diff --git a/GPU/GPUTracking/Base/GPUMemoryResource.h b/GPU/GPUTracking/Base/GPUMemoryResource.h
index 79960af9bc188..9f1e824488c4d 100644
--- a/GPU/GPUTracking/Base/GPUMemoryResource.h
+++ b/GPU/GPUTracking/Base/GPUMemoryResource.h
@@ -31,6 +31,7 @@ struct GPUMemoryReuse {
   };
   enum Group : uint16_t {
     ClustererScratch,
+    ClustererScratchClusters,
     NNClusterer,
     ClustererZS,
     TrackerScratch,
diff --git a/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx b/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx
index 1aa9690453b87..8ce90c740804a 100644
--- a/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx
+++ b/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx
@@ -1367,6 +1367,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
 
           if (doGPU && propagateMCLabels) {
             TransferMemoryResourceLinkToHost(RecoStep::TPCClusterFinding, clusterer.mScratchId, lane);
+            TransferMemoryResourceLinkToHost(RecoStep::TPCClusterFinding, clusterer.mScratchClustersId, lane);
             SynchronizeStream(lane);
             runKernel<GPUTPCCFClusterizer>({GetGrid(clusterer.mPmemory->counters.nClusters, lane, GPUReconstruction::krnlDeviceType::CPU), {iSector}}, 1); // Computes MC labels
           }
@@ -1382,7 +1383,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
             SynchronizeStream(lane);
           }
           if (doGPU && propagateMCLabels) {
-            TransferMemoryResourceLinkToHost(RecoStep::TPCClusterFinding, clusterer.mScratchId, lane);
+            TransferMemoryResourceLinkToHost(RecoStep::TPCClusterFinding, clusterer.mScratchClustersId, lane);
             SynchronizeStream(lane);
             runKernel<GPUTPCCFHIPClusterizer>({GetGrid(GPUTPCGeometry::NROWS, lane, GPUReconstruction::krnlDeviceType::CPU), {iSector}}, 1); // Computes MC labels
           }
diff --git a/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.cxx b/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.cxx
index c0c869d0dcc9b..80415dc043bf8 100644
--- a/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.cxx
+++ b/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.cxx
@@ -96,10 +96,14 @@ void* GPUTPCClusterFinder::SetPointersScratch(void* mem)
   computePointerWithAlignment(mem, mPisPeak, mNMaxDigitsFragment);
   computePointerWithAlignment(mem, mPchargeMap, TPCMapMemoryLayout<decltype(*mPchargeMap)>::items(mRec->GetProcessingSettings().overrideClusterizerFragmentLen));
   computePointerWithAlignment(mem, mPpeakMap, TPCMapMemoryLayout<decltype(*mPpeakMap)>::items(mRec->GetProcessingSettings().overrideClusterizerFragmentLen));
-  computePointerWithAlignment(mem, mPclusterByRow, GPUTPCGeometry::NROWS * mNMaxClusterPerRow);
   if ((mRec->GetRecoStepsGPU() & gpudatatypes::RecoStep::TPCClusterFinding)) {
     computePointerWithAlignment(mem, mPscanBuf, mBufSize * mNBufs);
   }
+  return mem;
+}
+void* GPUTPCClusterFinder::SetPointersScratchClusters(void* mem)
+{
+  computePointerWithAlignment(mem, mPclusterByRow, GPUTPCGeometry::NROWS * mNMaxClusterPerRow);
   // TODO: Use memory scalers for MaxHIPTails.
   // NOTE: Always allocate since Param() is not available during size computation.
   computePointerWithAlignment(mem, mPhipTailsByRow, GPUTPCGeometry::NROWS * GPUTPCCFHIPClusterizer::MaxHIPTailsPerRow);
@@ -117,6 +121,7 @@ void GPUTPCClusterFinder::RegisterMemoryAllocation()
     scratchType |= GPUMemoryResource::MEMORY_HOST | GPUMemoryResource::MEMORY_GPU;
   }
   mScratchId = mRec->RegisterMemoryAllocation(this, &GPUTPCClusterFinder::SetPointersScratch, scratchType, "TPCClustererScratch", GPUMemoryReuse{GPUMemoryReuse::REUSE_1TO1, GPUMemoryReuse::ClustererScratch, (uint16_t)(mISector % mRec->GetProcessingSettings().nTPCClustererLanes)});
+  mScratchClustersId = mRec->RegisterMemoryAllocation(this, &GPUTPCClusterFinder::SetPointersScratchClusters, scratchType, "TPCClustererScratchClusters", GPUMemoryReuse{GPUMemoryReuse::REUSE_1TO1, GPUMemoryReuse::ClustererScratchClusters, (uint16_t)(mISector % mRec->GetProcessingSettings().nTPCClustererLanes)});
 
   mMemoryId = mRec->RegisterMemoryAllocation(this, &GPUTPCClusterFinder::SetPointersMemory, GPUMemoryResource::MEMORY_PERMANENT, "TPCClustererMemory");
   mRec->RegisterMemoryAllocation(this, &GPUTPCClusterFinder::SetPointersOutput, GPUMemoryResource::MEMORY_OUTPUT | GPUMemoryResource::MEMORY_STACK, "TPCClustererOutput");
diff --git a/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.h b/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.h
index 513bb750abf9b..6922678d83a7d 100644
--- a/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.h
+++ b/GPU/GPUTracking/TPCClusterFinder/GPUTPCClusterFinder.h
@@ -85,6 +85,7 @@ class GPUTPCClusterFinder : public GPUProcessor
   void* SetPointersInput(void* mem);
   void* SetPointersOutput(void* mem);
   void* SetPointersScratch(void* mem);
+  void* SetPointersScratchClusters(void* mem);
   void* SetPointersMemory(void* mem);
   void* SetPointersZS(void* mem);
   void* SetPointersZSOffset(void* mem);
@@ -140,6 +141,7 @@ class GPUTPCClusterFinder : public GPUProcessor
 
   int16_t mMemoryId = -1;
   int16_t mScratchId = -1;
+  int16_t mScratchClustersId = -1;
   int16_t mZSId = -1;
   int16_t mZSOffsetId = -1;
   int16_t mOutputId = -1;