diff --git a/src/runtime/pprof/pprof.go b/src/runtime/pprof/pprof.go
index b702fd345d..901467fa9a 100644
--- a/src/runtime/pprof/pprof.go
+++ b/src/runtime/pprof/pprof.go
@@ -296,22 +296,25 @@ func printCountProfile(w io.Writer, debug int, name string, p countProfile) erro
 		}
 		return buf.String()
 	}
-	m := map[string]int{}
+	count := map[string]int{}
+	index := map[string]int{}
+	var keys []string
 	n := p.Len()
 	for i := 0; i < n; i++ {
-		m[key(p.Stack(i))]++
+		k := key(p.Stack(i))
+		if count[k] == 0 {
+			index[k] = i
+			keys = append(keys, k)
+		}
+		count[k]++
 	}
 
-	// Print stacks, listing count on first occurrence of a unique stack.
-	for i := 0; i < n; i++ {
-		stk := p.Stack(i)
-		s := key(stk)
-		if count := m[s]; count != 0 {
-			fmt.Fprintf(w, "%d %s\n", count, s)
-			if debug > 0 {
-				printStackRecord(w, stk, false)
-			}
-			delete(m, s)
+	sort.Sort(&keysByCount{keys, count})
+
+	for _, k := range keys {
+		fmt.Fprintf(w, "%d %s\n", count[k], k)
+		if debug > 0 {
+			printStackRecord(w, p.Stack(index[k]), false)
 		}
 	}
 
@@ -321,6 +324,23 @@ func printCountProfile(w io.Writer, debug int, name string, p countProfile) erro
 	return b.Flush()
 }
 
+// keysByCount sorts keys with higher counts first, breaking ties by key string order.
+type keysByCount struct {
+	keys  []string
+	count map[string]int
+}
+
+func (x *keysByCount) Len() int      { return len(x.keys) }
+func (x *keysByCount) Swap(i, j int) { x.keys[i], x.keys[j] = x.keys[j], x.keys[i] }
+func (x *keysByCount) Less(i, j int) bool {
+	ki, kj := x.keys[i], x.keys[j]
+	ci, cj := x.count[ki], x.count[kj]
+	if ci != cj {
+		return ci > cj
+	}
+	return ki < kj
+}
+
 // printStackRecord prints the function + source line information
 // for a single stack trace.
 func printStackRecord(w io.Writer, stk []uintptr, allFrames bool) {
diff --git a/src/runtime/pprof/pprof_test.go b/src/runtime/pprof/pprof_test.go
index d232257c20..fa0af59b37 100644
--- a/src/runtime/pprof/pprof_test.go
+++ b/src/runtime/pprof/pprof_test.go
@@ -578,3 +578,47 @@ func blockCond() {
 	c.Wait()
 	mu.Unlock()
 }
+
+func func1(c chan int) { <-c }
+func func2(c chan int) { <-c }
+func func3(c chan int) { <-c }
+func func4(c chan int) { <-c }
+
+func TestGoroutineCounts(t *testing.T) {
+	c := make(chan int)
+	for i := 0; i < 100; i++ {
+		if i%10 == 0 {
+			go func1(c)
+			continue
+		}
+		if i%2 == 0 {
+			go func2(c)
+			continue
+		}
+		go func3(c)
+	}
+	time.Sleep(10 * time.Millisecond) // let goroutines block on channel
+
+	var w bytes.Buffer
+	Lookup("goroutine").WriteTo(&w, 1)
+	prof := w.String()
+
+	if !containsInOrder(prof, "\n50 @ ", "\n40 @", "\n10 @", "\n1 @") {
+		t.Errorf("expected sorted goroutine counts:\n%s", prof)
+	}
+
+	close(c)
+
+	time.Sleep(10 * time.Millisecond) // let goroutines exit
+}
+
+func containsInOrder(s string, all ...string) bool {
+	for _, t := range all {
+		i := strings.Index(s, t)
+		if i < 0 {
+			return false
+		}
+		s = s[i+len(t):]
+	}
+	return true
+}