diff --git a/rpcs3/Emu/RSX/GL/GLGSRender.cpp b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
index ff0724f32e..2ecbf1e901 100644
--- a/rpcs3/Emu/RSX/GL/GLGSRender.cpp
+++ b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
@@ -15,7 +15,7 @@ u64 GLGSRender::get_cycles()
 
 GLGSRender::GLGSRender() : GSRender()
 {
-	m_shaders_cache = std::make_unique<gl::shader_cache>(m_prog_buffer, "opengl", "v1.91");
+	m_shaders_cache = std::make_unique<gl::shader_cache>(m_prog_buffer, "opengl", "v1.92");
 
 	if (g_cfg.video.disable_vertex_cache || g_cfg.video.multithreaded_rsx)
 		m_vertex_cache = std::make_unique<gl::null_vertex_cache>();
diff --git a/rpcs3/Emu/RSX/Program/ProgramStateCache.cpp b/rpcs3/Emu/RSX/Program/ProgramStateCache.cpp
index 24378816d6..9cabeda367 100644
--- a/rpcs3/Emu/RSX/Program/ProgramStateCache.cpp
+++ b/rpcs3/Emu/RSX/Program/ProgramStateCache.cpp
@@ -36,10 +36,10 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 	//u32 last_instruction_address = 0;
 	//u32 first_instruction_address = entry;
 
-	std::stack<u32> call_stack;
-	std::pair<u32, u32> instruction_range{umax, 0};
-	std::bitset<512> instructions_to_patch;
+	std::bitset<rsx::max_vertex_program_instructions> instructions_to_patch;
+	std::pair<u32, u32> instruction_range{ umax, 0 };
 	bool has_branch_instruction = false;
+	std::stack<u32> call_stack;
 
 	D3  d3;
 	D2  d2;
@@ -54,7 +54,7 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 
 		while (true)
 		{
-			ensure(current_instruction < 512);
+			ensure(current_instruction < rsx::max_vertex_program_instructions);
 
 			if (result.instruction_mask[current_instruction])
 			{
@@ -120,8 +120,9 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 				instructions_to_patch[current_instruction] = true;
 				has_branch_instruction = true;
 
+				d0.HEX = instruction._u32[0];
 				d2.HEX = instruction._u32[2];
-				const u32 jump_address = ((d2.iaddrh << 3) | d3.iaddrl);
+				const u32 jump_address = (d0.iaddrh2 << 9) | (d2.iaddrh << 3) | d3.iaddrl;
 
 				if (function_call)
 				{
@@ -162,7 +163,7 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 			}
 
 			if ((d3.end && (fast_exit || current_instruction >= instruction_range.second)) ||
-				(current_instruction + 1) == 512)
+				(current_instruction + 1) == rsx::max_vertex_program_instructions)
 			{
 				break;
 			}
@@ -183,7 +184,7 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 	{
 		fs::file dump(fs::get_cache_dir() + "shaderlog/vp_analyser.bin", fs::rewrite);
 		dump.write(&entry, 4);
-		dump.write(data, 512 * 16);
+		dump.write(data, rsx::max_vertex_program_instructions * 16);
 		dump.close();
 	}
 
@@ -215,14 +216,17 @@ vertex_program_utils::vertex_program_metadata vertex_program_utils::analyse_vert
 
 				if (instructions_to_patch[i])
 				{
+					d0.HEX = dst[0];
 					d2.HEX = dst[2];
 					d3.HEX = dst[3];
 
-					u32 address = ((d2.iaddrh << 3) | d3.iaddrl);
+					u32 address = (d0.iaddrh2 << 9) | (d2.iaddrh << 3) | d3.iaddrl;
 					address -= instruction_range.first;
 
-					d2.iaddrh = (address >> 3);
+					d0.iaddrh2 = (address >> 9) & 0x1;
+					d2.iaddrh = (address >> 3) & 0x3F;
 					d3.iaddrl = (address & 0x7);
+					dst[0] = d0.HEX;
 					dst[2] = d2.HEX;
 					dst[3] = d3.HEX;
 
diff --git a/rpcs3/Emu/RSX/Program/ProgramStateCache.h b/rpcs3/Emu/RSX/Program/ProgramStateCache.h
index 54b12fcb1c..235ff6e9ca 100644
--- a/rpcs3/Emu/RSX/Program/ProgramStateCache.h
+++ b/rpcs3/Emu/RSX/Program/ProgramStateCache.h
@@ -22,7 +22,7 @@ namespace program_hash_util
 	{
 		struct vertex_program_metadata
 		{
-			std::bitset<512> instruction_mask;
+			std::bitset<rsx::max_vertex_program_instructions> instruction_mask;
 			u32 ucode_length;
 			u32 referenced_textures_mask;
 		};
diff --git a/rpcs3/Emu/RSX/Program/RSXVertexProgram.h b/rpcs3/Emu/RSX/Program/RSXVertexProgram.h
index 92013bfde8..df7e243e07 100644
--- a/rpcs3/Emu/RSX/Program/RSXVertexProgram.h
+++ b/rpcs3/Emu/RSX/Program/RSXVertexProgram.h
@@ -93,6 +93,13 @@ union D0
 		u32 vec_result           : 1;
 		u32                      : 1;
 	};
+
+	struct
+	{
+		u32         : 23;
+		u32 iaddrh2 : 1;
+		u32         : 8;
+	};
 };
 
