diff --git a/DEPS b/DEPS
index 976639b42033..74fde4be9193 100644
--- a/DEPS
+++ b/DEPS
@@ -81,7 +81,7 @@ vars = {
   "args_rev": "73e8d3b55cbedc9765f8e266f3422d8914f8e62a",
   "async_rev": "f3ed5f690e2ec9dbe1bfc5184705575b4f6480e5",
   "bazel_worker_rev": "9710de6c9c70b1b583183db9d9721ba64e5a16fe",
-  "benchmark_harness_rev": "0ae822e264e410e9f1e927daea68601cc54906ef",
+  "benchmark_harness_rev": "f4ed0fcac4c284580263f0c548664dd548c9cca3",
   "boolean_selector_rev": "1d3565e2651d16566bb556955b96ea75018cbd0c",
   "boringssl_gen_rev": "ced85ef0a00bbca77ce5a91261a5f2ae61b1e62f",
   "boringssl_rev": "87f316d7748268eb56f2dc147bd593254ae93198",
diff --git a/runtime/tests/vm/dart/splay_ephemeron_test.dart b/runtime/tests/vm/dart/splay_ephemeron_test.dart
index 329d6d554d96..27853cc59139 100644
--- a/runtime/tests/vm/dart/splay_ephemeron_test.dart
+++ b/runtime/tests/vm/dart/splay_ephemeron_test.dart
@@ -103,7 +103,14 @@ class Splay extends BenchmarkBase {
 
   static void main() {
     mysetup();
-    new Splay().report();
+    // Don't use benchmark_harness - exercise runtime is not stable (so
+    // benchmark_harness measuring approach is meaningless for such benchmark
+    // anyway).
+    final sw = Stopwatch()..start();
+    final benchmark = new Splay();
+    while (sw.elapsedMilliseconds < 2000) {
+      benchmark.exercise();
+    }
     tearDown();
   }
 }
diff --git a/runtime/tests/vm/dart/splay_test.dart b/runtime/tests/vm/dart/splay_test.dart
index 1abb1979314b..e5085617045f 100644
--- a/runtime/tests/vm/dart/splay_test.dart
+++ b/runtime/tests/vm/dart/splay_test.dart
@@ -102,7 +102,14 @@ class Splay extends BenchmarkBase {
 
   static void main() {
     mysetup();
-    new Splay().report();
+    // Don't use benchmark_harness - exercise runtime is not stable (so
+    // benchmark_harness measuring approach is meaningless for such benchmark
+    // anyway).
+    final sw = Stopwatch()..start();
+    final benchmark = new Splay();
+    while (sw.elapsedMilliseconds < 2000) {
+      benchmark.exercise();
+    }
     tearDown();
   }
 }
diff --git a/runtime/tests/vm/dart_2/splay_ephemeron_test.dart b/runtime/tests/vm/dart_2/splay_ephemeron_test.dart
index 7ab290aa769b..580cd77bf467 100644
--- a/runtime/tests/vm/dart_2/splay_ephemeron_test.dart
+++ b/runtime/tests/vm/dart_2/splay_ephemeron_test.dart
@@ -102,7 +102,14 @@ class Splay extends BenchmarkBase {
 
   static void main() {
     mysetup();
-    new Splay().report();
+    // Don't use benchmark_harness - exercise runtime is not stable (so
+    // benchmark_harness measuring approach is meaningless for such benchmark
+    // anyway).
+    final sw = Stopwatch()..start();
+    final benchmark = new Splay();
+    while (sw.elapsedMilliseconds < 2000) {
+      benchmark.exercise();
+    }
     tearDown();
   }
 }
diff --git a/runtime/tests/vm/dart_2/splay_test.dart b/runtime/tests/vm/dart_2/splay_test.dart
index f1b401e47570..f0474f6b6d26 100644
--- a/runtime/tests/vm/dart_2/splay_test.dart
+++ b/runtime/tests/vm/dart_2/splay_test.dart
@@ -101,7 +101,14 @@ class Splay extends BenchmarkBase {
 
   static void main() {
     mysetup();
-    new Splay().report();
+    // Don't use benchmark_harness - exercise runtime is not stable (so
+    // benchmark_harness measuring approach is meaningless for such benchmark
+    // anyway).
+    final sw = Stopwatch()..start();
+    final benchmark = new Splay();
+    while (sw.elapsedMilliseconds < 2000) {
+      benchmark.exercise();
+    }
     tearDown();
   }
 }