diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index 9fe618f855d1..161c6a24960b 100644
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -5374,6 +5374,22 @@ static enum ggml_backend_dev_type ggml_backend_cuda_device_get_type(ggml_backend
         : GGML_BACKEND_DEVICE_TYPE_GPU;
 }
 
+static bool ggml_backend_cuda_host_buffer_supported() {
+    return !ggml_cuda_pinned_memory_disabled();
+}
+
+static bool ggml_backend_cuda_device_supports_cuda_host_buft(int device) {
+#if defined(GGML_USE_HIP)
+    if (ggml_cuda_info().devices[device].integrated) {
+        return false;
+    }
+#else
+    GGML_UNUSED(device);
+#endif
+
+    return ggml_backend_cuda_host_buffer_supported();
+}
+
 static void ggml_backend_cuda_device_get_props(ggml_backend_dev_t dev, ggml_backend_dev_props * props) {
     ggml_backend_cuda_device_context * ctx = (ggml_backend_cuda_device_context *)dev->context;
 
@@ -5412,6 +5428,10 @@ static ggml_backend_buffer_type_t ggml_backend_cuda_device_get_buffer_type(ggml_
 
 static ggml_backend_buffer_type_t ggml_backend_cuda_device_get_host_buffer_type(ggml_backend_dev_t dev) {
     GGML_UNUSED(dev);
+    if (!ggml_backend_cuda_host_buffer_supported()) {
+        return nullptr;
+    }
+
     return ggml_backend_cuda_host_buffer_type();
 }
 
@@ -5887,7 +5907,10 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
 static bool ggml_backend_cuda_device_supports_buft(ggml_backend_dev_t dev, ggml_backend_buffer_type_t buft) {
     ggml_backend_cuda_device_context * dev_ctx = (ggml_backend_cuda_device_context *) dev->context;
     const bool integrated = ggml_cuda_info().devices[dev_ctx->device].integrated;
-    return (ggml_backend_buft_is_cuda(buft) && buft->device == dev) || (integrated && ggml_backend_buft_is_cuda_host(buft));
+    return (ggml_backend_buft_is_cuda(buft) && buft->device == dev) ||
+        (integrated &&
+         ggml_backend_buft_is_cuda_host(buft) &&
+         ggml_backend_cuda_device_supports_cuda_host_buft(dev_ctx->device));
 }
 
 static int64_t get_op_batch_size(const ggml_tensor * op) {