> This page is for Health Validation.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/health-validation/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/health-validation/_mcp/server.

# Xid Checks

Xid covers NVRM Xid fault events logged by the NVIDIA driver to the kernel ring buffer. Xid measurements are collected separately from GPU telemetry and route only to Xid checks.

## Single-rule checks reading `xid_counts`

These 19 checks share the properties below and each has one rule.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr> |

| Check | Name | Condition | Severity | Recommended action |
|---|---|---|---|---|
| <nobr>`gpu_xid_46`</nobr> | GPU Xid 46 (GPU stopped processing) | <nobr>`46 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_48`</nobr> | GPU Xid 48 (double bit ECC error) | <nobr>`48 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_62`</nobr> | GPU Xid 62 (internal micro-controller halt) | <nobr>`62 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_64`</nobr> | GPU Xid 64 (GPU memory remapping failure) | <nobr>`64 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_79`</nobr> | GPU Xid 79 (GPU has fallen off the bus) | <nobr>`79 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_95`</nobr> | GPU Xid 95 (uncontained memory error) | <nobr>`95 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_109`</nobr> | GPU Xid 109 (context switch timeout error) | <nobr>`109 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_110`</nobr> | GPU Xid 110 (security fault error) | <nobr>`110 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_119`</nobr> | GPU Xid 119 (GSP RPC timeout) | <nobr>`119 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_120`</nobr> | GPU Xid 120 (GSP error) | <nobr>`120 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_140`</nobr> | GPU Xid 140 (ECC unrecovered error) | <nobr>`140 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_142`</nobr> | GPU Xid 142 (NVENC3 error) | <nobr>`142 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_143`</nobr> | GPU Xid 143 (GPU initialization error) | <nobr>`143 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_151`</nobr> | GPU Xid 151 (key rotation error) | <nobr>`151 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`restart_os`</nobr> |
| <nobr>`gpu_xid_154`</nobr> | GPU Xid 154 (GPU recovery action changed) | <nobr>`154 in xid_counts`</nobr> | INFO | <nobr>`drain`</nobr>: <nobr>`xid_154`</nobr> |
| <nobr>`gpu_xid_155`</nobr> | GPU Xid 155 (NVLink SW-defined error) | <nobr>`155 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_156`</nobr> | GPU Xid 156 (resource retirement event) | <nobr>`156 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_158`</nobr> | GPU Xid 158 (GPU fatal timeout) | <nobr>`158 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_159`</nobr> | GPU Xid 159 (CHI non-data error) | <nobr>`159 in xid_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |

## Single-rule checks reading `sub_xid_code_counts`

These 20 checks share the properties below and each has one rule.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`sub_xid_code_counts`</nobr> |

| Check | Name | Condition | Severity | Recommended action |
|---|---|---|---|---|
| <nobr>`gpu_xid_149_1`</nobr> | GPU Xid 149.1 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.1" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_2`</nobr> | GPU Xid 149.2 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.2" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_2a`</nobr> | GPU Xid 149.2A (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.2A" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_2b`</nobr> | GPU Xid 149.2B (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.2B" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_4`</nobr> | GPU Xid 149.4 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.4" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_10`</nobr> | GPU Xid 149.10 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.10" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_11`</nobr> | GPU Xid 149.11 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.11" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_12`</nobr> | GPU Xid 149.12 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.12" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_15`</nobr> | GPU Xid 149.15 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.15" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_16`</nobr> | GPU Xid 149.16 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.16" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_20`</nobr> | GPU Xid 149.20 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.20" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_21`</nobr> | GPU Xid 149.21 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.21" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_22`</nobr> | GPU Xid 149.22 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.22" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_23`</nobr> | GPU Xid 149.23 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.23" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_24`</nobr> | GPU Xid 149.24 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.24" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_25`</nobr> | GPU Xid 149.25 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.25" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_26`</nobr> | GPU Xid 149.26 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.26" in sub_xid_code_counts`</nobr> | WARNING | none |
| <nobr>`gpu_xid_149_28`</nobr> | GPU Xid 149.28 (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.28" in sub_xid_code_counts`</nobr> | WARNING | none |
| <nobr>`gpu_xid_149_a`</nobr> | GPU Xid 149.A (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.A" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |
| <nobr>`gpu_xid_149_f`</nobr> | GPU Xid 149.F (NETIR_LINK_EVT/NETIR_LINK_DOWN) | <nobr>`"149.F" in sub_xid_code_counts`</nobr> | CRITICAL | <nobr>`drain`</nobr>: <nobr>`reset_gpu`</nobr> → <nobr>`service_required`</nobr> |

## GPU Xid 144 (NVLink SAW error)

`gpu_xid_144`

Detects NVRM Xid 144 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `144 in fatal_xid_counts`
  - Message: GPU Xid 144 detected: NVLink SAW error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `144 in xid_counts && xid_counts[144] > (144 in fatal_xid_counts ? fatal_xid_counts[144] : 0)`
  - Message: GPU Xid 144 detected: NVLink SAW error (Nonfatal).

