diff --git a/src/control/server/ctl_storage_rpc_test.go b/src/control/server/ctl_storage_rpc_test.go
index f2a402cfd5e..b2d25ad8b2e 100644
--- a/src/control/server/ctl_storage_rpc_test.go
+++ b/src/control/server/ctl_storage_rpc_test.go
@@ -1569,7 +1569,10 @@ func TestServer_CtlSvc_StorageScan_PostEngineStart(t *testing.T) {
 			var engineCfgs []*engine.Config
 			for i, sc := range tc.storageCfgs {
 				log.Debugf("storage cfg contains bdevs %v for engine %d", sc.Bdevs(), i)
-				engineCfgs = append(engineCfgs, engine.MockConfig().WithStorage(sc...))
+				engineCfgs = append(engineCfgs,
+					engine.MockConfig().
+						WithStorage(sc...).
+						WithTargetCount(tc.engineTargetCount[i]))
 			}
 			sCfg := config.DefaultServer().WithEngines(engineCfgs...)
 			cs := mockControlService(t, log, sCfg, csbmbc, tc.smbc, tc.smsc)
@@ -1625,7 +1628,6 @@ func TestServer_CtlSvc_StorageScan_PostEngineStart(t *testing.T) {
 				}
 				te.setDrpcClient(newMockDrpcClient(dcc))
 				te._superblock.Rank = ranklist.NewRankPtr(uint32(idx + 1))
-				te.setTargetCount(tc.engineTargetCount[idx])
 				for _, tc := range te.storage.GetBdevConfigs() {
 					tc.Bdev.DeviceRoles.OptionBits = storage.OptionBits(storage.BdevRoleAll)
 				}
diff --git a/src/control/server/instance.go b/src/control/server/instance.go
index 3837860fabd..14f53cf3b5b 100644
--- a/src/control/server/instance.go
+++ b/src/control/server/instance.go
@@ -338,14 +338,6 @@ func (ei *EngineInstance) setHugepageSz(hpSizeMb int) {
 	ei.runner.GetConfig().HugepageSz = hpSizeMb
 }
 
-// setTargetCount updates target count in engine config.
-func (ei *EngineInstance) setTargetCount(numTargets int) {
-	ei.Lock()
-	defer ei.Unlock()
-
-	ei.runner.GetConfig().TargetCount = numTargets
-}
-
 // GetTargetCount returns the target count set for this instance.
 func (ei *EngineInstance) GetTargetCount() int {
 	ei.RLock()
diff --git a/src/control/server/instance_exec.go b/src/control/server/instance_exec.go
index 3a31d036137..ab22cb4504f 100644
--- a/src/control/server/instance_exec.go
+++ b/src/control/server/instance_exec.go
@@ -90,12 +90,6 @@ func (ei *EngineInstance) finishStartup(ctx context.Context, ready *srvpb.Notify
 	if err := ei.handleReady(ctx, ready); err != nil {
 		return err
 	}
-	// update engine target count to reflect allocated number of targets, not number requested
-	// when starting
-	// NOTE: Engine mem_size passed on engine invocation is based on the number of targets
-	//       requested in config so if number of targets allocated doesn't match the number of
-	//       targets requested the mem_size value may be inappropriate.
-	ei.setTargetCount(int(ready.GetNtgts()))
 
 	ei.ready.SetTrue()
 
diff --git a/src/engine/init.c b/src/engine/init.c
index a376488e62b..8d580188022 100644
--- a/src/engine/init.c
+++ b/src/engine/init.c
@@ -40,8 +40,7 @@
 static char		modules[MAX_MODULE_OPTIONS + 1];
 
 /**
- * Number of target threads the user would like to start
- * 0 means default value, see dss_tgt_nr_get();
+ * Number of target threads the user would like to start.
  */
 static unsigned int	nr_threads;
 
@@ -250,56 +249,61 @@ modules_load(void)
 	return rc;
 }
 
+static unsigned int
+ncores_needed(unsigned int tgt_nr, unsigned int nr_helpers)
+{
+	return DAOS_TGT0_OFFSET + tgt_nr + nr_helpers;
+}
+
 /**
- * Get the appropriate number of main XS based on the number of cores and
- * passed in preferred number of threads.
+ * Check if the #targets and #nr_xs_helpers is valid to start server, the #nr_xs_helpers possibly
+ * be reduced.
  */
 static int
-dss_tgt_nr_get(unsigned int ncores, unsigned int nr, bool oversubscribe)
+dss_tgt_nr_check(unsigned int ncores, unsigned int tgt_nr, bool oversubscribe)
 {
-	int tgt_nr;
-
 	D_ASSERT(ncores >= 1);
 
