14 files changed, 843 insertions, 210 deletions

diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
index e62b36822..3d328a250 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
@@ -48,43 +48,30 @@ void RefreshXfbState(VideoCommon::TransformFeedbackState& state, const Maxwell&
 }
 } // Anonymous namespace
 
-void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
-                                 bool has_extended_dynamic_state, bool has_dynamic_vertex_input) {
+void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFeatures& features) {
     const Maxwell& regs = maxwell3d.regs;
     const auto topology_ = maxwell3d.draw_manager->GetDrawState().topology;
-    const std::array enabled_lut{
-        regs.polygon_offset_point_enable,
-        regs.polygon_offset_line_enable,
-        regs.polygon_offset_fill_enable,
-    };
-    const u32 topology_index = static_cast<u32>(topology_);
 
     raw1 = 0;
-    extended_dynamic_state.Assign(has_extended_dynamic_state ? 1 : 0);
-    dynamic_vertex_input.Assign(has_dynamic_vertex_input ? 1 : 0);
+    extended_dynamic_state.Assign(features.has_extended_dynamic_state ? 1 : 0);
+    extended_dynamic_state_2.Assign(features.has_extended_dynamic_state_2 ? 1 : 0);
+    extended_dynamic_state_2_extra.Assign(features.has_extended_dynamic_state_2_extra ? 1 : 0);
+    extended_dynamic_state_3_blend.Assign(features.has_extended_dynamic_state_3_blend ? 1 : 0);
+    extended_dynamic_state_3_enables.Assign(features.has_extended_dynamic_state_3_enables ? 1 : 0);
+    dynamic_vertex_input.Assign(features.has_dynamic_vertex_input ? 1 : 0);
     xfb_enabled.Assign(regs.transform_feedback_enabled != 0);
-    primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0);
-    depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
-    depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
-                                regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
-                                regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
     ndc_minus_one_to_one.Assign(regs.depth_mode == Maxwell::DepthMode::MinusOneToOne ? 1 : 0);
     polygon_mode.Assign(PackPolygonMode(regs.polygon_mode_front));
-    patch_control_points_minus_one.Assign(regs.patch_vertices - 1);
     tessellation_primitive.Assign(static_cast<u32>(regs.tessellation.params.domain_type.Value()));
     tessellation_spacing.Assign(static_cast<u32>(regs.tessellation.params.spacing.Value()));
     tessellation_clockwise.Assign(regs.tessellation.params.output_primitives.Value() ==
                                   Maxwell::Tessellation::OutputPrimitives::Triangles_CW);
-    logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
-    logic_op.Assign(PackLogicOp(regs.logic_op.op));
+    patch_control_points_minus_one.Assign(regs.patch_vertices - 1);
     topology.Assign(topology_);
     msaa_mode.Assign(regs.anti_alias_samples_mode);
 
     raw2 = 0;
-    rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0);
+
     const auto test_func =
         regs.alpha_test_enabled != 0 ? regs.alpha_test_func : Maxwell::ComparisonOp::Always_GL;
     alpha_test_func.Assign(PackComparisonOp(test_func));
@@ -97,6 +84,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
     smooth_lines.Assign(regs.line_anti_alias_enable != 0 ? 1 : 0);
     alpha_to_coverage_enabled.Assign(regs.anti_alias_alpha_control.alpha_to_coverage != 0 ? 1 : 0);
     alpha_to_one_enabled.Assign(regs.anti_alias_alpha_control.alpha_to_one != 0 ? 1 : 0);
+    app_stage.Assign(maxwell3d.engine_state);
 
     for (size_t i = 0; i < regs.rt.size(); ++i) {
         color_formats[i] = static_cast<u8>(regs.rt[i].format);
@@ -105,7 +93,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
     point_size = Common::BitCast<u32>(regs.point_size);
 
     if (maxwell3d.dirty.flags[Dirty::VertexInput]) {
-        if (has_dynamic_vertex_input) {
+        if (features.has_dynamic_vertex_input) {
             // Dirty flag will be reset by the command buffer update
             static constexpr std::array LUT{
                 0u, // Invalid
@@ -144,12 +132,6 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
             }
         }
     }
-    if (maxwell3d.dirty.flags[Dirty::Blending]) {
-        maxwell3d.dirty.flags[Dirty::Blending] = false;
-        for (size_t index = 0; index < attachments.size(); ++index) {
-            attachments[index].Refresh(regs, index);
-        }
-    }
     if (maxwell3d.dirty.flags[Dirty::ViewportSwizzles]) {
         maxwell3d.dirty.flags[Dirty::ViewportSwizzles] = false;
         const auto& transform = regs.viewport_transform;
@@ -157,8 +139,27 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
             return static_cast<u16>(viewport.swizzle.raw);
         });
     }
+    dynamic_state.raw1 = 0;
+    dynamic_state.raw2 = 0;
     if (!extended_dynamic_state) {
         dynamic_state.Refresh(regs);
+        std::ranges::transform(regs.vertex_streams, vertex_strides.begin(), [](const auto& array) {
+            return static_cast<u16>(array.stride.Value());
+        });
+    }
+    if (!extended_dynamic_state_2_extra) {
+        dynamic_state.Refresh2(regs, topology, extended_dynamic_state_2);
+    }
+    if (!extended_dynamic_state_3_blend) {
+        if (maxwell3d.dirty.flags[Dirty::Blending]) {
+            maxwell3d.dirty.flags[Dirty::Blending] = false;
+            for (size_t index = 0; index < attachments.size(); ++index) {
+                attachments[index].Refresh(regs, index);
+            }
+        }
+    }
+    if (!extended_dynamic_state_3_enables) {
+        dynamic_state.Refresh3(regs);
     }
     if (xfb_enabled) {
         RefreshXfbState(xfb_state, regs);
@@ -175,12 +176,11 @@ void FixedPipelineState::BlendingAttachment::Refresh(const Maxwell& regs, size_t
     mask_a.Assign(mask.A);
 
     // TODO: C++20 Use templated lambda to deduplicate code
+    if (!regs.blend.enable[index]) {
+        return;
+    }
 
-    if (!regs.blend_per_target_enabled) {
-        if (!regs.blend.enable[index]) {
-            return;
-        }
-        const auto& src = regs.blend;
+    const auto setup_blend = [&]<typename T>(const T& src) {
         equation_rgb.Assign(PackBlendEquation(src.color_op));
         equation_a.Assign(PackBlendEquation(src.alpha_op));
         factor_source_rgb.Assign(PackBlendFactor(src.color_source));
@@ -188,20 +188,13 @@ void FixedPipelineState::BlendingAttachment::Refresh(const Maxwell& regs, size_t
         factor_source_a.Assign(PackBlendFactor(src.alpha_source));
         factor_dest_a.Assign(PackBlendFactor(src.alpha_dest));
         enable.Assign(1);
-        return;
-    }
+    };
 
-    if (!regs.blend.enable[index]) {
+    if (!regs.blend_per_target_enabled) {
+        setup_blend(regs.blend);
         return;
     }
-    const auto& src = regs.blend_per_target[index];
-    equation_rgb.Assign(PackBlendEquation(src.color_op));
-    equation_a.Assign(PackBlendEquation(src.alpha_op));
-    factor_source_rgb.Assign(PackBlendFactor(src.color_source));
-    factor_dest_rgb.Assign(PackBlendFactor(src.color_dest));
-    factor_source_a.Assign(PackBlendFactor(src.alpha_source));
-    factor_dest_a.Assign(PackBlendFactor(src.alpha_dest));
-    enable.Assign(1);
+    setup_blend(regs.blend_per_target[index]);
 }
 
 void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
@@ -211,8 +204,6 @@ void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
         packed_front_face = 1 - packed_front_face;
     }
 
