diff --git a/services/surfaceflinger/Scheduler/LayerInfoV2.cpp b/services/surfaceflinger/Scheduler/LayerInfoV2.cpp
index bf1fb88..b7d0bdd 100644
--- a/services/surfaceflinger/Scheduler/LayerInfoV2.cpp
+++ b/services/surfaceflinger/Scheduler/LayerInfoV2.cpp
@@ -111,23 +111,46 @@
 
     // Calculate the refresh rate by finding the average delta between frames
     nsecs_t totalPresentTimeDeltas = 0;
+    nsecs_t totalQueueTimeDeltas = 0;
+    auto missingPresentTime = false;
     for (auto it = mFrameTimes.begin(); it != mFrameTimes.end() - 1; ++it) {
-        // If there are no presentation timestamp provided we can't calculate the refresh rate
+        totalQueueTimeDeltas +=
+                std::max(((it + 1)->queueTime - it->queueTime), mHighRefreshRatePeriod);
+
         if (it->presetTime == 0 || (it + 1)->presetTime == 0) {
-            return std::nullopt;
+            missingPresentTime = true;
+            continue;
         }
 
         totalPresentTimeDeltas +=
                 std::max(((it + 1)->presetTime - it->presetTime), mHighRefreshRatePeriod);
     }
+
+    // If there are no presentation timestamps provided we can't calculate the refresh rate
+    if (missingPresentTime && mLastReportedRefreshRate == 0) {
+        return std::nullopt;
+    }
+
+    // Calculate the average frame time based on presentation timestamps. If those
+    // doesn't exist, we look at the time the buffer was queued only. We can do that only if
+    // we calculated a refresh rate based on presentation timestamps in the past. The reason
+    // we look at the queue time is to handle cases where hwui attaches presentation timestamps
+    // when implementing render ahead for specific refresh rates. When hwui no longer provides
+    // presentation timestamps we look at the queue time to see if the current refresh rate still
+    // matches the content.
     const float averageFrameTime =
-            static_cast<float>(totalPresentTimeDeltas) / (mFrameTimes.size() - 1);
+            static_cast<float>(missingPresentTime ? totalQueueTimeDeltas : totalPresentTimeDeltas) /
+            (mFrameTimes.size() - 1);
 
     // Now once we calculated the refresh rate we need to make sure that all the frames we captured
     // are evenly distributed and we don't calculate the average across some burst of frames.
     for (auto it = mFrameTimes.begin(); it != mFrameTimes.end() - 1; ++it) {
-        const nsecs_t presentTimeDeltas =
-                std::max(((it + 1)->presetTime - it->presetTime), mHighRefreshRatePeriod);
+        const auto presentTimeDeltas = [&] {
+            const auto delta = missingPresentTime ? (it + 1)->queueTime - it->queueTime
+                                                  : (it + 1)->presetTime - it->presetTime;
+            return std::max(delta, mHighRefreshRatePeriod);
+        }();
+
         if (std::abs(presentTimeDeltas - averageFrameTime) > 2 * averageFrameTime) {
             return std::nullopt;
         }
diff --git a/services/surfaceflinger/Scheduler/LayerInfoV2.h b/services/surfaceflinger/Scheduler/LayerInfoV2.h
index ad91f18..e36b7f7 100644
--- a/services/surfaceflinger/Scheduler/LayerInfoV2.h
+++ b/services/surfaceflinger/Scheduler/LayerInfoV2.h
@@ -88,6 +88,7 @@
         // buffer as Max as we don't know anything about this layer or Min as this layer is
         // posting infrequent updates.
         mFrameTimeValidSince = std::chrono::steady_clock::now();
+        mLastReportedRefreshRate = 0.0f;
     }
 
 private: