diff --git a/rpcs3/Emu/RSX/VK/VKRenderPass.cpp b/rpcs3/Emu/RSX/VK/VKRenderPass.cpp
index 4bb2a477bf..79b90a836c 100644
--- a/rpcs3/Emu/RSX/VK/VKRenderPass.cpp
+++ b/rpcs3/Emu/RSX/VK/VKRenderPass.cpp
@@ -194,28 +194,6 @@ namespace vk
 			attachment_references.push_back({ attachment_count, dsv_layout });
 		}
 
-		VkSubpassDependency null_subpass_dependencies[2] =
-		{
-			{
-				.srcSubpass = VK_SUBPASS_EXTERNAL,
-				.dstSubpass = 0,
-				.srcStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
-				.dstStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
-				.srcAccessMask = 0,
-				.dstAccessMask = 0,
-				.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
-			},
-			{
-				.srcSubpass = 0,
-				.dstSubpass = VK_SUBPASS_EXTERNAL,
-				.srcStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
-				.dstStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
-				.srcAccessMask = 0,
-				.dstAccessMask = 0,
-				.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
-			}
-		};
-
 		VkSubpassDescription subpass = {};
 		subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
 		subpass.colorAttachmentCount = attachment_count;
@@ -229,19 +207,6 @@ namespace vk
 		rp_info.subpassCount = 1;
 		rp_info.pSubpasses = &subpass;
 
-		if (vk::get_driver_vendor() == vk::driver_vendor::RADV)
-		{
-			// So, according to spec, if you do not explicitly define the relationship between a subpass and its neighbours,
-			// a full pipeline barrier will be inserted for you, which is awful for performance.
-			// However, only RADV seems to do this and it does not work for other vendors.
-			// NVIDIA specifically slows to a crawl even with a TOP_OF_PIPE->TOP_OF_PIPE dependency with no dependent access.
-			// RPCS3 does not actually want to declare any dependency here, we have explicit pipeline barriers for our tasks.
-			// Workaround: Only provide this null dep chain for RADV
-
-			rp_info.dependencyCount = 2;
-			rp_info.pDependencies = null_subpass_dependencies;
-		}
-
 		VkRenderPass result;
 		CHECK_RESULT(vkCreateRenderPass(dev, &rp_info, NULL, &result));