-    raw1 = 0;
-    raw2 = 0;
     front.action_stencil_fail.Assign(PackStencilOp(regs.stencil_front_op.fail));
     front.action_depth_fail.Assign(PackStencilOp(regs.stencil_front_op.zfail));
     front.action_depth_pass.Assign(PackStencilOp(regs.stencil_front_op.zpass));
@@ -236,9 +227,37 @@ void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
     depth_test_func.Assign(PackComparisonOp(regs.depth_test_func));
     cull_face.Assign(PackCullFace(regs.gl_cull_face));
     cull_enable.Assign(regs.gl_cull_test_enabled != 0 ? 1 : 0);
-    std::ranges::transform(regs.vertex_streams, vertex_strides.begin(), [](const auto& array) {
-        return static_cast<u16>(array.stride.Value());
-    });
+}
+
+void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs,
+                                                Maxwell::PrimitiveTopology topology_,
+                                                bool base_feautures_supported) {
+    logic_op.Assign(PackLogicOp(regs.logic_op.op));
+
+    if (base_feautures_supported) {
+        return;
+    }
+
+    const std::array enabled_lut{
+        regs.polygon_offset_point_enable,
+        regs.polygon_offset_line_enable,
+        regs.polygon_offset_fill_enable,
+    };
+    const u32 topology_index = static_cast<u32>(topology_);
+
+    rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0);
+    primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0);
+    depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
+}
+
+void FixedPipelineState::DynamicState::Refresh3(const Maxwell& regs) {
+    logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
+    depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
+                                    Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
+                                regs.viewport_clip_control.geometry_clip ==
+                                    Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
+                                regs.viewport_clip_control.geometry_clip ==
+                                    Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
 }
 
 size_t FixedPipelineState::Hash() const noexcept {
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
index ab79fb8f3..98ea20b42 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
@@ -17,6 +17,15 @@ namespace Vulkan {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
+struct DynamicFeatures {
+    bool has_extended_dynamic_state;
+    bool has_extended_dynamic_state_2;
+    bool has_extended_dynamic_state_2_extra;
+    bool has_extended_dynamic_state_3_blend;
+    bool has_extended_dynamic_state_3_enables;
+    bool has_dynamic_vertex_input;
+};
+
 struct FixedPipelineState {
     static u32 PackComparisonOp(Maxwell::ComparisonOp op) noexcept;
     static Maxwell::ComparisonOp UnpackComparisonOp(u32 packed) noexcept;
@@ -133,6 +142,17 @@ struct FixedPipelineState {
     struct DynamicState {
         union {
             u32 raw1;
+            BitField<0, 2, u32> cull_face;
+            BitField<2, 1, u32> cull_enable;
+            BitField<3, 1, u32> primitive_restart_enable;
+            BitField<4, 1, u32> depth_bias_enable;
+            BitField<5, 1, u32> rasterize_enable;
+            BitField<6, 4, u32> logic_op;
+            BitField<10, 1, u32> logic_op_enable;
+            BitField<11, 1, u32> depth_clamp_disabled;
+        };
+        union {
+            u32 raw2;
             StencilFace<0> front;
             StencilFace<12> back;
             BitField<24, 1, u32> stencil_enable;
@@ -142,15 +162,11 @@ struct FixedPipelineState {
             BitField<28, 1, u32> front_face;
             BitField<29, 3, u32> depth_test_func;
         };
-        union {
-            u32 raw2;
-            BitField<0, 2, u32> cull_face;
-            BitField<2, 1, u32> cull_enable;
-        };
-        // Vertex stride is a 12 bits value, we have 4 bits to spare per element
-        std::array<u16, Maxwell::NumVertexArrays> vertex_strides;
 
         void Refresh(const Maxwell& regs);
+        void Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology,
+                      bool base_feautures_supported);
+        void Refresh3(const Maxwell& regs);
 
         Maxwell::ComparisonOp DepthTestFunc() const noexcept {
             return UnpackComparisonOp(depth_test_func);
@@ -168,25 +184,24 @@ struct FixedPipelineState {
     union {
         u32 raw1;
         BitField<0, 1, u32> extended_dynamic_state;
-        BitField<1, 1, u32> dynamic_vertex_input;
-        BitField<2, 1, u32> xfb_enabled;
-        BitField<3, 1, u32> primitive_restart_enable;
-        BitField<4, 1, u32> depth_bias_enable;
-        BitField<5, 1, u32> depth_clamp_disabled;
-        BitField<6, 1, u32> ndc_minus_one_to_one;
-        BitField<7, 2, u32> polygon_mode;
-        BitField<9, 5, u32> patch_control_points_minus_one;
-        BitField<14, 2, u32> tessellation_primitive;
-        BitField<16, 2, u32> tessellation_spacing;
-        BitField<18, 1, u32> tessellation_clockwise;
-        BitField<19, 1, u32> logic_op_enable;
-        BitField<20, 4, u32> logic_op;
+        BitField<1, 1, u32> extended_dynamic_state_2;
+        BitField<2, 1, u32> extended_dynamic_state_2_extra;
+        BitField<3, 1, u32> extended_dynamic_state_3_blend;
+        BitField<4, 1, u32> extended_dynamic_state_3_enables;
+        BitField<5, 1, u32> dynamic_vertex_input;
+        BitField<6, 1, u32> xfb_enabled;
+        BitField<7, 1, u32> ndc_minus_one_to_one;
+        BitField<8, 2, u32> polygon_mode;
+        BitField<10, 2, u32> tessellation_primitive;
+        BitField<12, 2, u32> tessellation_spacing;
+        BitField<14, 1, u32> tessellation_clockwise;
+        BitField<15, 5, u32> patch_control_points_minus_one;
+
         BitField<24, 4, Maxwell::PrimitiveTopology> topology;
         BitField<28, 4, Tegra::Texture::MsaaMode> msaa_mode;
     };
     union {
         u32 raw2;
-        BitField<0, 1, u32> rasterize_enable;
         BitField<1, 3, u32> alpha_test_func;
         BitField<4, 1, u32> early_z;
         BitField<5, 1, u32> depth_enabled;
@@ -197,25 +212,28 @@ struct FixedPipelineState {
         BitField<14, 1, u32> smooth_lines;
         BitField<15, 1, u32> alpha_to_coverage_enabled;
         BitField<16, 1, u32> alpha_to_one_enabled;
+        BitField<17, 3, Tegra::Engines::Maxwell3D::EngineHint> app_stage;
     };
     std::array<u8, Maxwell::NumRenderTargets> color_formats;
 
     u32 alpha_test_ref;
     u32 point_size;
-    std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
     std::array<u16, Maxwell::NumViewports> viewport_swizzles;
     union {
         u64 attribute_types; // Used with VK_EXT_vertex_input_dynamic_state
         u64 enabled_divisors;
     };
+
+    DynamicState dynamic_state;
+    std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
     std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes;
     std::array<u32, Maxwell::NumVertexArrays> binding_divisors;
+    // Vertex stride is a 12 bits value, we have 4 bits to spare per element
+    std::array<u16, Maxwell::NumVertexArrays> vertex_strides;
 
-    DynamicState dynamic_state;
     VideoCommon::TransformFeedbackState xfb_state;
 
-    void Refresh(Tegra::Engines::Maxwell3D& maxwell3d, bool has_extended_dynamic_state,
-                 bool has_dynamic_vertex_input);
+    void Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFeatures& features);
 
     size_t Hash() const noexcept;
 
@@ -230,13 +248,17 @@ struct FixedPipelineState {
             // When transform feedback is enabled, use the whole struct
             return sizeof(*this);
         }
-        if (dynamic_vertex_input) {
+        if (dynamic_vertex_input && extended_dynamic_state_3_blend) {
             // Exclude dynamic state and attributes
+            return offsetof(FixedPipelineState, dynamic_state);
+        }
+        if (dynamic_vertex_input) {
+            // Exclude dynamic state
             return offsetof(FixedPipelineState, attributes);
         }
         if (extended_dynamic_state) {
             // Exclude dynamic state
-            return offsetof(FixedPipelineState, dynamic_state);
+            return offsetof(FixedPipelineState, vertex_strides);
         }
         // Default
         return offsetof(FixedPipelineState, xfb_state);
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
index 6b54d7111..487d8b416 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
@@ -56,7 +56,8 @@ vk::Buffer CreateBuffer(const Device& device, u64 size) {
         VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
         VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT |
         VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
-        VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
+        VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
+        VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
     if (device.IsExtTransformFeedbackSupported()) {
         flags |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
     }
@@ -516,6 +517,7 @@ void BufferCacheRuntime::ReserveNullBuffer() {
     if (device.IsExtTransformFeedbackSupported()) {
         create_info.usage |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
     }
+    create_info.usage |= VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
     null_buffer = device.GetLogical().CreateBuffer(create_info);
     if (device.HasDebuggingToolAttached()) {
         null_buffer.SetObjectNameEXT("Null buffer");
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index 515d8d869..d11383bf1 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -201,6 +201,22 @@ struct SimpleVertexSpec {
     static constexpr bool has_images = false;
 };
 
+struct SimpleStorageSpec {
+    static constexpr std::array<bool, 5> enabled_stages{true, false, false, false, true};
+    static constexpr bool has_storage_buffers = true;
+    static constexpr bool has_texture_buffers = false;
+    static constexpr bool has_image_buffers = false;
+    static constexpr bool has_images = false;
+};
+
+struct SimpleImageSpec {
+    static constexpr std::array<bool, 5> enabled_stages{true, false, false, false, true};
+    static constexpr bool has_storage_buffers = false;
+    static constexpr bool has_texture_buffers = false;
+    static constexpr bool has_image_buffers = false;
+    static constexpr bool has_images = true;
+};
+
 struct DefaultSpec {
     static constexpr std::array<bool, 5> enabled_stages{true, true, true, true, true};
     static constexpr bool has_storage_buffers = true;
@@ -211,7 +227,8 @@ struct DefaultSpec {
 
 ConfigureFuncPtr ConfigureFunc(const std::array<vk::ShaderModule, NUM_STAGES>& modules,
                                const std::array<Shader::Info, NUM_STAGES>& infos) {
-    return FindSpec<SimpleVertexSpec, SimpleVertexFragmentSpec, DefaultSpec>(modules, infos);
+    return FindSpec<SimpleVertexSpec, SimpleVertexFragmentSpec, SimpleStorageSpec, SimpleImageSpec,
+                    DefaultSpec>(modules, infos);
 }
 } // Anonymous namespace
 