## GPU Xid 145 (NVLink RLW error)

`gpu_xid_145`

Detects NVRM Xid 145 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `145 in fatal_xid_counts`
  - Message: GPU Xid 145 detected: NVLink RLW error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `145 in xid_counts && xid_counts[145] > (145 in fatal_xid_counts ? fatal_xid_counts[145] : 0)`
  - Message: GPU Xid 145 detected: NVLink RLW error (Nonfatal).

## GPU Xid 146 (NVLink TLW error)

`gpu_xid_146`

Detects NVRM Xid 146 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `146 in fatal_xid_counts`
  - Message: GPU Xid 146 detected: NVLink TLW error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `146 in xid_counts && xid_counts[146] > (146 in fatal_xid_counts ? fatal_xid_counts[146] : 0)`
  - Message: GPU Xid 146 detected: NVLink TLW error (Nonfatal).

## GPU Xid 147 (NVLink TREX error)

`gpu_xid_147`

Detects NVRM Xid 147 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `147 in fatal_xid_counts`
  - Message: GPU Xid 147 detected: NVLink TREX error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `147 in xid_counts && xid_counts[147] > (147 in fatal_xid_counts ? fatal_xid_counts[147] : 0)`
  - Message: GPU Xid 147 detected: NVLink TREX error (Nonfatal).

## GPU Xid 148 (NVLink NVLPW_CTRL error)

`gpu_xid_148`

Detects NVRM Xid 148 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `148 in fatal_xid_counts`
  - Message: GPU Xid 148 detected: NVLink NVLPW_CTRL error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `148 in xid_counts && xid_counts[148] > (148 in fatal_xid_counts ? fatal_xid_counts[148] : 0)`
  - Message: GPU Xid 148 detected: NVLink NVLPW_CTRL error (Nonfatal).

## GPU Xid 149 (NVLink NETIR error)

`gpu_xid_149`

Detects NVRM Xid 149 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `149 in fatal_xid_counts`
  - Message: GPU Xid 149 detected: NVLink NETIR error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `149 in xid_counts && xid_counts[149] > (149 in fatal_xid_counts ? fatal_xid_counts[149] : 0)`
  - Message: GPU Xid 149 detected: NVLink NETIR error (Nonfatal).

## GPU Xid 150 (NVLink MSE error)

`gpu_xid_150`

Detects NVRM Xid 150 in the system journal since boot, differentiated by the Fatal/Nonfatal severity NVRM assigns each occurrence.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr>, <nobr>`fatal_xid_counts`</nobr> |

**Rules**

- **CRITICAL** when `150 in fatal_xid_counts`
  - Message: GPU Xid 150 detected: NVLink MSE error (Fatal).
  - Action: operational state `drain`, then `reset_gpu` → `service_required`
- **WARNING** when `150 in xid_counts && xid_counts[150] > (150 in fatal_xid_counts ? fatal_xid_counts[150] : 0)`
  - Message: GPU Xid 150 detected: NVLink MSE error (Nonfatal).

## GPU Xid Warnings

`gpu_xid_warning`

Detects application-level and otherwise recoverable NVRM Xid errors in the system journal since boot, grouped by their NVIDIA Xid-catalog immediate-action resolution bucket.

| Property | Value |
|---|---|
| Provenance | <nobr>`nvidia`</nobr> |
| Condition scope | <nobr>`NODE`</nobr> |
| Execution class | <nobr>`PASSIVE`</nobr> |
| Scheduling cadence | <nobr>`CONTINUOUS`</nobr> |
| Maturity | <nobr>`stable`</nobr> |
| Execution contexts | <nobr>`OPERATOR_INFRA`</nobr>, <nobr>`TENANT_GUEST`</nobr> |
| Architectures | <nobr>`GB200`</nobr>, <nobr>`GB300`</nobr> |
| Required signals | <nobr>`xid_counts`</nobr> |

**Rules**

- **WARNING** when `xid_counts.exists(k, k in [8, 11, 13, 25, 31, 32, 39, 40, 41, 60, 68, 69, 70, 71, 72, 75, 76, 77, 82, 83, 84, 85, 86, 88, 89, 94, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 139])`
  - Message: Application-level GPU Xid error detected. Restart the faulting application; send the node for service if the error persists.
  - Action: operational state `drain`, then `restart_app` → `service_required`
- **WARNING** when `xid_counts.exists(k, k in [45])`
  - Message: GPU Xid 45 (preemptive removal) detected. Follow the Xid 45 workflow.
  - Action: operational state `drain`, then `workflow_xid_45`
- **WARNING** when `xid_counts.exists(k, k in [78])`
  - Message: GPU Xid 78 (vGPU start error) detected. Update GPU software/firmware.
  - Action: operational state `drain`, then `update_swfw`

**Recommended action:** operational state `drain`, then `restart_app` → `service_required`