diff --git a/source/op/descrpt_se_a_multi_device.cc b/source/op/descrpt_se_a_multi_device.cc
index 40f2c92eb0..af18909884 100644
--- a/source/op/descrpt_se_a_multi_device.cc
+++ b/source/op/descrpt_se_a_multi_device.cc
@@ -261,7 +261,7 @@ class DescrptSeAOp : public OpKernel {
             cudaErrcheck(cudaMemcpy(jlist,  jlist_host,  sizeof(int) * mesh_host[3], cudaMemcpyHostToDevice));
             
             max_nbor_size = 0;
-            for(int ii = 0; ii < mesh_host[2]; ii++) {
+            for(int ii = 0; ii < mesh_host[2] - 1; ii++) {
                 max_nbor_size = (jrange_host[ii + 1] - jrange_host[ii]) > max_nbor_size ? (jrange_host[ii + 1] - jrange_host[ii]) : max_nbor_size;
             }
             assert(max_nbor_size <= GPU_MAX_NBOR_SIZE);
diff --git a/source/op/descrpt_se_r_multi_device.cc b/source/op/descrpt_se_r_multi_device.cc
index 81d2603c79..a1f50fbb95 100644
--- a/source/op/descrpt_se_r_multi_device.cc
+++ b/source/op/descrpt_se_r_multi_device.cc
@@ -250,7 +250,7 @@ class DescrptSeROp : public OpKernel {
             cudaErrcheck(cudaMemcpy(jlist,  jlist_host,  sizeof(int) * mesh_host[3], cudaMemcpyHostToDevice));
 
             max_nbor_size = 0;
-            for(int ii = 0; ii < mesh_host[2]; ii++) {
+            for(int ii = 0; ii < mesh_host[2] - 1; ii++) {
                 max_nbor_size = (jrange_host[ii + 1] - jrange_host[ii]) > max_nbor_size ? (jrange_host[ii + 1] - jrange_host[ii]) : max_nbor_size;
             }
             assert(max_nbor_size <= GPU_MAX_NBOR_SIZE);