@@ -524,6 +541,8 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
     FixedPipelineState::DynamicState dynamic{};
     if (!key.state.extended_dynamic_state) {
         dynamic = key.state.dynamic_state;
+    } else {
+        dynamic.raw1 = key.state.dynamic_state.raw1;
     }
     static_vector<VkVertexInputBindingDescription, 32> vertex_bindings;
     static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors;
@@ -561,7 +580,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
                 instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
             vertex_bindings.push_back({
                 .binding = static_cast<u32>(index),
-                .stride = dynamic.vertex_strides[index],
+                .stride = key.state.vertex_strides[index],
                 .inputRate = rate,
             });
             if (instanced) {
@@ -625,7 +644,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
         .pNext = nullptr,
         .flags = 0,
         .topology = input_assembly_topology,
-        .primitiveRestartEnable = key.state.primitive_restart_enable != 0 &&
+        .primitiveRestartEnable = dynamic.primitive_restart_enable != 0 &&
                                   ((input_assembly_topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
                                     device.IsTopologyListPrimitiveRestartSupported()) ||
                                    SupportsPrimitiveRestart(input_assembly_topology) ||
@@ -672,15 +691,15 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
         .pNext = nullptr,
         .flags = 0,
         .depthClampEnable =
-            static_cast<VkBool32>(key.state.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE),
+            static_cast<VkBool32>(dynamic.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE),
         .rasterizerDiscardEnable =
-            static_cast<VkBool32>(key.state.rasterize_enable == 0 ? VK_TRUE : VK_FALSE),
+            static_cast<VkBool32>(dynamic.rasterize_enable == 0 ? VK_TRUE : VK_FALSE),
         .polygonMode =
             MaxwellToVK::PolygonMode(FixedPipelineState::UnpackPolygonMode(key.state.polygon_mode)),
         .cullMode = static_cast<VkCullModeFlags>(
             dynamic.cull_enable ? MaxwellToVK::CullFace(dynamic.CullFace()) : VK_CULL_MODE_NONE),
         .frontFace = MaxwellToVK::FrontFace(dynamic.FrontFace()),
-        .depthBiasEnable = key.state.depth_bias_enable,
+        .depthBiasEnable = (dynamic.depth_bias_enable == 0 ? VK_TRUE : VK_FALSE),
         .depthBiasConstantFactor = 0.0f,
         .depthBiasClamp = 0.0f,
         .depthBiasSlopeFactor = 0.0f,
@@ -782,13 +801,13 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
         .pNext = nullptr,
         .flags = 0,
-        .logicOpEnable = key.state.logic_op_enable != 0,
-        .logicOp = static_cast<VkLogicOp>(key.state.logic_op.Value()),
+        .logicOpEnable = dynamic.logic_op_enable != 0,
+        .logicOp = static_cast<VkLogicOp>(dynamic.logic_op.Value()),
         .attachmentCount = static_cast<u32>(cb_attachments.size()),
         .pAttachments = cb_attachments.data(),
         .blendConstants = {},
     };
-    static_vector<VkDynamicState, 19> dynamic_states{
+    static_vector<VkDynamicState, 28> dynamic_states{
         VK_DYNAMIC_STATE_VIEWPORT,           VK_DYNAMIC_STATE_SCISSOR,
         VK_DYNAMIC_STATE_DEPTH_BIAS,         VK_DYNAMIC_STATE_BLEND_CONSTANTS,
         VK_DYNAMIC_STATE_DEPTH_BOUNDS,       VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
@@ -811,6 +830,32 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
             dynamic_states.push_back(VK_DYNAMIC_STATE_VERTEX_INPUT_EXT);
         }
         dynamic_states.insert(dynamic_states.end(), extended.begin(), extended.end());
+        if (key.state.extended_dynamic_state_2) {
+            static constexpr std::array extended2{
+                VK_DYNAMIC_STATE_DEPTH_BIAS_ENABLE_EXT,
+                VK_DYNAMIC_STATE_PRIMITIVE_RESTART_ENABLE_EXT,
+                VK_DYNAMIC_STATE_RASTERIZER_DISCARD_ENABLE_EXT,
+            };
+            dynamic_states.insert(dynamic_states.end(), extended2.begin(), extended2.end());
+        }
+        if (key.state.extended_dynamic_state_2_extra) {
+            dynamic_states.push_back(VK_DYNAMIC_STATE_LOGIC_OP_EXT);
+        }
+        if (key.state.extended_dynamic_state_3_blend) {
+            static constexpr std::array extended3{
+                VK_DYNAMIC_STATE_COLOR_BLEND_ENABLE_EXT,
+                VK_DYNAMIC_STATE_COLOR_BLEND_EQUATION_EXT,
+                VK_DYNAMIC_STATE_COLOR_WRITE_MASK_EXT,
+            };
+            dynamic_states.insert(dynamic_states.end(), extended3.begin(), extended3.end());
+        }
+        if (key.state.extended_dynamic_state_3_enables) {
+            static constexpr std::array extended3{
+                VK_DYNAMIC_STATE_DEPTH_CLAMP_ENABLE_EXT,
+                VK_DYNAMIC_STATE_LOGIC_OP_ENABLE_EXT,
+            };
+            dynamic_states.insert(dynamic_states.end(), extended3.begin(), extended3.end());
+        }
     }
     const VkPipelineDynamicStateCreateInfo dynamic_state_ci{
         .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index e7262420c..3046b72ab 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -54,7 +54,7 @@ using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 8;
+constexpr u32 CACHE_VERSION = 10;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -351,6 +351,15 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         LOG_WARNING(Render_Vulkan, "maxVertexInputBindings is too low: {} < {}",
                     device.GetMaxVertexInputBindings(), Maxwell::NumVertexArrays);
     }
+
+    dynamic_features = DynamicFeatures{
+        .has_extended_dynamic_state = device.IsExtExtendedDynamicStateSupported(),
+        .has_extended_dynamic_state_2 = device.IsExtExtendedDynamicState2Supported(),
+        .has_extended_dynamic_state_2_extra = device.IsExtExtendedDynamicState2ExtrasSupported(),
+        .has_extended_dynamic_state_3_blend = device.IsExtExtendedDynamicState3BlendingSupported(),
+        .has_extended_dynamic_state_3_enables = device.IsExtExtendedDynamicState3EnablesSupported(),
+        .has_dynamic_vertex_input = device.IsExtVertexInputDynamicStateSupported(),
+    };
 }
 
 PipelineCache::~PipelineCache() = default;
@@ -362,8 +371,7 @@ GraphicsPipeline* PipelineCache::CurrentGraphicsPipeline() {
         current_pipeline = nullptr;
         return nullptr;
     }
-    graphics_key.state.Refresh(*maxwell3d, device.IsExtExtendedDynamicStateSupported(),
-                               device.IsExtVertexInputDynamicStateSupported());
+    graphics_key.state.Refresh(*maxwell3d, dynamic_features);
 
     if (current_pipeline) {
         GraphicsPipeline* const next{current_pipeline->Next(graphics_key)};
@@ -439,14 +447,21 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
         });
         ++state.total;
     }};
-    const bool extended_dynamic_state = device.IsExtExtendedDynamicStateSupported();
-    const bool dynamic_vertex_input = device.IsExtVertexInputDynamicStateSupported();
     const auto load_graphics{[&](std::ifstream& file, std::vector<FileEnvironment> envs) {
         GraphicsPipelineCacheKey key;
         file.read(reinterpret_cast<char*>(&key), sizeof(key));
 
-        if ((key.state.extended_dynamic_state != 0) != extended_dynamic_state ||
-            (key.state.dynamic_vertex_input != 0) != dynamic_vertex_input) {
+        if ((key.state.extended_dynamic_state != 0) !=
+                dynamic_features.has_extended_dynamic_state ||
+            (key.state.extended_dynamic_state_2 != 0) !=
+                dynamic_features.has_extended_dynamic_state_2 ||
+            (key.state.extended_dynamic_state_2_extra != 0) !=
+                dynamic_features.has_extended_dynamic_state_2_extra ||
+            (key.state.extended_dynamic_state_3_blend != 0) !=
+                dynamic_features.has_extended_dynamic_state_3_blend ||
+            (key.state.extended_dynamic_state_3_enables != 0) !=
+                dynamic_features.has_extended_dynamic_state_3_enables ||
+            (key.state.dynamic_vertex_input != 0) != dynamic_features.has_dynamic_vertex_input) {
             return;
         }
         workers.QueueWork([this, key, envs = std::move(envs), &state, &callback]() mutable {
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index 61f9e9366..b4f593ef5 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -160,6 +160,7 @@ private:
 
     Common::ThreadWorker workers;
     Common::ThreadWorker serialization_thread;
+    DynamicFeatures dynamic_features;
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index ac1eb9895..242bf9602 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -180,7 +180,8 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
 
