diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 07dc81f3..4764192c 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -42,10 +42,13 @@ add_library(popsift SHARED
 target_link_libraries(popsift
                       PUBLIC
 		      CUDA::cudart
-		      CUDA::nvToolsExt
 		      Threads::Threads)
-# expected library to link was: "CUDA::nvtx3" but it appears to be simply nvToolsExt
-# or maybe ${NVTX_LIBRARY} ???
+
+if(PopSift_USE_NVTX_PROFILING)
+target_link_libraries(popsift
+                      PUBLIC
+		      CUDA::nvtx3)
+endif()
 
 # It is necessary to choose between shared and static, otherwise the CUDA_RUNTIME_LIBRARY target property
 # will not be set.
diff --git a/src/application/CMakeLists.txt b/src/application/CMakeLists.txt
index 0b6bdf20..2379c57d 100755
--- a/src/application/CMakeLists.txt
+++ b/src/application/CMakeLists.txt
@@ -60,10 +60,6 @@ else()
   set(PD_COMPILE_OPTIONS "" )
 endif()
 
-if(PopSift_USE_NVTX_PROFILING)
-  list(APPEND PD_LINK_LIBS ${CUDA_NVTX_LIBRARY})
-endif(PopSift_USE_NVTX_PROFILING)
-
 #############################################################
 # popsift-demo
 #############################################################
diff --git a/src/popsift/popsift.h b/src/popsift/popsift.h
index 3b5f72b8..5654cc76 100755
--- a/src/popsift/popsift.h
+++ b/src/popsift/popsift.h
@@ -24,7 +24,7 @@
 #include <vector>
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangeStartA(a)
 #define nvtxRangeEnd(a)
diff --git a/src/popsift/s_filtergrid.cu b/src/popsift/s_filtergrid.cu
index a766c2de..bfe2e64e 100644
--- a/src/popsift/s_filtergrid.cu
+++ b/src/popsift/s_filtergrid.cu
@@ -10,7 +10,7 @@
 #include "sift_pyramid.h"
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangePushA(a)
 #define nvtxRangePop()
diff --git a/src/popsift/s_image.cu b/src/popsift/s_image.cu
index f26b8e3e..a966dd39 100755
--- a/src/popsift/s_image.cu
+++ b/src/popsift/s_image.cu
@@ -16,7 +16,7 @@
 #include <iostream>
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangePushA(a)
 #define nvtxRangePop()
diff --git a/src/popsift/s_orientation.cu b/src/popsift/s_orientation.cu
index f6b36fcd..b34aaaa1 100644
--- a/src/popsift/s_orientation.cu
+++ b/src/popsift/s_orientation.cu
@@ -19,7 +19,7 @@
 #include <cstdio>
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangePushA(a)
 #define nvtxRangePop()
diff --git a/src/popsift/sift_desc.cu b/src/popsift/sift_desc.cu
index b0eb0bd1..f533df35 100644
--- a/src/popsift/sift_desc.cu
+++ b/src/popsift/sift_desc.cu
@@ -22,7 +22,7 @@
 #include <iostream>
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangePushA(a)
 #define nvtxRangePop()
diff --git a/src/popsift/sift_pyramid.cu b/src/popsift/sift_pyramid.cu
index 06060052..c03b0d61 100644
--- a/src/popsift/sift_pyramid.cu
+++ b/src/popsift/sift_pyramid.cu
@@ -26,7 +26,7 @@
 #endif
 
 #if POPSIFT_IS_DEFINED(POPSIFT_USE_NVTX)
-#include <nvToolsExtCuda.h>
+#include <nvtx3/nvToolsExtCuda.h>
 #else
 #define nvtxRangePushA(a)
 #define nvtxRangePop()