diff --git a/example/ck_tile/18_flatmm/flatmm_basic.cpp b/example/ck_tile/18_flatmm/flatmm_basic.cpp
index 224631db0a..05d0c73b7e 100644
--- a/example/ck_tile/18_flatmm/flatmm_basic.cpp
+++ b/example/ck_tile/18_flatmm/flatmm_basic.cpp
@@ -20,10 +20,10 @@ float flatmm_calc(const ck_tile::FlatmmHostArgs& args, const ck_tile::stream_con
     constexpr bool kPadN = false;
     constexpr bool kPadK = false;
 
-    constexpr int kBlockPerCu = 1;
+    constexpr int kBlockPerCu = 2;
 
     // This part comes from the Codegen
-    constexpr ck_tile::index_t M_Tile = 32;
+    constexpr ck_tile::index_t M_Tile = 128;
     constexpr ck_tile::index_t N_Tile = 128;
     constexpr ck_tile::index_t K_Tile = 64;
 
diff --git a/example/ck_tile/18_flatmm/run_flatmm_example.inc b/example/ck_tile/18_flatmm/run_flatmm_example.inc
index ebcdf58df6..864d888074 100644
--- a/example/ck_tile/18_flatmm/run_flatmm_example.inc
+++ b/example/ck_tile/18_flatmm/run_flatmm_example.inc
@@ -143,8 +143,8 @@ int run_flatmm_example_with_layouts(int argc,
         ck_tile::host_tensor_descriptor(M, N, stride_C, is_row_major(CLayout{})));
 
     // TODO: add different init types
-    ck_tile::FillUniformDistribution<ADataType>{-5.f, 5.f}(a_host);
-    ck_tile::FillUniformDistribution<BDataType>{-5.f, 5.f}(b_origin_host);
+    ck_tile::FillUniformDistribution<ADataType>{-.5f, .5f}(a_host);
+    ck_tile::FillUniformDistribution<BDataType>{-.5f, .5f}(b_origin_host);
 
     ck_tile::DeviceMem a_dev_buf(a_host.get_element_space_size_in_bytes());
     ck_tile::DeviceMem c_dev_buf(c_rslt_host.get_element_space_size_in_bytes());