diff --git a/src/video_core/amdgpu/liverpool.h b/src/video_core/amdgpu/liverpool.h
index 837b73d89..070253ca9 100644
--- a/src/video_core/amdgpu/liverpool.h
+++ b/src/video_core/amdgpu/liverpool.h
@@ -899,7 +899,8 @@ struct Liverpool {
             // There is a small difference between T# and CB number types, account for it.
             return RemapNumberFormat(info.number_type == NumberFormat::SnormNz
                                          ? NumberFormat::Srgb
-                                         : info.number_type.Value());
+                                         : info.number_type.Value(),
+                                     info.format);
         }
 
         [[nodiscard]] NumberConversion GetNumberConversion() const {
diff --git a/src/video_core/amdgpu/resource.h b/src/video_core/amdgpu/resource.h
index 3dc1eadde..467cecf10 100644
--- a/src/video_core/amdgpu/resource.h
+++ b/src/video_core/amdgpu/resource.h
@@ -54,7 +54,7 @@ struct Buffer {
     }
 
     NumberFormat GetNumberFmt() const noexcept {
-        return RemapNumberFormat(NumberFormat(num_format));
+        return RemapNumberFormat(NumberFormat(num_format), DataFormat(data_format));
     }
 
     DataFormat GetDataFmt() const noexcept {
@@ -264,7 +264,7 @@ struct Image {
     }
 
     NumberFormat GetNumberFmt() const noexcept {
-        return RemapNumberFormat(NumberFormat(num_format));
+        return RemapNumberFormat(NumberFormat(num_format), DataFormat(data_format));
     }
 
     NumberConversion GetNumberConversion() const noexcept {
diff --git a/src/video_core/amdgpu/types.h b/src/video_core/amdgpu/types.h
index a19e53256..57f97418a 100644
--- a/src/video_core/amdgpu/types.h
+++ b/src/video_core/amdgpu/types.h
@@ -252,7 +252,7 @@ inline DataFormat RemapDataFormat(const DataFormat format) {
     }
 }
 
-inline NumberFormat RemapNumberFormat(const NumberFormat format) {
+inline NumberFormat RemapNumberFormat(const NumberFormat format, const DataFormat data_format) {
     switch (format) {
     case NumberFormat::Uscaled:
         return NumberFormat::Uint;
@@ -260,6 +260,14 @@ inline NumberFormat RemapNumberFormat(const NumberFormat format) {
         return NumberFormat::Sint;
     case NumberFormat::Ubnorm:
         return NumberFormat::Unorm;
+    case NumberFormat::Float:
+        if (data_format == DataFormat::Format8) {
+            // Games may ask for 8-bit float when they want to access the stencil component
+            // of a depth-stencil image. Change to unsigned int to match the stencil format.
+            // This is also the closest approximation to pass the bits through unconverted.
+            return NumberFormat::Uint;
+        }
+        [[fallthrough]];
     default:
         return format;
     }