diff --git a/.pick_status.json b/.pick_status.json
index e1d7f7a4239..54cd1033b41 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -924,7 +924,7 @@
         "description": "rusticl/kernel: properly respect device thread limits per dimension",
         "nominated": true,
         "nomination_type": 0,
-        "resolution": 0,
+        "resolution": 1,
         "main_sha": null,
         "because_sha": null,
         "notes": null
diff --git a/src/gallium/frontends/rusticl/core/kernel.rs b/src/gallium/frontends/rusticl/core/kernel.rs
index 1ddaee62dc5..5e50a63af77 100644
--- a/src/gallium/frontends/rusticl/core/kernel.rs
+++ b/src/gallium/frontends/rusticl/core/kernel.rs
@@ -898,7 +898,7 @@ impl Kernel {
         let total_threads = block.iter().take(work_dim).product::<usize>();
         if threads != 1 && total_threads < subgroups {
             for i in 0..work_dim {
-                if grid[i] * total_threads < threads {
+                if grid[i] * total_threads < threads && grid[i] * block[i] <= dim_threads[i] {
                     block[i] *= grid[i];
                     grid[i] = 1;
                     // can only do it once as nothing is cleanly divisible