 	/* at most 2 helper XS per target */
-	if (dss_tgt_offload_xs_nr > 2 * nr)
-		dss_tgt_offload_xs_nr = 2 * nr;
-	else if (dss_tgt_offload_xs_nr == 0)
+	if (dss_tgt_offload_xs_nr > 2 * tgt_nr) {
+		D_PRINT("#nr_xs_helpers(%d) cannot exceed 2 times #targets (2 x %d = %d).\n",
+			dss_tgt_offload_xs_nr, tgt_nr, 2 * tgt_nr);
+		dss_tgt_offload_xs_nr = 2 * tgt_nr;
+	} else if (dss_tgt_offload_xs_nr == 0) {
 		D_WARN("Suggest to config at least 1 helper XS per DAOS engine\n");
+	}
 
-	/* Each system XS uses one core, and  with dss_tgt_offload_xs_nr
-	 * offload XS. Calculate the tgt_nr as the number of main XS based
-	 * on number of cores.
-	 */
-retry:
-	tgt_nr = ncores - DAOS_TGT0_OFFSET - dss_tgt_offload_xs_nr;
-	if (tgt_nr <= 0)
-		tgt_nr = 1;
-
-	/* If user requires less target threads then set it as dss_tgt_nr,
-	 * if user oversubscribes, then:
-	 *      . if oversubscribe is enabled, use the required number
-	 *      . if oversubscribe is disabled(default),
-	 *        use the number calculated above
-	 * Note: oversubscribing  may hurt performance.
-	 */
-	if (nr >= 1 && ((nr < tgt_nr) || oversubscribe)) {
-		tgt_nr = nr;
-		if (dss_tgt_offload_xs_nr > 2 * tgt_nr)
-			dss_tgt_offload_xs_nr = 2 * tgt_nr;
-	} else if (dss_tgt_offload_xs_nr > 2 * tgt_nr) {
-		dss_tgt_offload_xs_nr--;
-		goto retry;
+	if (oversubscribe) {
+		if (ncores_needed(tgt_nr, dss_tgt_offload_xs_nr) > ncores) {
+			if (ncores > DAOS_TGT0_OFFSET + tgt_nr)
+				dss_tgt_offload_xs_nr = ncores - DAOS_TGT0_OFFSET - tgt_nr;
+			else
+				dss_tgt_offload_xs_nr = 0;
+
+			D_PRINT("Force to start engine with %d targets on %d cores, #nr_xs_helpers "
+				"set as %d.\n",
+				tgt_nr, ncores, dss_tgt_offload_xs_nr);
+		}
+		goto out;
 	}
 
-	if (tgt_nr != nr)
-		D_PRINT("%d target XS(xstream) requested (#cores %d); "
-			"use (%d) target XS\n", nr, ncores, tgt_nr);
+	if (ncores_needed(tgt_nr, dss_tgt_offload_xs_nr) > ncores) {
+		if (ncores < DAOS_TGT0_OFFSET + tgt_nr) {
+			D_ERROR("cannot start engine with %d targets on %d cores, may try with "
+				"DAOS_TARGET_OVERSUBSCRIBE=1\n",
+				tgt_nr, ncores);
+			return -DER_INVAL;
+		}
+		dss_tgt_offload_xs_nr = ncores - DAOS_TGT0_OFFSET - tgt_nr;
+		D_PRINT("Start engine with %d targets on %d cores, #nr_xs_helpers set as %d.\n",
+			tgt_nr, ncores, dss_tgt_offload_xs_nr);
+	}
 
+out:
 	if (dss_tgt_offload_xs_nr % tgt_nr != 0)
 		dss_helper_pool = true;
 
