X

  • A100D-80C

  • All MIG-backed vGPUs

NVIDIA A100 HGX 80GB

  • A100DX-80C

  • All MIG-backed vGPUs

NVIDIA A100 PCIe 40GB

  • A100-40C

  • All MIG-backed vGPUs

NVIDIA A100 HGX 40GB

  • A100X-40C

  • All MIG-backed vGPUs

NVIDIA A40

A40-48C

  • NVIDIA A30

  • NVIDIA A30X

  • NVIDIA A30 liquid-cooled

  • A30-24C

  • All MIG-backed vGPUs

NVIDIA A16

A16-16C

NVIDIA A10

A10-24C

NVIDIA RTX A6000

A6000-48C

NVIDIA RTX A5500

A5500-24C

NVIDIA RTX A5000

A5000-24C