diff --git a/src/Pipeline/VertexProgram.cpp b/src/Pipeline/VertexProgram.cpp
index 5bb3918..55c7b41 100644
--- a/src/Pipeline/VertexProgram.cpp
+++ b/src/Pipeline/VertexProgram.cpp
@@ -70,13 +70,9 @@
 		{
 			assert(it->second.SizeInComponents == 1);
 
-			Int4 indices;
-			indices = Insert(indices, As<Int>(batch[0]), 0);
-			indices = Insert(indices, As<Int>(batch[1]), 1);
-			indices = Insert(indices, As<Int>(batch[2]), 2);
-			indices = Insert(indices, As<Int>(batch[3]), 3);
 			routine.getVariable(it->second.Id)[it->second.FirstComponent] =
-					As<Float4>(indices + Int4(*Pointer<Int>(data + OFFSET(DrawData, baseVertex))));
+					As<Float4>(*Pointer<Int4>(As<Pointer<Int4>>(batch)) +
+					           Int4(*Pointer<Int>(data + OFFSET(DrawData, baseVertex))));
 		}
 
 		auto activeLaneMask = SIMD::Int(0xFFFFFFFF);
diff --git a/src/Pipeline/VertexRoutine.cpp b/src/Pipeline/VertexRoutine.cpp
index baa69e4..1c90a6e 100644
--- a/src/Pipeline/VertexRoutine.cpp
+++ b/src/Pipeline/VertexRoutine.cpp
@@ -94,13 +94,14 @@
 			{
 				Pointer<Byte> input = *Pointer<Pointer<Byte>>(data + OFFSET(DrawData, input) + sizeof(void*) * (i / 4));
 				UInt stride = *Pointer<UInt>(data + OFFSET(DrawData, stride) + sizeof(uint32_t) * (i / 4));
+				Int baseVertex = *Pointer<Int>(data + OFFSET(DrawData, baseVertex));
 				UInt robustnessSize(0);
 				if(state.robustBufferAccess)
 				{
 					robustnessSize = *Pointer<UInt>(data + OFFSET(DrawData, robustnessSize) + sizeof(uint32_t) * (i / 4));
 				}
 
-				auto value = readStream(input, stride, state.input[i / 4], batch, state.robustBufferAccess, robustnessSize);
+				auto value = readStream(input, stride, state.input[i / 4], batch, state.robustBufferAccess, robustnessSize, baseVertex);
 				routine.inputs[i + 0] = value.x;
 				routine.inputs[i + 1] = value.y;
 				routine.inputs[i + 2] = value.z;
@@ -143,10 +144,15 @@
 	}
 
 	Vector4f VertexRoutine::readStream(Pointer<Byte> &buffer, UInt &stride, const Stream &stream, Pointer<UInt> &batch,
-	                                   bool robustBufferAccess, UInt & robustnessSize)
+	                                   bool robustBufferAccess, UInt & robustnessSize, Int baseVertex)
 	{
 		Vector4f v;
-		UInt4 offsets = *Pointer<UInt4>(As<Pointer<UInt4>>(batch)) * UInt4(stride);
+		// Because of the following rule in the Vulkan spec, we do not care if a very large negative
+		// baseVertex would overflow all the way back into a valid region of the index buffer:
+		// "Out-of-bounds buffer loads will return any of the following values :
+		//  - Values from anywhere within the memory range(s) bound to the buffer (possibly including
+		//    bytes of memory past the end of the buffer, up to the end of the bound range)."
+		UInt4 offsets = (*Pointer<UInt4>(As<Pointer<UInt4>>(batch)) + As<UInt4>(Int4(baseVertex))) * UInt4(stride);
 
 		Pointer<Byte> source0 = buffer + offsets.x;
 		Pointer<Byte> source1 = buffer + offsets.y;
diff --git a/src/Pipeline/VertexRoutine.hpp b/src/Pipeline/VertexRoutine.hpp
index a7d3103..69dfa7c 100644
--- a/src/Pipeline/VertexRoutine.hpp
+++ b/src/Pipeline/VertexRoutine.hpp
@@ -67,7 +67,7 @@
 		typedef VertexProcessor::State::Input Stream;
 
 		Vector4f readStream(Pointer<Byte> &buffer, UInt &stride, const Stream &stream, Pointer<UInt> &batch,
-		                    bool robustBufferAccess, UInt& robustnessSize);
+		                    bool robustBufferAccess, UInt& robustnessSize, Int baseVertex);
 		void readInput(Pointer<UInt> &batch);
 		void computeClipFlags();
 		void writeCache(Pointer<Byte> &vertexCache, Pointer<UInt> &tagCache, Pointer<UInt> &batch);
diff --git a/src/Vulkan/VkCommandBuffer.cpp b/src/Vulkan/VkCommandBuffer.cpp
index da411cd..de7057d 100644
--- a/src/Vulkan/VkCommandBuffer.cpp
+++ b/src/Vulkan/VkCommandBuffer.cpp
@@ -411,7 +411,7 @@
 	uint32_t reference;
 };
 
-void CommandBuffer::ExecutionState::bindVertexInputs(sw::Context& context, int firstVertex, int firstInstance)
+void CommandBuffer::ExecutionState::bindVertexInputs(sw::Context& context, int firstInstance)
 {
 	for(uint32_t i = 0; i < MAX_VERTEX_INPUT_BINDINGS; i++)
 	{
@@ -420,7 +420,6 @@
 		{
 			const auto &vertexInput = vertexInputBindings[attrib.binding];
 			VkDeviceSize offset = attrib.offset + vertexInput.offset +
-			                      attrib.vertexStride * firstVertex +
 			                      attrib.instanceStride * firstInstance;
 			attrib.buffer = vertexInput.buffer ? vertexInput.buffer->getOffsetPointer(offset) : nullptr;
 
@@ -524,7 +523,7 @@
 
 		sw::Context context = pipeline->getContext();
 
-		executionState.bindVertexInputs(context, vertexOffset, firstInstance);
+		executionState.bindVertexInputs(context, firstInstance);
 
 		context.descriptorSets = pipelineState.descriptorSets;
 		context.descriptorDynamicOffsets = pipelineState.descriptorDynamicOffsets;
diff --git a/src/Vulkan/VkCommandBuffer.hpp b/src/Vulkan/VkCommandBuffer.hpp
index b592add..01664e4 100644
--- a/src/Vulkan/VkCommandBuffer.hpp
+++ b/src/Vulkan/VkCommandBuffer.hpp
@@ -180,7 +180,7 @@
 		uint32_t subpassIndex = 0;
 
 		void bindAttachments(sw::Context& context);
-		void bindVertexInputs(sw::Context& context, int firstVertex, int firstInstance);
+		void bindVertexInputs(sw::Context& context, int firstInstance);
 	};
 
 	void submit(CommandBuffer::ExecutionState& executionState);