diff --git a/renderdoc/driver/shaders/spirv/spirv_debug.h b/renderdoc/driver/shaders/spirv/spirv_debug.h
index 2cd114603..5876b50d6 100644
--- a/renderdoc/driver/shaders/spirv/spirv_debug.h
+++ b/renderdoc/driver/shaders/spirv/spirv_debug.h
@@ -240,6 +240,8 @@ struct ThreadState
   uint32_t quadNeighbours[4] = {~0U, ~0U, ~0U, ~0U};
   // index in the workgroup
   uint32_t workgroupIndex = 0;
+  // index in the subgroup
+  uint32_t subgroupId = 0;
   bool helperInvocation = false;
   bool dead = true;
   bool elected = false;
@@ -377,7 +379,7 @@ public:
                                const rdcstr &entryPoint, const rdcarray<SpecConstant> &specInfo,
                                const std::map<size_t, uint32_t> &instructionLines,
                                const SPIRVPatchData &patchData, uint32_t activeIndex,
-                               uint32_t workgroupSize);
+                               uint32_t threadsInWorkgroup, uint32_t threadsInSubgroup);
 
   rdcarray<ShaderDebugState> ContinueDebug();
 
@@ -416,6 +418,7 @@ public:
   const rdcarray<Id> &GetLiveGlobals() { return liveGlobals; }
   ThreadState &GetActiveLane() { return workgroup[activeLaneIndex]; }
   const ThreadState &GetActiveLane() const { return workgroup[activeLaneIndex]; }
+  uint32_t GetSubgroupSize() const { return subgroupSize; }
 private:
   virtual void PreParse(uint32_t maxId);
   virtual void PostParse();
@@ -447,6 +450,7 @@ private:
   Id convergeBlock;
 
   uint32_t activeLaneIndex = 0;
+  uint32_t subgroupSize = 0;
   ShaderStage stage;
 
   int steps = 0;
diff --git a/renderdoc/driver/shaders/spirv/spirv_debug_setup.cpp b/renderdoc/driver/shaders/spirv/spirv_debug_setup.cpp
index f7d9fe372..ddd74b265 100644
--- a/renderdoc/driver/shaders/spirv/spirv_debug_setup.cpp
+++ b/renderdoc/driver/shaders/spirv/spirv_debug_setup.cpp
@@ -839,7 +839,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
                                        const rdcarray<SpecConstant> &specInfo,
                                        const std::map<size_t, uint32_t> &instructionLines,
                                        const SPIRVPatchData &patchData, uint32_t activeIndex,
-                                       uint32_t workgroupSize)
+                                       uint32_t threadsInWorkgroup, uint32_t threadsInSubgroup)
 {
   Id entryId = entryLookup[ShaderEntryPoint(entryPoint, shaderStage)];
 
@@ -893,10 +893,11 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
   ret->debugger = this;
   ret->stage = shaderStage;
   activeLaneIndex = activeIndex;
+  subgroupSize = threadsInSubgroup;
   stage = shaderStage;
   apiWrapper = api;
 
-  for(uint32_t i = 0; i < workgroupSize; i++)
+  for(uint32_t i = 0; i < threadsInWorkgroup; i++)
     workgroup.push_back(ThreadState(*this, global));
 
   ThreadState &active = GetActiveLane();
@@ -1069,7 +1070,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
 
       if(isInput)
       {
-        for(laneIndex = 0; laneIndex < workgroupSize; laneIndex++)
+        for(laneIndex = 0; laneIndex < threadsInWorkgroup; laneIndex++)
         {
           // create the opaque storage
           workgroup[laneIndex].inputs.push_back(var);
@@ -1486,7 +1487,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
 
   std::sort(liveGlobals.begin(), liveGlobals.end());
 
-  for(uint32_t i = 0; i < workgroupSize; i++)
+  for(uint32_t i = 0; i < threadsInWorkgroup; i++)
   {
     ThreadState &lane = workgroup[i];
     lane.workgroupIndex = i;
@@ -1505,6 +1506,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
       lane.quadId = apiWrapper->GetThreadProperty(i, ThreadProperty::QuadId);
     }
 
