diff --git a/docs/features.txt b/docs/features.txt
index e285a368ed2..7041e7afa50 100644
--- a/docs/features.txt
+++ b/docs/features.txt
@@ -904,7 +904,7 @@ Rusticl OpenCL 2.0 -- all DONE:
   Extended 2D images creation                           in progress
   - CL_ABGR                                             DONE
   - cl_khr_image2d_from_buffer                          DONE (iris, llvmpipe)
-  - cl_khr_depth_images                                 not started
+  - cl_khr_depth_images                                 DONE
   - from sRGB images                                    in progress
   clCreateSamplerWithProperties                         DONE
   Non-uniform work-group sizes                          not started
diff --git a/docs/relnotes/new_features.txt b/docs/relnotes/new_features.txt
index e69de29bb2d..9a6b48db7c3 100644
--- a/docs/relnotes/new_features.txt
+++ b/docs/relnotes/new_features.txt
@@ -0,0 +1 @@
+cl_khr_depth_images in rusticl
\ No newline at end of file
diff --git a/src/gallium/frontends/rusticl/core/format.rs b/src/gallium/frontends/rusticl/core/format.rs
index 253e7c360ad..34783273878 100644
--- a/src/gallium/frontends/rusticl/core/format.rs
+++ b/src/gallium/frontends/rusticl/core/format.rs
@@ -183,9 +183,11 @@ cl_format_table!([
 // broken
 //  (CL_sRGBx,     CL_UNORM_INT8)         => pipe_format::PIPE_FORMAT_R8G8B8X8_SRGB,
 
-// broken
-//  (CL_DEPTH,     CL_FLOAT)              => pipe_format::PIPE_FORMAT_Z32_FLOAT,
-//  (CL_DEPTH,     CL_UNORM_INT16)        => pipe_format::PIPE_FORMAT_Z16_UNORM,
+    // yes, we use non depth formats for CL_DEPTH, because OpenCL requires normal pixel semantics
+    // and not implicit depth format semantics (e.g. implicit value clamping). Intel NEO does the
+    // same.
+    (CL_DEPTH,     CL_FLOAT)              => pipe_format::PIPE_FORMAT_R32_FLOAT,
+    (CL_DEPTH,     CL_UNORM_INT16)        => pipe_format::PIPE_FORMAT_R16_UNORM,
 
     (CL_LUMINANCE, CL_HALF_FLOAT)         => pipe_format::PIPE_FORMAT_L16_FLOAT,
     (CL_LUMINANCE, CL_FLOAT)              => pipe_format::PIPE_FORMAT_L32_FLOAT,