diff --git a/docs/release-notes/0.9.4.md b/docs/release-notes/0.9.4.md
index d89997b7..e187726b 100644
--- a/docs/release-notes/0.9.4.md
+++ b/docs/release-notes/0.9.4.md
@@ -1,7 +1,8 @@
-### 0.9.4 {small}`the future`
+### 0.9.4
 
 ```{rubric} Features
 ```
 * {func}`~rapids_singlecell.tl.leiden` now provides `random_state` parameter for initializing the optimization {pr}`102` {smaller}`J Pintar & S Dicks`
 * {func}`~rapids_singlecell.tl.louvain` now provides `threshold` parameter for setting the minimum modularity gain between levels {pr}`103` {smaller}`J Pintar & S Dicks`
 * added a kernel for sparse sum along the major axis {pr}`112` {smaller}`S Dicks`
+* {func}`~rapids_singlecell.pp.calculate_qc_metrics` now uses less atomics and global memory during kernels {pr}`113` {smaller}`S Dicks`
diff --git a/src/rapids_singlecell/preprocessing/_kernels/_qc_kernels.py b/src/rapids_singlecell/preprocessing/_kernels/_qc_kernels.py
index 36820521..18bbfdc8 100644
--- a/src/rapids_singlecell/preprocessing/_kernels/_qc_kernels.py
+++ b/src/rapids_singlecell/preprocessing/_kernels/_qc_kernels.py
@@ -12,14 +12,17 @@
         int start_idx = indptr[gene];
         int stop_idx = indptr[gene+1];
 
+        {0} sums_genes_i = 0;
+        int gene_ex_i = 0;
         for(int cell = start_idx; cell < stop_idx; cell++){
             {0} value = data[cell];
             int cell_number = index[cell];
-            atomicAdd(&sums_genes[gene], value);
+            sums_genes_i += value;
             atomicAdd(&sums_cells[cell_number], value);
-            atomicAdd(&gene_ex[gene], 1);
+            gene_ex_i += 1;
             atomicAdd(&cell_ex[cell_number], 1);
-
+        sums_genes[gene] = sums_genes_i;
+        gene_ex[gene] = gene_ex_i;
         }
     }
 """
@@ -36,14 +39,17 @@
         int start_idx = indptr[cell];
         int stop_idx = indptr[cell+1];
 
+        {0} sums_cells_i = 0;
+        int cell_ex_i = 0;
         for(int gene = start_idx; gene < stop_idx; gene++){
             {0} value = data[gene];
             int gene_number = index[gene];
             atomicAdd(&sums_genes[gene_number], value);
-            atomicAdd(&sums_cells[cell], value);
+            sums_cells_i += value;
             atomicAdd(&gene_ex[gene_number], 1);
-            atomicAdd(&cell_ex[cell], 1);
-
+            cell_ex_i += 1;
+        sums_cells[cell] = sums_cells_i;
+        cell_ex[cell] = cell_ex_i;
         }
     }
 """
@@ -58,8 +64,6 @@
         if(cell >= n_cells || gene >=n_genes){
             return;
         }
-
-
         long long int index = static_cast<long long int>(cell) * n_genes + gene;
         {0} value = data[index];
         if (value>0.0){
@@ -103,12 +107,14 @@
         int start_idx = indptr[cell];
         int stop_idx = indptr[cell+1];
 
+        {0} sums_cells_i = 0;
         for(int gene = start_idx; gene < stop_idx; gene++){
             int gene_number = index[gene];
             if (mask[gene_number]==true){
-                atomicAdd(&sums_cells[cell], data[gene]);
+                sums_cells_i += data[gene];
 
             }
+        sums_cells[cell] = sums_cells_i;
         }
     }
 """