diff --git a/RecoLocalCalo/EcalRecProducers/plugins/EcalUncalibRecHitProducerGPU.cc b/RecoLocalCalo/EcalRecProducers/plugins/EcalUncalibRecHitProducerGPU.cc
index d043d0f8e6e50..1ecbb9f5e5ece 100644
--- a/RecoLocalCalo/EcalRecProducers/plugins/EcalUncalibRecHitProducerGPU.cc
+++ b/RecoLocalCalo/EcalRecProducers/plugins/EcalUncalibRecHitProducerGPU.cc
@@ -72,7 +72,6 @@ class EcalUncalibRecHitProducerGPU : public edm::stream::EDProducer<edm::Externa
   // event data
   ecal::multifit::EventOutputDataGPU eventOutputDataGPU_;
   ecal::multifit::EventDataForScratchGPU eventDataForScratchGPU_;
-  bool shouldTransferToHost_{true};
 
   cms::cuda::ContextState cudaState_;
 
@@ -128,7 +127,6 @@ void EcalUncalibRecHitProducerGPU::fillDescriptions(edm::ConfigurationDescriptio
   desc.add<double>("amplitudeThresholdEB", 10);
   desc.add<double>("amplitudeThresholdEE", 10);
   desc.add<uint32_t>("maxNumberHits", 20000);  //---- AM TEST
-  desc.add<bool>("shouldTransferToHost", true);
   desc.add<std::vector<uint32_t>>("kernelMinimizeThreads", {32, 1, 1});
   // ---- default false or true? It was set to true, but at HLT it is false
   desc.add<bool>("shouldRunTimingComputation", false);
@@ -174,9 +172,6 @@ EcalUncalibRecHitProducerGPU::EcalUncalibRecHitProducerGPU(const edm::ParameterS
   // max number of digis to allocate for
   maxNumberHits_ = ps.getParameter<uint32_t>("maxNumberHits");
 
-  // transfer to host switch
-  shouldTransferToHost_ = ps.getParameter<bool>("shouldTransferToHost");
-
   // switch to run timing computation kernels
   configParameters_.shouldRunTimingComputation = ps.getParameter<bool>("shouldRunTimingComputation");