 RasterizerVulkan::~RasterizerVulkan() = default;
 
-void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
+template <typename Func>
+void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
     MICROPROFILE_SCOPE(Vulkan_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -201,20 +202,67 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
 
     UpdateDynamicStates();
 
-    const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
-    const u32 num_instances{instance_count};
-    const DrawParams draw_params{MakeDrawParams(draw_state, num_instances, is_indexed)};
-    scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
-        if (draw_params.is_indexed) {
-            cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
-                               draw_params.first_index, draw_params.base_vertex,
-                               draw_params.base_instance);
-        } else {
-            cmdbuf.Draw(draw_params.num_vertices, draw_params.num_instances,
-                        draw_params.base_vertex, draw_params.base_instance);
+    draw_func();
+
+    EndTransformFeedback();
+}
+
+void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
+    PrepareDraw(is_indexed, [this, is_indexed, instance_count] {
+        const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
+        const u32 num_instances{instance_count};
+        const DrawParams draw_params{MakeDrawParams(draw_state, num_instances, is_indexed)};
+        scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
+            if (draw_params.is_indexed) {
+                cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
+                                   draw_params.first_index, draw_params.base_vertex,
+                                   draw_params.base_instance);
+            } else {
+                cmdbuf.Draw(draw_params.num_vertices, draw_params.num_instances,
+                            draw_params.base_vertex, draw_params.base_instance);
+            }
+        });
+    });
+}
+
+void RasterizerVulkan::DrawIndirect() {
+    const auto& params = maxwell3d->draw_manager->GetIndirectParams();
+    buffer_cache.SetDrawIndirect(&params);
+    PrepareDraw(params.is_indexed, [this, &params] {
+        const auto indirect_buffer = buffer_cache.GetDrawIndirectBuffer();
+        const auto& buffer = indirect_buffer.first;
+        const auto& offset = indirect_buffer.second;
+        if (params.include_count) {
+            const auto count = buffer_cache.GetDrawIndirectCount();
+            const auto& draw_buffer = count.first;
+            const auto& offset_base = count.second;
+            scheduler.Record([draw_buffer_obj = draw_buffer->Handle(),
+                              buffer_obj = buffer->Handle(), offset_base, offset,
+                              params](vk::CommandBuffer cmdbuf) {
+                if (params.is_indexed) {
+                    cmdbuf.DrawIndexedIndirectCount(
+                        buffer_obj, offset, draw_buffer_obj, offset_base,
+                        static_cast<u32>(params.max_draw_counts), static_cast<u32>(params.stride));
+                } else {
+                    cmdbuf.DrawIndirectCount(buffer_obj, offset, draw_buffer_obj, offset_base,
+                                             static_cast<u32>(params.max_draw_counts),
+                                             static_cast<u32>(params.stride));
+                }
+            });
+            return;
         }
+        scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf) {
+            if (params.is_indexed) {
+                cmdbuf.DrawIndexedIndirect(buffer_obj, offset,
+                                           static_cast<u32>(params.max_draw_counts),
+                                           static_cast<u32>(params.stride));
+            } else {
+                cmdbuf.DrawIndirect(buffer_obj, offset, static_cast<u32>(params.max_draw_counts),
+                                    static_cast<u32>(params.stride));
+            }
+        });
     });
-    EndTransformFeedback();
+    buffer_cache.SetDrawIndirect(nullptr);
 }
 
 void RasterizerVulkan::Clear(u32 layer_count) {
@@ -379,44 +427,58 @@ void Vulkan::RasterizerVulkan::DisableGraphicsUniformBuffer(size_t stage, u32 in
 
 void RasterizerVulkan::FlushAll() {}
 
-void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) {
+void RasterizerVulkan::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
     }
-    {
+    if (True(which & VideoCommon::CacheType::TextureCache)) {
         std::scoped_lock lock{texture_cache.mutex};
         texture_cache.DownloadMemory(addr, size);
     }
-    {
+    if ((True(which & VideoCommon::CacheType::BufferCache))) {
         std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.DownloadMemory(addr, size);
     }
-    query_cache.FlushRegion(addr, size);
+    if ((True(which & VideoCommon::CacheType::QueryCache))) {
+        query_cache.FlushRegion(addr, size);
+    }
 }
 
-bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size) {
-    std::scoped_lock lock{texture_cache.mutex, buffer_cache.mutex};
+bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+    if ((True(which & VideoCommon::CacheType::BufferCache))) {
+        std::scoped_lock lock{buffer_cache.mutex};
+        if (buffer_cache.IsRegionGpuModified(addr, size)) {
+            return true;
+        }
+    }
     if (!Settings::IsGPULevelHigh()) {
-        return buffer_cache.IsRegionGpuModified(addr, size);
+        return false;
+    }
+    if (True(which & VideoCommon::CacheType::TextureCache)) {
+        std::scoped_lock lock{texture_cache.mutex};
+        return texture_cache.IsRegionGpuModified(addr, size);
     }
-    return texture_cache.IsRegionGpuModified(addr, size) ||
-           buffer_cache.IsRegionGpuModified(addr, size);
+    return false;
 }
 
-void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
+void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
     }
-    {
+    if (True(which & VideoCommon::CacheType::TextureCache)) {
         std::scoped_lock lock{texture_cache.mutex};
         texture_cache.WriteMemory(addr, size);
     }
-    {
+    if ((True(which & VideoCommon::CacheType::BufferCache))) {
         std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.WriteMemory(addr, size);
     }
-    pipeline_cache.InvalidateRegion(addr, size);
-    query_cache.InvalidateRegion(addr, size);
+    if ((True(which & VideoCommon::CacheType::QueryCache))) {
+        query_cache.InvalidateRegion(addr, size);
+    }
+    if ((True(which & VideoCommon::CacheType::ShaderCache))) {
+        pipeline_cache.InvalidateRegion(addr, size);
+    }
 }
 
 void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
@@ -481,11 +543,12 @@ void RasterizerVulkan::ReleaseFences() {
     fence_manager.WaitPendingFences();
 }
 
-void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size,
+                                                VideoCommon::CacheType which) {
     if (Settings::IsGPULevelExtreme()) {
-        FlushRegion(addr, size);
+        FlushRegion(addr, size, which);
     }
-    InvalidateRegion(addr, size);
+    InvalidateRegion(addr, size, which);
 }
 
 void RasterizerVulkan::WaitForIdle() {
@@ -541,6 +604,21 @@ void RasterizerVulkan::TickFrame() {
     }
 }
 
+bool RasterizerVulkan::AccelerateConditionalRendering() {
+    if (Settings::IsGPULevelHigh()) {
+        // TODO(Blinkhawk): Reimplement Host conditional rendering.
+        return false;
+    }
+    // Medium / Low Hack: stub any checks on queries writen into the buffer cache.
+    const GPUVAddr condition_address{maxwell3d->regs.render_enable.Address()};
+    Maxwell::ReportSemaphore::Compare cmp;
+    if (gpu_memory->IsMemoryDirty(condition_address, sizeof(cmp),
+                                  VideoCommon::CacheType::BufferCache)) {
+        return true;
+    }
+    return false;
+}
+
 bool RasterizerVulkan::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
                                              const Tegra::Engines::Fermi2D::Surface& dst,
                                              const Tegra::Engines::Fermi2D::Config& copy_config) {
@@ -561,7 +639,7 @@ void RasterizerVulkan::AccelerateInlineToMemory(GPUVAddr address, size_t copy_si
     }
     gpu_memory->WriteBlockUnsafe(address, memory.data(), copy_size);
     {
-        std::unique_lock<std::mutex> lock{buffer_cache.mutex};
+        std::unique_lock<std::recursive_mutex> lock{buffer_cache.mutex};
         if (!buffer_cache.InlineMemory(*cpu_addr, copy_size, memory)) {
             buffer_cache.WriteMemory(*cpu_addr, copy_size);
         }
@@ -639,16 +717,35 @@ void RasterizerVulkan::UpdateDynamicStates() {
     UpdateLineWidth(regs);
     if (device.IsExtExtendedDynamicStateSupported()) {
         UpdateCullMode(regs);
-        UpdateDepthBoundsTestEnable(regs);
-        UpdateDepthTestEnable(regs);
-        UpdateDepthWriteEnable(regs);
         UpdateDepthCompareOp(regs);
         UpdateFrontFace(regs);
         UpdateStencilOp(regs);
-        UpdateStencilTestEnable(regs);
+
         if (device.IsExtVertexInputDynamicStateSupported()) {
             UpdateVertexInput(regs);
         }
+
+        if (state_tracker.TouchStateEnable()) {
+            UpdateDepthBoundsTestEnable(regs);
+            UpdateDepthTestEnable(regs);
+            UpdateDepthWriteEnable(regs);
+            UpdateStencilTestEnable(regs);
+            if (device.IsExtExtendedDynamicState2Supported()) {
+                UpdatePrimitiveRestartEnable(regs);
+                UpdateRasterizerDiscardEnable(regs);
+                UpdateDepthBiasEnable(regs);
+            }
+            if (device.IsExtExtendedDynamicState3EnablesSupported()) {
+                UpdateLogicOpEnable(regs);
+                UpdateDepthClampEnable(regs);
+            }
+        }
+        if (device.IsExtExtendedDynamicState2ExtrasSupported()) {
+            UpdateLogicOp(regs);
+        }
+        if (device.IsExtExtendedDynamicState3Supported()) {
+            UpdateBlending(regs);
+        }
     }
 }
 
