diff --git a/Source/Core/Core/CoreTiming.cpp b/Source/Core/Core/CoreTiming.cpp
index af5bc4731ce..a692e8db6f3 100644
--- a/Source/Core/Core/CoreTiming.cpp
+++ b/Source/Core/Core/CoreTiming.cpp
@@ -15,6 +15,7 @@
 #include "Common/ChunkFile.h"
 #include "Common/Logging/Log.h"
 #include "Common/SPSCQueue.h"
+#include "Common/ScopeGuard.h"
 
 #include "Core/AchievementManager.h"
 #include "Core/CPUThreadConfigCallback.h"
@@ -449,6 +450,12 @@ void CoreTimingManager::Throttle(const s64 target_cycle)
   if (skip_throttle)
     return;
 
+  // Measure current performance after throttling.
+  Common::ScopeGuard perf_marker{[&] {
+    g_perf_metrics.CountPerformanceMarker(target_cycle,
+                                          m_system.GetSystemTimers().GetTicksPerSecond());
+  }};
+
   if (IsSpeedUnlimited())
   {
     ResetThrottle(target_cycle);
diff --git a/Source/Core/Core/HW/VideoInterface.cpp b/Source/Core/Core/HW/VideoInterface.cpp
index f97852df7aa..48b81916a5c 100644
--- a/Source/Core/Core/HW/VideoInterface.cpp
+++ b/Source/Core/Core/HW/VideoInterface.cpp
@@ -950,9 +950,6 @@ void VideoInterfaceManager::Update(u64 ticks)
     // Throttle before SI poll so user input is taken just before needed. (lower input latency)
     core_timing.Throttle(ticks);
 
-    // This is a nice place to measure performance so we don't have to Throttle elsewhere.
-    g_perf_metrics.CountPerformanceMarker(ticks, m_system.GetSystemTimers().GetTicksPerSecond());
-
     Core::UpdateInputGate(!Config::Get(Config::MAIN_INPUT_BACKGROUND_INPUT),
                           Config::Get(Config::MAIN_LOCK_CURSOR));
     auto& si = m_system.GetSerialInterface();