diff --git a/drivers/gpu/nvgpu/hal/pmu/pmu_gk20a.c b/drivers/gpu/nvgpu/hal/pmu/pmu_gk20a.c
index 2a5514f89..9ee1e9390 100644
--- a/drivers/gpu/nvgpu/hal/pmu/pmu_gk20a.c
+++ b/drivers/gpu/nvgpu/hal/pmu/pmu_gk20a.c
@@ -85,6 +85,10 @@ void gk20a_pmu_init_perfmon_counter(struct gk20a *g)
 		pwr_pmu_idle_mask_gr_enabled_f() |
 		pwr_pmu_idle_mask_ce_2_enabled_f());
 
+	/* assign same mask setting from GR ELPG to counter #3 */
+	data = gk20a_readl(g, pwr_pmu_idle_mask_1_supp_r(0));
+	gk20a_writel(g, pwr_pmu_idle_mask_1_r(3), data);
+
 	/* disable idle filtering for counters 3 and 6 */
 	data = gk20a_readl(g, pwr_pmu_idle_ctrl_r(3));
 	data = set_field(data, pwr_pmu_idle_ctrl_value_m() |
diff --git a/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.c b/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.c
index f3f00a2f6..f7838786c 100644
--- a/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.c
+++ b/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.c
@@ -33,6 +33,7 @@
 #include <nvgpu/pmu/cmd.h>
 #endif
 
+#include "pmu_gk20a.h"
 #include "pmu_gv11b.h"
 
 #include <nvgpu/hw/gv11b/hw_pwr_gv11b.h>
@@ -113,6 +114,17 @@ static struct pg_init_sequence_list _pginitseq_gv11b[] = {
 	{0x00020004U, 0x00000000U} ,
 };
 
+void gv11b_pmu_init_perfmon_counter(struct gk20a *g)
+{
+        u32 data;
+
+        gk20a_pmu_init_perfmon_counter(g);
+
+        /* assign same mask setting from GR ELPG to counter #3 */
+        data = gk20a_readl(g, pwr_pmu_idle_mask_2_supp_r(0));
+        gk20a_writel(g, pwr_pmu_idle_mask_2_r(3), data);
+}
+
 void gv11b_pmu_setup_elpg(struct gk20a *g)
 {
 	size_t reg_writes;
diff --git a/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.h b/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.h
index 7e9b8a1c4..17bc2e6ff 100644
--- a/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.h
+++ b/drivers/gpu/nvgpu/hal/pmu/pmu_gv11b.h
@@ -42,6 +42,7 @@ void gv11b_pmu_handle_ext_irq(struct gk20a *g, u32 intr0);
 #ifdef CONFIG_NVGPU_LS_PMU
 int gv11b_pmu_bootstrap(struct gk20a *g, struct nvgpu_pmu *pmu,
 	u32 args_offset);
+void gv11b_pmu_init_perfmon_counter(struct gk20a *g);
 void gv11b_pmu_setup_elpg(struct gk20a *g);
 void gv11b_secured_pmu_start(struct gk20a *g);
 void gv11b_write_dmatrfbase(struct gk20a *g, u32 addr);