diff --git a/src/compiler/execution.jl b/src/compiler/execution.jl
index 663572140..239ebe0bd 100644
--- a/src/compiler/execution.jl
+++ b/src/compiler/execution.jl
@@ -265,15 +265,22 @@ end
 
 @autoreleasepool function (kernel::HostKernel)(args...; groups=1, threads=1,
                                                queue=global_queue(device()))
-    threadgroupsPerGrid = MTLSize(groups)
-    threadsPerThreadgroup = MTLSize(threads)
-    (threadgroupsPerGrid.width>0 && threadgroupsPerGrid.height>0 && threadgroupsPerGrid.depth>0) ||
+    gs = MTLSize(groups)
+    ts = MTLSize(threads)
+    (gs.width>0 && gs.height>0 && gs.depth>0) ||
         throw(ArgumentError("All group dimensions should be non-zero"))
-    (threadsPerThreadgroup.width>0 && threadsPerThreadgroup.height>0 && threadsPerThreadgroup.depth>0) ||
+    (ts.width>0 && ts.height>0 && ts.depth>0) ||
         throw(ArgumentError("All thread dimensions should be non-zero"))
 
-    (threadsPerThreadgroup.width * threadsPerThreadgroup.height * threadsPerThreadgroup.depth) > kernel.pipeline.maxTotalThreadsPerThreadgroup &&
-        throw(ArgumentError("Number of threads in group ($(threadsPerThreadgroup.width * threadsPerThreadgroup.height * threadsPerThreadgroup.depth)) should not exceed $(kernel.pipeline.maxTotalThreadsPerThreadgroup)"))
+    (ts.width * ts.height * ts.depth) > kernel.pipeline.maxTotalThreadsPerThreadgroup &&
+        throw(ArgumentError("Number of threads in group ($(ts.width * ts.height * ts.depth)) should not exceed $(kernel.pipeline.maxTotalThreadsPerThreadgroup)"))
+
+    (gs.width * ts.width) > typemax(UInt32) &&
+        throw(ArgumentError("Total threads per grid in a dimension (threads.width($(gs.width)) * groups.width($(ts.width)) = $(gs.width * ts.width)) must not exceed $(typemax(UInt32))"))
+    (gs.height * ts.height) > typemax(UInt32) &&
+        throw(ArgumentError("Total threads per grid in a dimension (threads.height($(gs.height)) * groups.height($(ts.height)) = $(gs.height * ts.height)) must not exceed $(typemax(UInt32))"))
+    (gs.depth * ts.depth) > typemax(UInt32) &&
+        throw(ArgumentError("Total threads per grid in a dimension (threads.depth($(gs.depth)) * groups.depth($(ts.depth)) = $(gs.depth * ts.depth)) must not exceed $(typemax(UInt32))"))
 
     kernel_state = KernelState(Random.rand(UInt32))
 
@@ -283,7 +290,7 @@ end
     argument_buffers = try
         MTL.set_function!(cce, kernel.pipeline)
         bufs = encode_arguments!(cce, kernel, kernel_state, kernel.f, args...)
-        MTL.append_current_function!(cce, threadgroupsPerGrid, threadsPerThreadgroup)
+        MTL.append_current_function!(cce, gs, ts)
         bufs
     finally
         close(cce)
diff --git a/test/execution.jl b/test/execution.jl
index 1124f5abc..2749de80c 100644
--- a/test/execution.jl
+++ b/test/execution.jl
@@ -164,6 +164,10 @@ end
     @test_throws InexactError @metal groups=(-2) tester(bufferA)
     @test_throws ArgumentError @metal threads=(1025) tester(bufferA)
     @test_throws ArgumentError @metal threads=(1000,2) tester(bufferA)
+
+    @test_throws ArgumentError @metal threads=(1024,1,1) groups=(4194304,1,1) tester(bufferA)
+    @test_throws ArgumentError @metal threads=(1,1024,1) groups=(1,4194304,1) tester(bufferA)
+    @test_throws ArgumentError @metal threads=(1,1,1024) groups=(1,1,4194304) tester(bufferA)
 end
 
 ############################################################################################