diff --git a/src/Cafe/CMakeLists.txt b/src/Cafe/CMakeLists.txt
index 8dd401b7..c7360458 100644
--- a/src/Cafe/CMakeLists.txt
+++ b/src/Cafe/CMakeLists.txt
@@ -558,6 +558,8 @@ if(ENABLE_METAL)
         HW/Latte/Renderer/Metal/MetalBufferAllocator.h
         HW/Latte/Renderer/Metal/MetalMemoryManager.cpp
         HW/Latte/Renderer/Metal/MetalMemoryManager.h
+        HW/Latte/Renderer/Metal/MetalOutputShaderCache.cpp
+        HW/Latte/Renderer/Metal/MetalOutputShaderCache.h
         HW/Latte/Renderer/Metal/MetalPipelineCache.cpp
         HW/Latte/Renderer/Metal/MetalPipelineCache.h
         HW/Latte/Renderer/Metal/MetalDepthStencilCache.cpp
diff --git a/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.cpp b/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.cpp
new file mode 100644
index 00000000..8a69a442
--- /dev/null
+++ b/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.cpp
@@ -0,0 +1,38 @@
+#include "Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h"
+#include "Cafe/HW/Latte/Renderer/Metal/RendererShaderMtl.h"
+
+MetalOutputShaderCache::~MetalOutputShaderCache()
+{
+    for (uint8 i = 0; i < METAL_OUTPUT_SHADER_CACHE_SIZE; i++)
+    {
+        if (m_cache[i])
+            m_cache[i]->release();
+    }
+}
+
+MTL::RenderPipelineState* MetalOutputShaderCache::GetPipeline(RendererOutputShader* shader, uint8 shaderIndex, bool usesSRGB)
+{
+    uint8 cacheIndex = (usesSRGB ? METAL_SHADER_TYPE_COUNT : 0) + shaderIndex;
+    auto& renderPipelineState = m_cache[cacheIndex];
+    if (renderPipelineState)
+        return renderPipelineState;
+
+    // Create a new render pipeline state
+    auto vertexShaderMtl = static_cast<RendererShaderMtl*>(shader->GetVertexShader())->GetFunction();
+    auto fragmentShaderMtl = static_cast<RendererShaderMtl*>(shader->GetFragmentShader())->GetFunction();
+
+    auto renderPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
+    renderPipelineDescriptor->setVertexFunction(vertexShaderMtl);
+    renderPipelineDescriptor->setFragmentFunction(fragmentShaderMtl);
+    renderPipelineDescriptor->colorAttachments()->object(0)->setPixelFormat(usesSRGB ? MTL::PixelFormatBGRA8Unorm_sRGB : MTL::PixelFormatBGRA8Unorm);
+
+    NS::Error* error = nullptr;
+    renderPipelineState = m_mtlr->GetDevice()->newRenderPipelineState(renderPipelineDescriptor, &error);
+    if (error)
+    {
+        cemuLog_log(LogType::Force, "error creating output render pipeline state: {}", error->localizedDescription()->utf8String());
+        error->release();
+    }
+
+    return renderPipelineState;
+}
diff --git a/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h b/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h
new file mode 100644
index 00000000..85b9e8b2
--- /dev/null
+++ b/src/Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h
@@ -0,0 +1,20 @@
+#pragma once
+
+#include "Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h"
+
+constexpr uint8 METAL_SHADER_TYPE_COUNT = 6;
+constexpr uint8 METAL_OUTPUT_SHADER_CACHE_SIZE = 2 * METAL_SHADER_TYPE_COUNT;
+
+class MetalOutputShaderCache
+{
+public:
+    MetalOutputShaderCache(class MetalRenderer* metalRenderer) : m_mtlr{metalRenderer} {}
+    ~MetalOutputShaderCache();
+
+    MTL::RenderPipelineState* GetPipeline(RendererOutputShader* shader, uint8 shaderIndex, bool usesSRGB);
+
+private:
+    class MetalRenderer* m_mtlr;
+
+    MTL::RenderPipelineState* m_cache[METAL_OUTPUT_SHADER_CACHE_SIZE] = {nullptr};
+};
diff --git a/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp b/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
index 9e0b4641..11724544 100644
--- a/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
+++ b/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.cpp
@@ -4,6 +4,7 @@
 #include "Cafe/HW/Latte/Renderer/Metal/LatteTextureViewMtl.h"
 #include "Cafe/HW/Latte/Renderer/Metal/RendererShaderMtl.h"
 #include "Cafe/HW/Latte/Renderer/Metal/CachedFBOMtl.h"
+#include "Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h"
 #include "Cafe/HW/Latte/Renderer/Metal/MetalPipelineCache.h"
 #include "Cafe/HW/Latte/Renderer/Metal/MetalDepthStencilCache.h"
 #include "Cafe/HW/Latte/Renderer/Metal/MetalSamplerCache.h"
@@ -81,6 +82,7 @@ MetalRenderer::MetalRenderer()
     textureDescriptor->release();
 
