diff --git a/Makefile.am b/Makefile.am
index 2cf49ba..ea8c9b7 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -105,7 +105,7 @@ x11/cuda_x11_luffa512.o: x11/cuda_x11_luffa512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
x11/cuda_x11_luffa512_Cubehash.o: x11/cuda_x11_luffa512_Cubehash.cu
- $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
+ $(NVCC) $(nvcc_FLAGS) --maxrregcount=76 -o $@ -c $<
x13/cuda_x13_hamsi512.o: x13/cuda_x13_hamsi512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=72 -o $@ -c $<
diff --git a/ccminer.vcxproj b/ccminer.vcxproj
index d42e263..235e36c 100644
--- a/ccminer.vcxproj
+++ b/ccminer.vcxproj
@@ -455,7 +455,7 @@
- 80
+ 76
128
diff --git a/x11/cuda_x11_luffa512_Cubehash.cu b/x11/cuda_x11_luffa512_Cubehash.cu
index 1e7b914..cab0062 100644
--- a/x11/cuda_x11_luffa512_Cubehash.cu
+++ b/x11/cuda_x11_luffa512_Cubehash.cu
@@ -731,8 +731,6 @@ static void finalization512(uint32_t *statebuffer, uint32_t *statechainv, uint32
__global__
#if __CUDA_ARCH__ > 500
__launch_bounds__(256, 4)
-#else
-__launch_bounds__(256, 3)
#endif
void x11_luffaCubehash512_gpu_hash_64(uint32_t threads, uint32_t *g_hash)
{