@@ -789,32 +886,92 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
     if (!state_tracker.TouchStencilProperties()) {
         return;
     }
-    if (regs.stencil_two_side_enable) {
-        // Separate values per face
-        scheduler.Record(
-            [front_ref = regs.stencil_front_ref, front_write_mask = regs.stencil_front_mask,
-             front_test_mask = regs.stencil_front_func_mask, back_ref = regs.stencil_back_ref,
-             back_write_mask = regs.stencil_back_mask,
-             back_test_mask = regs.stencil_back_func_mask](vk::CommandBuffer cmdbuf) {
+    bool update_references = state_tracker.TouchStencilReference();
+    bool update_write_mask = state_tracker.TouchStencilWriteMask();
+    bool update_compare_masks = state_tracker.TouchStencilCompare();
+    if (state_tracker.TouchStencilSide(regs.stencil_two_side_enable != 0)) {
+        update_references = true;
+        update_write_mask = true;
+        update_compare_masks = true;
+    }
+    if (update_references) {
+        [&]() {
+            if (regs.stencil_two_side_enable) {
+                if (!state_tracker.CheckStencilReferenceFront(regs.stencil_front_ref) &&
+                    !state_tracker.CheckStencilReferenceBack(regs.stencil_back_ref)) {
+                    return;
+                }
+            } else {
+                if (!state_tracker.CheckStencilReferenceFront(regs.stencil_front_ref)) {
+                    return;
+                }
+            }
+            scheduler.Record([front_ref = regs.stencil_front_ref, back_ref = regs.stencil_back_ref,
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                const bool set_back = two_sided && front_ref != back_ref;
                 // Front face
-                cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_BIT, front_ref);
-                cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_BIT, front_write_mask);
-                cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_BIT, front_test_mask);
-
-                // Back face
-                cmdbuf.SetStencilReference(VK_STENCIL_FACE_BACK_BIT, back_ref);
-                cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_BACK_BIT, back_write_mask);
-                cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_BACK_BIT, back_test_mask);
+                cmdbuf.SetStencilReference(set_back ? VK_STENCIL_FACE_FRONT_BIT
+                                                    : VK_STENCIL_FACE_FRONT_AND_BACK,
+                                           front_ref);
+                if (set_back) {
+                    cmdbuf.SetStencilReference(VK_STENCIL_FACE_BACK_BIT, back_ref);
+                }
             });
-    } else {
-        // Front face defines both faces
-        scheduler.Record([ref = regs.stencil_front_ref, write_mask = regs.stencil_front_mask,
-                          test_mask = regs.stencil_front_func_mask](vk::CommandBuffer cmdbuf) {
-            cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_AND_BACK, ref);
-            cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_AND_BACK, write_mask);
-            cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_AND_BACK, test_mask);
-        });
+        }();
+    }
+    if (update_write_mask) {
+        [&]() {
+            if (regs.stencil_two_side_enable) {
+                if (!state_tracker.CheckStencilWriteMaskFront(regs.stencil_front_mask) &&
+                    !state_tracker.CheckStencilWriteMaskBack(regs.stencil_back_mask)) {
+                    return;
+                }
+            } else {
+                if (!state_tracker.CheckStencilWriteMaskFront(regs.stencil_front_mask)) {
+                    return;
+                }
+            }
+            scheduler.Record([front_write_mask = regs.stencil_front_mask,
+                              back_write_mask = regs.stencil_back_mask,
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                const bool set_back = two_sided && front_write_mask != back_write_mask;
+                // Front face
+                cmdbuf.SetStencilWriteMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
+                                                    : VK_STENCIL_FACE_FRONT_AND_BACK,
+                                           front_write_mask);
+                if (set_back) {
+                    cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_BACK_BIT, back_write_mask);
+                }
+            });
+        }();
+    }
+    if (update_compare_masks) {
+        [&]() {
+            if (regs.stencil_two_side_enable) {
+                if (!state_tracker.CheckStencilCompareMaskFront(regs.stencil_front_func_mask) &&
+                    !state_tracker.CheckStencilCompareMaskBack(regs.stencil_back_func_mask)) {
+                    return;
+                }
+            } else {
+                if (!state_tracker.CheckStencilCompareMaskFront(regs.stencil_front_func_mask)) {
+                    return;
+                }
+            }
+            scheduler.Record([front_test_mask = regs.stencil_front_func_mask,
+                              back_test_mask = regs.stencil_back_func_mask,
+                              two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
+                const bool set_back = two_sided && front_test_mask != back_test_mask;
+                // Front face
+                cmdbuf.SetStencilCompareMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
+                                                      : VK_STENCIL_FACE_FRONT_AND_BACK,
+                                             front_test_mask);
+                if (set_back) {
+                    cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_BACK_BIT, back_test_mask);
+                }
+            });
+        }();
     }
+    state_tracker.ClearStencilReset();
 }
 
 void RasterizerVulkan::UpdateLineWidth(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -868,6 +1025,82 @@ void RasterizerVulkan::UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& r
     });
 }
 
+void RasterizerVulkan::UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchPrimitiveRestartEnable()) {
+        return;
+    }
+    scheduler.Record([enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf) {
+        cmdbuf.SetPrimitiveRestartEnableEXT(enable);
+    });
+}
+
+void RasterizerVulkan::UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchRasterizerDiscardEnable()) {
+        return;
+    }
+    scheduler.Record([disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf) {
+        cmdbuf.SetRasterizerDiscardEnableEXT(disable == 0);
+    });
+}
+
+void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthBiasEnable()) {
+        return;
+    }
+    constexpr size_t POINT = 0;
+    constexpr size_t LINE = 1;
+    constexpr size_t POLYGON = 2;
+    static constexpr std::array POLYGON_OFFSET_ENABLE_LUT = {
+        POINT,   // Points
+        LINE,    // Lines
+        LINE,    // LineLoop
+        LINE,    // LineStrip
+        POLYGON, // Triangles
+        POLYGON, // TriangleStrip
+        POLYGON, // TriangleFan
+        POLYGON, // Quads
+        POLYGON, // QuadStrip
+        POLYGON, // Polygon
+        LINE,    // LinesAdjacency
+        LINE,    // LineStripAdjacency
+        POLYGON, // TrianglesAdjacency
+        POLYGON, // TriangleStripAdjacency
+        POLYGON, // Patches
+    };
+    const std::array enabled_lut{
+        regs.polygon_offset_point_enable,
+        regs.polygon_offset_line_enable,
+        regs.polygon_offset_fill_enable,
+    };
+    const u32 topology_index = static_cast<u32>(maxwell3d->draw_manager->GetDrawState().topology);
+    const u32 enable = enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]];
+    scheduler.Record(
+        [enable](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBiasEnableEXT(enable != 0); });
+}
+
+void RasterizerVulkan::UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchLogicOpEnable()) {
+        return;
+    }
+    scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf) {
+        cmdbuf.SetLogicOpEnableEXT(enable != 0);
+    });
+}
+
+void RasterizerVulkan::UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthClampEnable()) {
+        return;
+    }
+    bool is_enabled = !(regs.viewport_clip_control.geometry_clip ==
+                            Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
+                        regs.viewport_clip_control.geometry_clip ==
+                            Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
+                        regs.viewport_clip_control.geometry_clip ==
+                            Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
+    scheduler.Record(
+        [is_enabled](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthClampEnableEXT(is_enabled); });
+}
+
 void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchDepthCompareOp()) {
         return;
@@ -925,6 +1158,78 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
     }
 }
 