+    lane.subgroupId = apiWrapper->GetThreadProperty(i, ThreadProperty::SubgroupId);
     lane.dead = apiWrapper->GetThreadProperty(i, ThreadProperty::Active) == 0;
 
     if(patchData.threadScope & ThreadScope::Subgroup)
@@ -1518,7 +1520,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
   // find quad neighbours
   {
     rdcarray<uint32_t> processedQuads;
-    for(uint32_t i = 0; i < workgroupSize; i++)
+    for(uint32_t i = 0; i < threadsInWorkgroup; i++)
     {
       uint32_t desiredQuad = workgroup[i].quadId;
 
@@ -1540,7 +1542,7 @@ ShaderDebugTrace *Debugger::BeginDebug(DebugAPIWrapper *api, const ShaderStage s
           ~0U,
           ~0U,
       };
-      for(uint32_t j = i + 1, t = 1; j < workgroupSize && t < 4; j++)
+      for(uint32_t j = i + 1, t = 1; j < threadsInWorkgroup && t < 4; j++)
       {
         if(workgroup[j].quadId == desiredQuad)
           threads[t++] = j;
diff --git a/renderdoc/driver/vulkan/vk_shaderdebug.cpp b/renderdoc/driver/vulkan/vk_shaderdebug.cpp
index 51ee5d7bd..e4a69a5cc 100644
--- a/renderdoc/driver/vulkan/vk_shaderdebug.cpp
+++ b/renderdoc/driver/vulkan/vk_shaderdebug.cpp
@@ -5125,7 +5125,7 @@ ShaderDebugTrace *VulkanReplay::DebugVertex(uint32_t eventId, uint32_t vertid, u
 
     ShaderDebugTrace *ret = debugger->BeginDebug(apiWrapper, ShaderStage::Vertex, entryPoint, spec,
                                                  shadRefl.instructionLines, shadRefl.patchData,
-                                                 winner->laneIndex, numThreads);
+                                                 winner->laneIndex, numThreads, numThreads);
     apiWrapper->ResetReplay();
 
     return ret;
@@ -5298,9 +5298,9 @@ ShaderDebugTrace *VulkanReplay::DebugVertex(uint32_t eventId, uint32_t vertid, u
 
     rdcspv::Debugger *debugger = new rdcspv::Debugger;
     debugger->Parse(shader.spirv.GetSPIRV());
-    ShaderDebugTrace *ret =
-        debugger->BeginDebug(apiWrapper, ShaderStage::Vertex, entryPoint, spec,
-                             shadRefl.instructionLines, shadRefl.patchData, laneIndex, numThreads);
+    ShaderDebugTrace *ret = debugger->BeginDebug(apiWrapper, ShaderStage::Vertex, entryPoint, spec,
+                                                 shadRefl.instructionLines, shadRefl.patchData,
+                                                 laneIndex, numThreads, numThreads);
     apiWrapper->ResetReplay();
 
     return ret;
@@ -5797,7 +5797,7 @@ ShaderDebugTrace *VulkanReplay::DebugPixel(uint32_t eventId, uint32_t x, uint32_
 
     ret = debugger->BeginDebug(apiWrapper, ShaderStage::Pixel, entryPoint, spec,
                                shadRefl.instructionLines, shadRefl.patchData, winner->laneIndex,
-                               numThreads);
+                               numThreads, numThreads);
     apiWrapper->ResetReplay();
   }
   else
@@ -6189,9 +6189,9 @@ ShaderDebugTrace *VulkanReplay::DebugComputeCommon(ShaderStage stage, uint32_t e
       }
     }
 
-    ShaderDebugTrace *ret =
-        debugger->BeginDebug(apiWrapper, stage, entryPoint, spec, shadRefl.instructionLines,
-                             shadRefl.patchData, winner->laneIndex, numThreads);
+    ShaderDebugTrace *ret = debugger->BeginDebug(
+        apiWrapper, stage, entryPoint, spec, shadRefl.instructionLines, shadRefl.patchData,
+        winner->laneIndex, numThreads, winner->subgroupSize);
     apiWrapper->ResetReplay();
 
     return ret;
@@ -6246,7 +6246,7 @@ ShaderDebugTrace *VulkanReplay::DebugComputeCommon(ShaderStage stage, uint32_t e
     debugger->Parse(shader.spirv.GetSPIRV());
     ShaderDebugTrace *ret =
         debugger->BeginDebug(apiWrapper, stage, entryPoint, spec, shadRefl.instructionLines,
-                             shadRefl.patchData, laneIndex, numThreads);
+                             shadRefl.patchData, laneIndex, numThreads, 1);
     apiWrapper->ResetReplay();
 
     return ret;