diff --git a/sys/dev/drm/amd/amdgpu/amdgpu_ttm.c b/sys/dev/drm/amd/amdgpu/amdgpu_ttm.c
--- a/sys/dev/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/sys/dev/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2180,9 +2180,14 @@
 	mm_node = bo->tbo.mem.mm_node;
 	num_loops = 0;
 	while (num_pages) {
-		uint32_t byte_count = mm_node->size << PAGE_SHIFT;
+		uint64_t byte_count = (uint64_t)mm_node->size << PAGE_SHIFT;
+		uint32_t bc32 = byte_count > max_bytes ? max_bytes : (uint32_t)byte_count;
 
-		num_loops += DIV_ROUND_UP(byte_count, max_bytes);
+		/* mm_node->size << PAGE_SHIFT can exceed UINT32_MAX; never
+		 * report zero bytes to DIV_ROUND_UP or the VRAM clear is
+		 * skipped entirely (DF-1332). */
+		num_loops += DIV_ROUND_UP(bc32, max_bytes) +
+		    (byte_count > UINT32_MAX ? byte_count / max_bytes : 0);
 		num_pages -= mm_node->size;
 		++mm_node;
 	}
@@ -2208,12 +2213,13 @@
 	mm_node = bo->tbo.mem.mm_node;
 
 	while (num_pages) {
-		uint32_t byte_count = mm_node->size << PAGE_SHIFT;
+		uint64_t byte_count = (uint64_t)mm_node->size << PAGE_SHIFT;
 		uint64_t dst_addr;
 
 		dst_addr = amdgpu_mm_node_addr(&bo->tbo, mm_node, &bo->tbo.mem);
 		while (byte_count) {
-			uint32_t cur_size_in_bytes = min(byte_count, max_bytes);
+			uint32_t cur_size_in_bytes = byte_count > max_bytes ?
+			    max_bytes : (uint32_t)byte_count;
 
 			amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data,
 						dst_addr, cur_size_in_bytes);