diff --git a/bitsandbytes/cuda_specs.py b/bitsandbytes/cuda_specs.py
index f9fdd295d..3db42d536 100644
--- a/bitsandbytes/cuda_specs.py
+++ b/bitsandbytes/cuda_specs.py
@@ -1,6 +1,7 @@
 import dataclasses
 from functools import lru_cache
 import logging
+import os
 import platform
 import re
 import subprocess
@@ -97,6 +98,14 @@ def get_rocm_gpu_arch() -> str:
             if match:
                 return "gfx" + match.group(1)
             else:
+                override = os.environ.get("BNB_ROCM_GPU_ARCH")
+                if override:
+                    override = override.strip()
+                    if override:
+                        if not override.startswith("gfx"):
+                            override = f"gfx{override}"
+                        logger.info("Using ROCm GPU arch override: %s", override)
+                        return override
                 return "unknown"
         else:
             return "unknown"
@@ -108,6 +117,14 @@ def get_rocm_gpu_arch() -> str:
 ROCm GPU architecture detection failed despite ROCm being available.
                 """,
             )
+        override = os.environ.get("BNB_ROCM_GPU_ARCH")
+        if override:
+            override = override.strip()
+            if override:
+                if not override.startswith("gfx"):
+                    override = f"gfx{override}"
+                logger.info("Using ROCm GPU arch override: %s", override)
+                return override
         return "unknown"
 
 
diff --git a/docs/source/installation.mdx b/docs/source/installation.mdx
index c97996b75..61c143b37 100644
--- a/docs/source/installation.mdx
+++ b/docs/source/installation.mdx
@@ -189,6 +189,9 @@ pip install -e .
 * All features are supported for both consumer RDNA devices and Data Center CDNA products.
 * A compatible PyTorch version with AMD ROCm support is required. It is recommended to use the latest stable release. See [PyTorch on ROCm](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/3rd-party/pytorch-install.html) for guidance.
 
+> [!TIP]
+> If `rocminfo` is unavailable or fails in your runtime environment, set `BNB_ROCM_GPU_ARCH` (for example `BNB_ROCM_GPU_ARCH=gfx90a`) as a fallback.
+
 ### Installation from PyPI[[rocm-pip]]
 
 This is the most straightforward and recommended installation option.