+void RasterizerVulkan::UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchLogicOp()) {
+        return;
+    }
+    const auto op_value = static_cast<u32>(regs.logic_op.op);
+    auto op = op_value >= 0x1500 && op_value < 0x1510 ? static_cast<VkLogicOp>(op_value - 0x1500)
+                                                      : VK_LOGIC_OP_NO_OP;
+    scheduler.Record([op](vk::CommandBuffer cmdbuf) { cmdbuf.SetLogicOpEXT(op); });
+}
+
+void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchBlending()) {
+        return;
+    }
+
+    if (state_tracker.TouchColorMask()) {
+        std::array<VkColorComponentFlags, Maxwell::NumRenderTargets> setup_masks{};
+        for (size_t index = 0; index < Maxwell::NumRenderTargets; index++) {
+            const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : index];
+            auto& current = setup_masks[index];
+            if (mask.R) {
+                current |= VK_COLOR_COMPONENT_R_BIT;
+            }
+            if (mask.G) {
+                current |= VK_COLOR_COMPONENT_G_BIT;
+            }
+            if (mask.B) {
+                current |= VK_COLOR_COMPONENT_B_BIT;
+            }
+            if (mask.A) {
+                current |= VK_COLOR_COMPONENT_A_BIT;
+            }
+        }
+        scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf) {
+            cmdbuf.SetColorWriteMaskEXT(0, setup_masks);
+        });
+    }
+
+    if (state_tracker.TouchBlendEnable()) {
+        std::array<VkBool32, Maxwell::NumRenderTargets> setup_enables{};
+        std::ranges::transform(
+            regs.blend.enable, setup_enables.begin(),
+            [&](const auto& is_enabled) { return is_enabled != 0 ? VK_TRUE : VK_FALSE; });
+        scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf) {
+            cmdbuf.SetColorBlendEnableEXT(0, setup_enables);
+        });
+    }
+
+    if (state_tracker.TouchBlendEquations()) {
+        std::array<VkColorBlendEquationEXT, Maxwell::NumRenderTargets> setup_blends{};
+        for (size_t index = 0; index < Maxwell::NumRenderTargets; index++) {
+            const auto blend_setup = [&]<typename T>(const T& guest_blend) {
+                auto& host_blend = setup_blends[index];
+                host_blend.srcColorBlendFactor = MaxwellToVK::BlendFactor(guest_blend.color_source);
+                host_blend.dstColorBlendFactor = MaxwellToVK::BlendFactor(guest_blend.color_dest);
+                host_blend.colorBlendOp = MaxwellToVK::BlendEquation(guest_blend.color_op);
+                host_blend.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(guest_blend.alpha_source);
+                host_blend.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(guest_blend.alpha_dest);
+                host_blend.alphaBlendOp = MaxwellToVK::BlendEquation(guest_blend.alpha_op);
+            };
+            if (!regs.blend_per_target_enabled) {
+                blend_setup(regs.blend);
+                continue;
+            }
+            blend_setup(regs.blend_per_target[index]);
+        }
+        scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf) {
+            cmdbuf.SetColorBlendEquationEXT(0, setup_blends);
+        });
+    }
+}
+
 void RasterizerVulkan::UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchStencilTestEnable()) {
         return;
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index ee483cfd9..c661e5b19 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -65,6 +65,7 @@ public:
     ~RasterizerVulkan() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
+    void DrawIndirect() override;
     void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -72,9 +73,12 @@ public:
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
     void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
     void FlushAll() override;
-    void FlushRegion(VAddr addr, u64 size) override;
-    bool MustFlushRegion(VAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushRegion(VAddr addr, u64 size,
+                     VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    bool MustFlushRegion(VAddr addr, u64 size,
+                         VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    void InvalidateRegion(VAddr addr, u64 size,
+                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void OnCPUWrite(VAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
     void UnmapMemory(VAddr addr, u64 size) override;
@@ -84,12 +88,14 @@ public:
     void SignalSyncPoint(u32 value) override;
     void SignalReference() override;
     void ReleaseFences() override;
-    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(
+        VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void WaitForIdle() override;
     void FragmentBarrier() override;
     void TiledCacheBarrier() override;
     void FlushCommands() override;
     void TickFrame() override;
+    bool AccelerateConditionalRendering() override;
     bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
                                const Tegra::Engines::Fermi2D::Surface& dst,
                                const Tegra::Engines::Fermi2D::Config& copy_config) override;
@@ -114,6 +120,9 @@ private:
 
     static constexpr VkDeviceSize DEFAULT_BUFFER_SIZE = 4 * sizeof(float);
 
+    template <typename Func>
+    void PrepareDraw(bool is_indexed, Func&&);
+
     void FlushWork();
 
     void UpdateDynamicStates();
@@ -135,9 +144,16 @@ private:
     void UpdateDepthTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);
     void UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& regs);
     void UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& regs);
     void UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs);
     void UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs);
     void UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs);
 
     void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);
 
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index 06f68d09a..74ca77216 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -1,5 +1,5 @@
-// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
 
 #include <algorithm>
 #include <utility>
@@ -94,7 +94,8 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
         .flags = 0,
         .size = STREAM_BUFFER_SIZE,
         .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
-                 VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
+                 VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
+                 VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT,
         .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
         .queueFamilyIndexCount = 0,
         .pQueueFamilyIndices = nullptr,
@@ -142,11 +143,23 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
 
 StagingBufferPool::~StagingBufferPool() = default;
 
-StagingBufferRef StagingBufferPool::Request(size_t size, MemoryUsage usage) {
-    if (usage == MemoryUsage::Upload && size <= MAX_STREAM_BUFFER_REQUEST_SIZE) {
+StagingBufferRef StagingBufferPool::Request(size_t size, MemoryUsage usage, bool deferred) {
+    if (!deferred && usage == MemoryUsage::Upload && size <= MAX_STREAM_BUFFER_REQUEST_SIZE) {
         return GetStreamBuffer(size);
     }
-    return GetStagingBuffer(size, usage);
+    return GetStagingBuffer(size, usage, deferred);
+}
+
+void StagingBufferPool::FreeDeferred(StagingBufferRef& ref) {
+    auto& entries = GetCache(ref.usage)[ref.log2_level].entries;
+    const auto is_this_one = [&ref](const StagingBuffer& entry) {
+        return entry.index == ref.index;
+    };
+    auto it = std::find_if(entries.begin(), entries.end(), is_this_one);
+    ASSERT(it != entries.end());
+    ASSERT(it->deferred);
+    it->tick = scheduler.CurrentTick();
+    it->deferred = false;
 }
 
 void StagingBufferPool::TickFrame() {
@@ -187,6 +200,9 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
         .buffer = *stream_buffer,
         .offset = static_cast<VkDeviceSize>(offset),
         .mapped_span = std::span<u8>(stream_pointer + offset, size),
+        .usage{},
+        .log2_level{},
+        .index{},
     };
 }
 
@@ -196,19 +212,21 @@ bool StagingBufferPool::AreRegionsActive(size_t region_begin, size_t region_end)
                        [gpu_tick](u64 sync_tick) { return gpu_tick < sync_tick; });
 };
 
-StagingBufferRef StagingBufferPool::GetStagingBuffer(size_t size, MemoryUsage usage) {
-    if (const std::optional<StagingBufferRef> ref = TryGetReservedBuffer(size, usage)) {
+StagingBufferRef StagingBufferPool::GetStagingBuffer(size_t size, MemoryUsage usage,
+                                                     bool deferred) {
+    if (const std::optional<StagingBufferRef> ref = TryGetReservedBuffer(size, usage, deferred)) {
         return *ref;
     }
-    return CreateStagingBuffer(size, usage);
+    return CreateStagingBuffer(size, usage, deferred);
 }
 
 std::optional<StagingBufferRef> StagingBufferPool::TryGetReservedBuffer(size_t size,
-                                                                        MemoryUsage usage) {
+                                                                        MemoryUsage usage,
+                                                                        bool deferred) {
     StagingBuffers& cache_level = GetCache(usage)[Common::Log2Ceil64(size)];
 
     const auto is_free = [this](const StagingBuffer& entry) {
-        return scheduler.IsFree(entry.tick);
+        return !entry.deferred && scheduler.IsFree(entry.tick);
     };
     auto& entries = cache_level.entries;
     const auto hint_it = entries.begin() + cache_level.iterate_index;
@@ -220,11 +238,14 @@ std::optional<StagingBufferRef> StagingBufferPool::TryGetReservedBuffer(size_t s
         }
     }
     cache_level.iterate_index = std::distance(entries.begin(), it) + 1;
-    it->tick = scheduler.CurrentTick();
+    it->tick = deferred ? std::numeric_limits<u64>::max() : scheduler.CurrentTick();
+    ASSERT(!it->deferred);
+    it->deferred = deferred;
     return it->Ref();
 }
 
-StagingBufferRef StagingBufferPool::CreateStagingBuffer(size_t size, MemoryUsage usage) {
+StagingBufferRef StagingBufferPool::CreateStagingBuffer(size_t size, MemoryUsage usage,
+                                                        bool deferred) {
     const u32 log2 = Common::Log2Ceil64(size);
     vk::Buffer buffer = device.GetLogical().CreateBuffer({
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
@@ -233,7 +254,8 @@ StagingBufferRef StagingBufferPool::CreateStagingBuffer(size_t size, MemoryUsage
         .size = 1ULL << log2,
         .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
                  VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
-                 VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT,
+                 VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
+                 VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT,
         .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
         .queueFamilyIndexCount = 0,
         .pQueueFamilyIndices = nullptr,
@@ -249,7 +271,11 @@ StagingBufferRef StagingBufferPool::CreateStagingBuffer(size_t size, MemoryUsage
         .buffer = std::move(buffer),
         .commit = std::move(commit),
         .mapped_span = mapped_span,
-        .tick = scheduler.CurrentTick(),
+        .usage = usage,
+        .log2_level = log2,
+        .index = unique_ids++,
+        .tick = deferred ? std::numeric_limits<u64>::max() : scheduler.CurrentTick(),
+        .deferred = deferred,
     });
     return entry.Ref();
 }
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
index 91dc84da8..4fd15f11a 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
@@ -1,5 +1,5 @@
-// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
 