-	return tgt_nr;
+	return 0;
 }
 
 static int
@@ -321,14 +325,12 @@ dss_topo_init()
 	depth = hwloc_get_type_depth(dss_topo, HWLOC_OBJ_NUMANODE);
 	numa_node_nr = hwloc_get_nbobjs_by_depth(dss_topo, depth);
 	d_getenv_bool("DAOS_TARGET_OVERSUBSCRIBE", &tgt_oversub);
+	dss_tgt_nr = nr_threads;
 
 	/* if no NUMA node was specified, or NUMA data unavailable */
 	/* fall back to the legacy core allocation algorithm */
 	if (dss_numa_node == -1 || numa_node_nr <= 0) {
 		D_PRINT("Using legacy core allocation algorithm\n");
-		dss_tgt_nr = dss_tgt_nr_get(dss_core_nr, nr_threads,
-					    tgt_oversub);
-
 		if (dss_core_offset >= dss_core_nr) {
 			D_ERROR("invalid dss_core_offset %u "
 				"(set by \"-f\" option),"
@@ -336,7 +338,8 @@ dss_topo_init()
 				dss_core_offset, dss_core_nr - 1);
 			return -DER_INVAL;
 		}
-		return 0;
+
+		return dss_tgt_nr_check(dss_core_nr, dss_tgt_nr, tgt_oversub);
 	}
 
 	if (dss_numa_node > numa_node_nr) {
@@ -381,17 +384,15 @@ dss_topo_init()
 	hwloc_bitmap_asprintf(&cpuset, core_allocation_bitmap);
 	free(cpuset);
 
-	dss_tgt_nr = dss_tgt_nr_get(dss_num_cores_numa_node, nr_threads,
-				    tgt_oversub);
 	if (dss_core_offset >= dss_num_cores_numa_node) {
 		D_ERROR("invalid dss_core_offset %d (set by \"-f\" option), "
 			"should within range [0, %d]", dss_core_offset,
 			dss_num_cores_numa_node - 1);
 		return -DER_INVAL;
 	}
-
 	D_PRINT("Using NUMA core allocation algorithm\n");
-	return 0;
+
+	return dss_tgt_nr_check(dss_num_cores_numa_node, dss_tgt_nr, tgt_oversub);
 }
 
 static ABT_mutex		server_init_state_mutex;
diff --git a/src/tests/ftest/control/dmg_server_set_logmasks.yaml b/src/tests/ftest/control/dmg_server_set_logmasks.yaml
index b4e3a1ddfeb..68e58cc7488 100644
--- a/src/tests/ftest/control/dmg_server_set_logmasks.yaml
+++ b/src/tests/ftest/control/dmg_server_set_logmasks.yaml
@@ -6,6 +6,7 @@ server_config:
   engines_per_host: 1
   engines:
     0:
+      targets: 4
       storage:
         0:
           class: ram
diff --git a/src/tests/ftest/harness/core_files.yaml b/src/tests/ftest/harness/core_files.yaml
index 8133398a42f..04cb67f7a11 100644
--- a/src/tests/ftest/harness/core_files.yaml
+++ b/src/tests/ftest/harness/core_files.yaml
@@ -5,6 +5,7 @@ server_config:
   engines_per_host: 1
   engines:
     0:
+      targets: 4
       storage:
         0:
           class: ram
diff --git a/src/tests/ftest/pool/create_all_vm.yaml b/src/tests/ftest/pool/create_all_vm.yaml
index 0e030ee8cbc..2c053e5b038 100644
--- a/src/tests/ftest/pool/create_all_vm.yaml
+++ b/src/tests/ftest/pool/create_all_vm.yaml
@@ -32,7 +32,7 @@ server_config:
   engines_per_host: 1
   engines:
     0:
-      targets: 5
+      targets: 4
       nr_xs_helpers: 0
       storage:
         0:
diff --git a/utils/nlt_server.yaml b/utils/nlt_server.yaml
index 4b9a1a9ffd8..5d0d2d9b3ed 100644
--- a/utils/nlt_server.yaml
+++ b/utils/nlt_server.yaml
@@ -14,6 +14,7 @@ engines:
       - DAOS_MD_CAP=1024
       - DAOS_STRICT_SHUTDOWN=1
       - CRT_CTX_SHARE_ADDR=0
+      - DAOS_TARGET_OVERSUBSCRIBE=1
       - ABT_STACK_OVERFLOW_CHECK=mprotect
     storage:
       -