> ## Documentation Index
> Fetch the complete documentation index at: https://docs.polystack.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Multi-Vendor Accelerator Support

> Run NVIDIA GPUs, Intel Gaudi HPUs, and AMD GPUs on one platform, each with a dedicated flavor, cluster template, and vendor-specific operator.

## Overview

The On-Prem AI Infrastructure Platform is vendor-neutral. **NVIDIA** GPUs, **Intel Gaudi**
HPUs, and **AMD** GPUs run side by side on the same platform. Each vendor gets its own
compute flavor, Kubernetes cluster template, and accelerator operator, so every workload runs
on the accelerator that fits it best.

<Note>
  Accelerators are presented through PCI passthrough at the [compute layer](/services/ai-platform/gpu-vms-and-containers),
  and a vendor-specific operator runs on each Kubernetes node group.
</Note>

***

## Vendor Model

| Layer | NVIDIA | Intel Gaudi | AMD |
| - | - | - | - |
| **Compute** | PCI passthrough, vGPU | PCI passthrough | PCI passthrough, SR-IOV |
| **Flavor** | Dedicated NVIDIA flavor | Dedicated Intel Gaudi flavor | Dedicated AMD flavor |
| **Cluster template** | Dedicated NVIDIA template | Dedicated Intel Gaudi template | Dedicated AMD template |
| **Kubernetes operator** | NVIDIA GPU Operator | Intel Gaudi Operator | AMD GPU Operator (ROCm) |

***

## Architecture

```mermaid theme={null}
graph TD
    subgraph NV[NVIDIA]
        NVF[NVIDIA Flavor] --> NVT[NVIDIA Cluster Template] --> NVO[NVIDIA GPU Operator]
    end
    subgraph IG[Intel Gaudi]
        IGF[Gaudi Flavor] --> IGT[Gaudi Cluster Template] --> IGO[Intel Gaudi Operator]
    end
    subgraph AM[AMD]
        AMF[AMD Flavor] --> AMT[AMD Cluster Template] --> AMO[AMD GPU Operator]
    end
    HOSTS[Accelerator Hosts] -->|PCI passthrough| NVF
    HOSTS -->|PCI passthrough| IGF
    HOSTS -->|PCI passthrough| AMF
```

***

## Vendors

<Tabs>
  <Tab title="NVIDIA" icon="microchip">
    NVIDIA GPUs are available as dedicated passthrough devices or as shared vGPUs. In
    Kubernetes, the NVIDIA GPU Operator manages drivers and runtime, MIG partitioning,
    time-slicing, and DCGM monitoring.
  </Tab>

  <Tab title="Intel Gaudi" icon="microchip">
    Intel Gaudi HPUs are available as dedicated passthrough devices. In Kubernetes, the Intel
    Gaudi Operator manages the Gaudi software stack on each node.
  </Tab>

  <Tab title="AMD" icon="microchip">
    AMD GPUs are available as dedicated passthrough devices or as shared SR-IOV virtual
    functions. In Kubernetes, the AMD GPU Operator manages the ROCm stack on each node.
  </Tab>
</Tabs>

***

## Benefits

<CardGroup cols={2}>
  <Card title="No vendor lock-in" icon="unlock" color="#bf9667">
    Add accelerators from any supported vendor without changing how workloads are deployed.
  </Card>

  <Card title="Right accelerator per workload" icon="bullseye" color="#bf9667">
    Match training, inference, and development workloads to the accelerator that suits them.
  </Card>

  <Card title="Clean separation" icon="table-columns" color="#bf9667">
    Separate flavors and templates keep each vendor's drivers and operators isolated to its
    own node groups.
  </Card>

  <Card title="One operating model" icon="window-maximize" color="#bf9667">
    Every vendor is managed through the same console, flavors, and cluster templates.
  </Card>
</CardGroup>

***

## Next Steps

<CardGroup cols={2}>
  <Card title="GPU and HPU Performance Optimization" icon="gauge-high" href="/services/ai-platform/performance-optimization" color="#bf9667">
    Tuning and telemetry for each vendor's accelerators.
  </Card>

  <Card title="Unified Console" icon="window-maximize" href="/services/ai-platform/unified-console" color="#bf9667">
    Manage every vendor's workloads from one portal.
  </Card>

  <Card title="GPU-Capable VMs and Containers" icon="server" href="/services/ai-platform/gpu-vms-and-containers" color="#bf9667">
    Passthrough, vGPU, and SR-IOV in detail.
  </Card>

  <Card title="Cluster Templates" icon="file-lines" href="/services/kubernetes/user-guide/cluster-templates" color="#bf9667">
    How cluster templates are defined in Polystack K8SaaS.
  </Card>
</CardGroup>