diff --git a/build_files/buildbot/config/user-config-cuda-glibc211-i686.py b/build_files/buildbot/config/user-config-cuda-glibc211-i686.py
index ae5cbac8cc7..580e67ee7e3 100644
--- a/build_files/buildbot/config/user-config-cuda-glibc211-i686.py
+++ b/build_files/buildbot/config/user-config-cuda-glibc211-i686.py
@@ -2,4 +2,5 @@ BF_BUILDDIR = '../blender-build/linux-glibc211-i686'
 BF_INSTALLDIR = '../blender-install/linux-glibc211-i686'
 BF_NUMJOBS = 1
 
-BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50', 'sm_52']
+#BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50', 'sm_52']
+BF_CYCLES_CUDA_BINARIES_ARCH = []
diff --git a/intern/cycles/kernel/kernel_light.h b/intern/cycles/kernel/kernel_light.h
index eb2d39d5020..24d4b01302c 100644
--- a/intern/cycles/kernel/kernel_light.h
+++ b/intern/cycles/kernel/kernel_light.h
@@ -620,12 +620,7 @@ ccl_device void lamp_light_sample(KernelGlobals *kg, int lamp,
 	}
 }
 
-#if defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86))
-ccl_device_noinline
-#else
-ccl_device
-#endif
-bool lamp_light_eval(KernelGlobals *kg, int lamp, float3 P, float3 D, float t, LightSample *ls)
+ccl_device bool lamp_light_eval(KernelGlobals *kg, int lamp, float3 P, float3 D, float t, LightSample *ls)
 {
 	float4 data0 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 0);
 	float4 data1 = kernel_tex_fetch(__light_data, lamp*LIGHT_SIZE + 1);
diff --git a/intern/cycles/util/util_math.h b/intern/cycles/util/util_math.h
index 97c93516496..7d6dfd34e0e 100644
--- a/intern/cycles/util/util_math.h
+++ b/intern/cycles/util/util_math.h
@@ -1447,14 +1447,9 @@ ccl_device bool ray_triangle_intersect_uv(
 	return true;
 }
 
-#if defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86)) && (defined(__KERNEL_EXPERIMENTAL__) || __CUDA_ARCH__ == 500)
-ccl_device_noinline
-#else
-ccl_device
-#endif
-bool ray_quad_intersect(float3 ray_P, float3 ray_D, float ray_t,
-                        float3 quad_P, float3 quad_u, float3 quad_v,
-                        float3 *isect_P, float *isect_t)
+ccl_device bool ray_quad_intersect(float3 ray_P, float3 ray_D, float ray_t,
+                                   float3 quad_P, float3 quad_u, float3 quad_v,
+                                   float3 *isect_P, float *isect_t)
 {
 	float3 v0 = quad_P - quad_u*0.5f - quad_v*0.5f;
 	float3 v1 = quad_P + quad_u*0.5f - quad_v*0.5f;