 #pragma once
 
@@ -20,6 +20,9 @@ struct StagingBufferRef {
     VkBuffer buffer;
     VkDeviceSize offset;
     std::span<u8> mapped_span;
+    MemoryUsage usage;
+    u32 log2_level;
+    u64 index;
 };
 
 class StagingBufferPool {
@@ -30,7 +33,8 @@ public:
                                Scheduler& scheduler);
     ~StagingBufferPool();
 
-    StagingBufferRef Request(size_t size, MemoryUsage usage);
+    StagingBufferRef Request(size_t size, MemoryUsage usage, bool deferred = false);
+    void FreeDeferred(StagingBufferRef& ref);
 
     void TickFrame();
 
@@ -44,13 +48,20 @@ private:
         vk::Buffer buffer;
         MemoryCommit commit;
         std::span<u8> mapped_span;
+        MemoryUsage usage;
+        u32 log2_level;
+        u64 index;
         u64 tick = 0;
+        bool deferred{};
 
         StagingBufferRef Ref() const noexcept {
             return {
                 .buffer = *buffer,
                 .offset = 0,
                 .mapped_span = mapped_span,
+                .usage = usage,
+                .log2_level = log2_level,
+                .index = index,
             };
         }
     };
@@ -68,11 +79,12 @@ private:
 
     bool AreRegionsActive(size_t region_begin, size_t region_end) const;
 
-    StagingBufferRef GetStagingBuffer(size_t size, MemoryUsage usage);
+    StagingBufferRef GetStagingBuffer(size_t size, MemoryUsage usage, bool deferred = false);
 
-    std::optional<StagingBufferRef> TryGetReservedBuffer(size_t size, MemoryUsage usage);
+    std::optional<StagingBufferRef> TryGetReservedBuffer(size_t size, MemoryUsage usage,
+                                                         bool deferred);
 
-    StagingBufferRef CreateStagingBuffer(size_t size, MemoryUsage usage);
+    StagingBufferRef CreateStagingBuffer(size_t size, MemoryUsage usage, bool deferred);
 
     StagingBuffersCache& GetCache(MemoryUsage usage);
 
@@ -99,6 +111,7 @@ private:
 
     size_t current_delete_level = 0;
     u64 buffer_index = 0;
+    u64 unique_ids{};
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.cpp b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
index edb41b171..d56558a83 100644
--- a/src/video_core/renderer_vulkan/vk_state_tracker.cpp
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
@@ -27,10 +27,37 @@ using Flags = Maxwell3D::DirtyState::Flags;
 
 Flags MakeInvalidationFlags() {
     static constexpr int INVALIDATION_FLAGS[]{
-        Viewports,         Scissors,       DepthBias, BlendConstants,    DepthBounds,
-        StencilProperties, LineWidth,      CullMode,  DepthBoundsEnable, DepthTestEnable,
-        DepthWriteEnable,  DepthCompareOp, FrontFace, StencilOp,         StencilTestEnable,
-        VertexBuffers,     VertexInput,
+        Viewports,
+        Scissors,
+        DepthBias,
+        BlendConstants,
+        DepthBounds,
+        StencilProperties,
+        StencilReference,
+        StencilWriteMask,
+        StencilCompare,
+        LineWidth,
+        CullMode,
+        DepthBoundsEnable,
+        DepthTestEnable,
+        DepthWriteEnable,
+        DepthCompareOp,
+        FrontFace,
+        StencilOp,
+        StencilTestEnable,
+        VertexBuffers,
+        VertexInput,
+        StateEnable,
+        PrimitiveRestartEnable,
+        RasterizerDiscardEnable,
+        DepthBiasEnable,
+        LogicOpEnable,
+        DepthClampEnable,
+        LogicOp,
+        Blending,
+        ColorMask,
+        BlendEquations,
+        BlendEnable,
     };
     Flags flags{};
     for (const int flag : INVALIDATION_FLAGS) {
@@ -75,14 +102,17 @@ void SetupDirtyDepthBounds(Tables& tables) {
 }
 
 void SetupDirtyStencilProperties(Tables& tables) {
-    auto& table = tables[0];
-    table[OFF(stencil_two_side_enable)] = StencilProperties;
-    table[OFF(stencil_front_ref)] = StencilProperties;
-    table[OFF(stencil_front_mask)] = StencilProperties;
-    table[OFF(stencil_front_func_mask)] = StencilProperties;
-    table[OFF(stencil_back_ref)] = StencilProperties;
-    table[OFF(stencil_back_mask)] = StencilProperties;
-    table[OFF(stencil_back_func_mask)] = StencilProperties;
+    const auto setup = [&](size_t position, u8 flag) {
+        tables[0][position] = flag;
+        tables[1][position] = StencilProperties;
+    };
+    tables[0][OFF(stencil_two_side_enable)] = StencilProperties;
+    setup(OFF(stencil_front_ref), StencilReference);
+    setup(OFF(stencil_front_mask), StencilWriteMask);
+    setup(OFF(stencil_front_func_mask), StencilCompare);
+    setup(OFF(stencil_back_ref), StencilReference);
+    setup(OFF(stencil_back_mask), StencilWriteMask);
+    setup(OFF(stencil_back_func_mask), StencilCompare);
 }
 
 void SetupDirtyLineWidth(Tables& tables) {
@@ -96,16 +126,22 @@ void SetupDirtyCullMode(Tables& tables) {
     table[OFF(gl_cull_test_enabled)] = CullMode;
 }
 
-void SetupDirtyDepthBoundsEnable(Tables& tables) {
-    tables[0][OFF(depth_bounds_enable)] = DepthBoundsEnable;
-}
-
-void SetupDirtyDepthTestEnable(Tables& tables) {
-    tables[0][OFF(depth_test_enable)] = DepthTestEnable;
-}
-
-void SetupDirtyDepthWriteEnable(Tables& tables) {
-    tables[0][OFF(depth_write_enabled)] = DepthWriteEnable;
+void SetupDirtyStateEnable(Tables& tables) {
+    const auto setup = [&](size_t position, u8 flag) {
+        tables[0][position] = flag;
+        tables[1][position] = StateEnable;
+    };
+    setup(OFF(depth_bounds_enable), DepthBoundsEnable);
+    setup(OFF(depth_test_enable), DepthTestEnable);
+    setup(OFF(depth_write_enabled), DepthWriteEnable);
+    setup(OFF(stencil_enable), StencilTestEnable);
+    setup(OFF(primitive_restart.enabled), PrimitiveRestartEnable);
+    setup(OFF(rasterize_enable), RasterizerDiscardEnable);
+    setup(OFF(polygon_offset_point_enable), DepthBiasEnable);
+    setup(OFF(polygon_offset_line_enable), DepthBiasEnable);
+    setup(OFF(polygon_offset_fill_enable), DepthBiasEnable);
+    setup(OFF(logic_op.enable), LogicOpEnable);
+    setup(OFF(viewport_clip_control.geometry_clip), DepthClampEnable);
 }
 
 void SetupDirtyDepthCompareOp(Tables& tables) {
@@ -133,16 +169,22 @@ void SetupDirtyStencilOp(Tables& tables) {
     tables[1][OFF(stencil_two_side_enable)] = StencilOp;
 }
 
-void SetupDirtyStencilTestEnable(Tables& tables) {
-    tables[0][OFF(stencil_enable)] = StencilTestEnable;
-}
-
 void SetupDirtyBlending(Tables& tables) {
     tables[0][OFF(color_mask_common)] = Blending;
+    tables[1][OFF(color_mask_common)] = ColorMask;
     tables[0][OFF(blend_per_target_enabled)] = Blending;
+    tables[1][OFF(blend_per_target_enabled)] = BlendEquations;
     FillBlock(tables[0], OFF(color_mask), NUM(color_mask), Blending);
+    FillBlock(tables[1], OFF(color_mask), NUM(color_mask), ColorMask);
     FillBlock(tables[0], OFF(blend), NUM(blend), Blending);
+    FillBlock(tables[1], OFF(blend), NUM(blend), BlendEquations);
+    FillBlock(tables[1], OFF(blend.enable), NUM(blend.enable), BlendEnable);
     FillBlock(tables[0], OFF(blend_per_target), NUM(blend_per_target), Blending);
+    FillBlock(tables[1], OFF(blend_per_target), NUM(blend_per_target), BlendEquations);
+}
+
+void SetupDirtySpecialOps(Tables& tables) {
+    tables[0][OFF(logic_op.op)] = LogicOp;
 }
 
 void SetupDirtyViewportSwizzles(Tables& tables) {
@@ -185,17 +227,15 @@ void StateTracker::SetupTables(Tegra::Control::ChannelState& channel_state) {
     SetupDirtyStencilProperties(tables);
     SetupDirtyLineWidth(tables);
     SetupDirtyCullMode(tables);
-    SetupDirtyDepthBoundsEnable(tables);
-    SetupDirtyDepthTestEnable(tables);
-    SetupDirtyDepthWriteEnable(tables);
+    SetupDirtyStateEnable(tables);
     SetupDirtyDepthCompareOp(tables);
     SetupDirtyFrontFace(tables);
     SetupDirtyStencilOp(tables);
-    SetupDirtyStencilTestEnable(tables);
     SetupDirtyBlending(tables);
     SetupDirtyViewportSwizzles(tables);
     SetupDirtyVertexAttributes(tables);
     SetupDirtyVertexBindings(tables);
+    SetupDirtySpecialOps(tables);
 }
 
 void StateTracker::ChangeChannel(Tegra::Control::ChannelState& channel_state) {
@@ -204,6 +244,8 @@ void StateTracker::ChangeChannel(Tegra::Control::ChannelState& channel_state) {
 
 void StateTracker::InvalidateState() {
     flags->set();
+    current_topology = INVALID_TOPOLOGY;
+    stencil_reset = true;
 }
 