     m_memoryManager = new MetalMemoryManager(this);
+    m_outputShaderCache = new MetalOutputShaderCache(this);
     m_pipelineCache = new MetalPipelineCache(this);
     m_depthStencilCache = new MetalDepthStencilCache(this);
     m_samplerCache = new MetalSamplerCache(this);
@@ -174,6 +176,7 @@ MetalRenderer::~MetalRenderer()
     m_presentPipelineLinear->release();
     m_presentPipelineSRGB->release();
 
+    delete m_outputShaderCache;
     delete m_pipelineCache;
     delete m_depthStencilCache;
     delete m_samplerCache;
@@ -276,7 +279,6 @@ void MetalRenderer::SwapBuffers(bool swapTV, bool swapDRC)
     m_performanceMonitor.ResetPerFrameData();
 }
 
-// TODO: use `shader` for drawing
 void MetalRenderer::DrawBackbufferQuad(LatteTextureView* texView, RendererOutputShader* shader, bool useLinearTexFilter,
 								sint32 imageX, sint32 imageY, sint32 imageWidth, sint32 imageHeight,
 								bool padView, bool clearBackground)
@@ -299,21 +301,20 @@ void MetalRenderer::DrawBackbufferQuad(LatteTextureView* texView, RendererOutput
     renderPassDescriptor->release();
 
     // Get a render pipeline
-    auto vertexShaderMtl = static_cast<RendererShaderMtl*>(shader->GetVertexShader())->GetFunction();
-    auto fragmentShaderMtl = static_cast<RendererShaderMtl*>(shader->GetFragmentShader())->GetFunction();
 
-    auto renderPipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
-    renderPipelineDescriptor->setVertexFunction(vertexShaderMtl);
-    renderPipelineDescriptor->setFragmentFunction(fragmentShaderMtl);
-    renderPipelineDescriptor->colorAttachments()->object(0)->setPixelFormat(m_state.m_usesSRGB ? MTL::PixelFormatBGRA8Unorm_sRGB : MTL::PixelFormatBGRA8Unorm);
+    // Find out which shader we are using
+    uint8 shaderIndex = 255;
+    if (shader == RendererOutputShader::s_copy_shader) shaderIndex = 0;
+    else if (shader == RendererOutputShader::s_bicubic_shader) shaderIndex = 1;
+    else if (shader == RendererOutputShader::s_hermit_shader) shaderIndex = 2;
+    else if (shader == RendererOutputShader::s_copy_shader_ud) shaderIndex = 3;
+    else if (shader == RendererOutputShader::s_bicubic_shader_ud) shaderIndex = 4;
+    else if (shader == RendererOutputShader::s_hermit_shader_ud) shaderIndex = 5;
 
-    NS::Error* error = nullptr;
-    auto renderPipelineState = m_device->newRenderPipelineState(renderPipelineDescriptor, &error);
-    if (error)
-    {
-        printf("AAA: %s\n", error->localizedDescription()->utf8String());
-        error->release();
-    }
+    uint8 shaderType = shaderIndex % 3;
+
+    // Get the render pipeline state
+    auto renderPipelineState = m_outputShaderCache->GetPipeline(shader, shaderIndex, m_state.m_usesSRGB);
 
     // Draw to Metal layer
     renderCommandEncoder->setRenderPipelineState(renderPipelineState);
diff --git a/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h b/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h
index 8eb0e319..04956ac7 100644
--- a/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h
+++ b/src/Cafe/HW/Latte/Renderer/Metal/MetalRenderer.h
@@ -4,6 +4,7 @@
 
 #include "Cafe/HW/Latte/Renderer/Metal/MetalLayerHandle.h"
 #include "Cafe/HW/Latte/Renderer/Metal/MetalPerformanceMonitor.h"
+#include "Cafe/HW/Latte/Renderer/Metal/MetalOutputShaderCache.h"
 
 struct MetalBufferAllocation
 {
@@ -460,6 +461,7 @@ private:
 
 	// Managers and caches
 	class MetalMemoryManager* m_memoryManager;
+	class MetalOutputShaderCache* m_outputShaderCache;
 	class MetalPipelineCache* m_pipelineCache;
 	class MetalDepthStencilCache* m_depthStencilCache;
 	class MetalSamplerCache* m_samplerCache;