diff --git a/epochX/cudacpp/gg_tt.mad/SubProcesses/cudacpp.mk b/epochX/cudacpp/gg_tt.mad/SubProcesses/cudacpp.mk
index 364264b8fe..1b086329a7 100644
--- a/epochX/cudacpp/gg_tt.mad/SubProcesses/cudacpp.mk
+++ b/epochX/cudacpp/gg_tt.mad/SubProcesses/cudacpp.mk
@@ -74,12 +74,13 @@ endif
 
 #=== Configure the C++ compiler
 
-CXXFLAGS = $(OPTFLAGS) -std=c++17 $(INCFLAGS) $(USE_NVTX) -Wall -Wshadow -Wextra
+CXXFLAGS = $(OPTFLAGS) -std=c++17 $(INCFLAGS) -Wall -Wshadow -Wextra
 ifeq ($(shell $(CXX) --version | grep ^nvc++),)
 CXXFLAGS+= -ffast-math # see issue #117
 endif
 ###CXXFLAGS+= -Ofast # performance is not different from --fast-math
 ###CXXFLAGS+= -g # FOR DEBUGGING ONLY
+###CXXFLAGS+= $(USE_NVTX) # OPTIONALLY enable NVTX in C++ builds (but disable it by default #679)
 
 # Optionally add debug flags to display the full list of flags (eg on Darwin)
 ###CXXFLAGS+= -v