 StateTracker::StateTracker()
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.h b/src/video_core/renderer_vulkan/vk_state_tracker.h
index 2296dea60..8010ad26c 100644
--- a/src/video_core/renderer_vulkan/vk_state_tracker.h
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.h
@@ -35,6 +35,9 @@ enum : u8 {
     BlendConstants,
     DepthBounds,
     StencilProperties,
+    StencilReference,
+    StencilWriteMask,
+    StencilCompare,
     LineWidth,
 
     CullMode,
@@ -45,8 +48,18 @@ enum : u8 {
     FrontFace,
     StencilOp,
     StencilTestEnable,
+    PrimitiveRestartEnable,
+    RasterizerDiscardEnable,
+    DepthBiasEnable,
+    StateEnable,
+    LogicOp,
+    LogicOpEnable,
+    DepthClampEnable,
 
     Blending,
+    BlendEnable,
+    BlendEquations,
+    ColorMask,
     ViewportSwizzles,
 
     Last,
@@ -64,6 +77,7 @@ public:
     void InvalidateCommandBufferState() {
         (*flags) |= invalidation_flags;
         current_topology = INVALID_TOPOLOGY;
+        stencil_reset = true;
     }
 
     void InvalidateViewports() {
@@ -103,6 +117,57 @@ public:
         return Exchange(Dirty::StencilProperties, false);
     }
 
+    bool TouchStencilReference() {
+        return Exchange(Dirty::StencilReference, false);
+    }
+
+    bool TouchStencilWriteMask() {
+        return Exchange(Dirty::StencilWriteMask, false);
+    }
+
+    bool TouchStencilCompare() {
+        return Exchange(Dirty::StencilCompare, false);
+    }
+
+    template <typename T>
+    bool ExchangeCheck(T& old_value, T new_value) {
+        bool result = old_value != new_value;
+        old_value = new_value;
+        return result;
+    }
+
+    bool TouchStencilSide(bool two_sided_stencil_new) {
+        return ExchangeCheck(two_sided_stencil, two_sided_stencil_new) || stencil_reset;
+    }
+
+    bool CheckStencilReferenceFront(u32 new_value) {
+        return ExchangeCheck(front.ref, new_value) || stencil_reset;
+    }
+
+    bool CheckStencilReferenceBack(u32 new_value) {
+        return ExchangeCheck(back.ref, new_value) || stencil_reset;
+    }
+
+    bool CheckStencilWriteMaskFront(u32 new_value) {
+        return ExchangeCheck(front.write_mask, new_value) || stencil_reset;
+    }
+
+    bool CheckStencilWriteMaskBack(u32 new_value) {
+        return ExchangeCheck(back.write_mask, new_value) || stencil_reset;
+    }
+
+    bool CheckStencilCompareMaskFront(u32 new_value) {
+        return ExchangeCheck(front.compare_mask, new_value) || stencil_reset;
+    }
+
+    bool CheckStencilCompareMaskBack(u32 new_value) {
+        return ExchangeCheck(back.compare_mask, new_value) || stencil_reset;
+    }
+
+    void ClearStencilReset() {
+        stencil_reset = false;
+    }
+
     bool TouchLineWidth() const {
         return Exchange(Dirty::LineWidth, false);
     }
@@ -111,6 +176,10 @@ public:
         return Exchange(Dirty::CullMode, false);
     }
 
+    bool TouchStateEnable() {
+        return Exchange(Dirty::StateEnable, false);
+    }
+
     bool TouchDepthBoundsTestEnable() {
         return Exchange(Dirty::DepthBoundsEnable, false);
     }
@@ -123,6 +192,26 @@ public:
         return Exchange(Dirty::DepthWriteEnable, false);
     }
 
+    bool TouchPrimitiveRestartEnable() {
+        return Exchange(Dirty::PrimitiveRestartEnable, false);
+    }
+
+    bool TouchRasterizerDiscardEnable() {
+        return Exchange(Dirty::RasterizerDiscardEnable, false);
+    }
+
+    bool TouchDepthBiasEnable() {
+        return Exchange(Dirty::DepthBiasEnable, false);
+    }
+
+    bool TouchLogicOpEnable() {
+        return Exchange(Dirty::LogicOpEnable, false);
+    }
+
+    bool TouchDepthClampEnable() {
+        return Exchange(Dirty::DepthClampEnable, false);
+    }
+
     bool TouchDepthCompareOp() {
         return Exchange(Dirty::DepthCompareOp, false);
     }
@@ -135,10 +224,30 @@ public:
         return Exchange(Dirty::StencilOp, false);
     }
 
+    bool TouchBlending() {
+        return Exchange(Dirty::Blending, false);
+    }
+
+    bool TouchBlendEnable() {
+        return Exchange(Dirty::BlendEnable, false);
+    }
+
+    bool TouchBlendEquations() {
+        return Exchange(Dirty::BlendEquations, false);
+    }
+
+    bool TouchColorMask() {
+        return Exchange(Dirty::ColorMask, false);
+    }
+
     bool TouchStencilTestEnable() {
         return Exchange(Dirty::StencilTestEnable, false);
     }
 
+    bool TouchLogicOp() {
+        return Exchange(Dirty::LogicOp, false);
+    }
+
     bool ChangePrimitiveTopology(Maxwell::PrimitiveTopology new_topology) {
         const bool has_changed = current_topology != new_topology;
         current_topology = new_topology;
@@ -160,10 +269,20 @@ private:
         return is_dirty;
     }
 
+    struct StencilProperties {
+        u32 ref = 0;
+        u32 write_mask = 0;
+        u32 compare_mask = 0;
+    };
+
     Tegra::Engines::Maxwell3D::DirtyState::Flags* flags;
     Tegra::Engines::Maxwell3D::DirtyState::Flags default_flags;
     Tegra::Engines::Maxwell3D::DirtyState::Flags invalidation_flags;
     Maxwell::PrimitiveTopology current_topology = INVALID_TOPOLOGY;
+    bool two_sided_stencil = false;
+    StencilProperties front{};
+    StencilProperties back{};
+    bool stencil_reset = false;
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index a65bbeb1c..d39372ec4 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -812,8 +812,12 @@ StagingBufferRef TextureCacheRuntime::UploadStagingBuffer(size_t size) {
     return staging_buffer_pool.Request(size, MemoryUsage::Upload);
 }
 
-StagingBufferRef TextureCacheRuntime::DownloadStagingBuffer(size_t size) {
-    return staging_buffer_pool.Request(size, MemoryUsage::Download);
+StagingBufferRef TextureCacheRuntime::DownloadStagingBuffer(size_t size, bool deferred) {
+    return staging_buffer_pool.Request(size, MemoryUsage::Download, deferred);
+}
+
+void TextureCacheRuntime::FreeDeferredStagingBuffer(StagingBufferRef& ref) {
+    staging_buffer_pool.FreeDeferred(ref);
 }
 
 bool TextureCacheRuntime::ShouldReinterpret(Image& dst, Image& src) {
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h
index 7ec0df134..1f27a3589 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.h
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.h
@@ -51,7 +51,9 @@ public:
 
     StagingBufferRef UploadStagingBuffer(size_t size);
 
-    StagingBufferRef DownloadStagingBuffer(size_t size);
+    StagingBufferRef DownloadStagingBuffer(size_t size, bool deferred = false);
+
+    void FreeDeferredStagingBuffer(StagingBufferRef& ref);
 
     void TickFrame();
 
@@ -347,6 +349,7 @@ struct TextureCacheParams {
     static constexpr bool FRAMEBUFFER_BLITS = false;
     static constexpr bool HAS_EMULATED_COPIES = false;
     static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
+    static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = true;
 
     using Runtime = Vulkan::TextureCacheRuntime;
     using Image = Vulkan::Image;
@@ -354,6 +357,7 @@ struct TextureCacheParams {
     using ImageView = Vulkan::ImageView;
     using Sampler = Vulkan::Sampler;
     using Framebuffer = Vulkan::Framebuffer;
+    using AsyncBuffer = Vulkan::StagingBufferRef;
 };
 
 using TextureCache = VideoCommon::TextureCache<TextureCacheParams>;