 union D1
@@ -239,7 +246,7 @@ struct RSXVertexProgram
 
 	u32 base_address;
 	u32 entry;
-	std::bitset<512> instruction_mask;
+	std::bitset<rsx::max_vertex_program_instructions> instruction_mask;
 	std::set<u32> jump_table;
 
 	rsx::texture_dimension_extended get_texture_dimension(u8 id) const
diff --git a/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.cpp b/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.cpp
index 79af20991c..feb6b4218c 100644
--- a/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.cpp
+++ b/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.cpp
@@ -330,7 +330,7 @@ std::string VertexProgramDecompiler::AddCondReg()
 
 u32 VertexProgramDecompiler::GetAddr() const
 {
-	return (d2.iaddrh << 3) | d3.iaddrl;
+	return (d0.iaddrh2 << 9) | (d2.iaddrh << 3) | d3.iaddrl;
 }
 
 void VertexProgramDecompiler::AddCode(const std::string& code)
diff --git a/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.h b/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.h
index 1ce4f96fd0..35a3c69f24 100644
--- a/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.h
+++ b/rpcs3/Emu/RSX/Program/VertexProgramDecompiler.h
@@ -53,8 +53,7 @@ struct VertexProgramDecompiler
 		}
 	};
 
-	static const usz m_max_instr_count = 512;
-	Instruction m_instructions[m_max_instr_count];
+	Instruction m_instructions[rsx::max_vertex_program_instructions];
 	Instruction* m_cur_instr;
 	usz m_instr_count;
 
diff --git a/rpcs3/Emu/RSX/Program/program_util.h b/rpcs3/Emu/RSX/Program/program_util.h
index f74af801b1..c99ab6ac48 100644
--- a/rpcs3/Emu/RSX/Program/program_util.h
+++ b/rpcs3/Emu/RSX/Program/program_util.h
@@ -5,6 +5,11 @@
 
 namespace rsx
 {
+	enum program_limits
+	{
+		max_vertex_program_instructions = 544
+	};
+
 #pragma pack(push, 1)
 	// NOTE: This structure must be packed to match GPU layout.
 	struct fragment_program_texture_config
diff --git a/rpcs3/Emu/RSX/VK/VKGSRender.cpp b/rpcs3/Emu/RSX/VK/VKGSRender.cpp
index 240b2df2ab..6153b691c6 100644
--- a/rpcs3/Emu/RSX/VK/VKGSRender.cpp
+++ b/rpcs3/Emu/RSX/VK/VKGSRender.cpp
@@ -501,7 +501,7 @@ VKGSRender::VKGSRender() : GSRender()
 	else
 		m_vertex_cache = std::make_unique<vk::weak_vertex_cache>();
 
-	m_shaders_cache = std::make_unique<vk::shader_cache>(*m_prog_buffer, "vulkan", "v1.91");
+	m_shaders_cache = std::make_unique<vk::shader_cache>(*m_prog_buffer, "vulkan", "v1.92");
 
 	open_command_buffer();
 
diff --git a/rpcs3/Emu/RSX/rsx_cache.h b/rpcs3/Emu/RSX/rsx_cache.h
index ba629ca7dc..d972841e94 100644
--- a/rpcs3/Emu/RSX/rsx_cache.h
+++ b/rpcs3/Emu/RSX/rsx_cache.h
@@ -30,7 +30,7 @@ namespace rsx
 
 			u32 vp_ctrl;
 			u32 vp_texture_dimensions;
-			u64 vp_instruction_mask[8];
+			u64 vp_instruction_mask[9];
 
 			u32 vp_base_address;
 			u32 vp_entry;
@@ -365,7 +365,7 @@ namespace rsx
 			vp.base_address = data.vp_base_address;
 			vp.entry = data.vp_entry;
 
-			pack_bitset<512>(vp.instruction_mask, data.vp_instruction_mask);
+			pack_bitset<max_vertex_program_instructions>(vp.instruction_mask, data.vp_instruction_mask);
 
 			for (u8 index = 0; index < 32; ++index)
 			{
@@ -403,7 +403,7 @@ namespace rsx
 			data_block.vp_base_address = vp.base_address;
 			data_block.vp_entry = vp.entry;
 
-			unpack_bitset<512>(vp.instruction_mask, data_block.vp_instruction_mask);
+			unpack_bitset<max_vertex_program_instructions>(vp.instruction_mask, data_block.vp_instruction_mask);
 
 			u8 index = 0;
 			while (index < 32)
diff --git a/rpcs3/Emu/RSX/rsx_methods.cpp b/rpcs3/Emu/RSX/rsx_methods.cpp
index 4cce0bf476..1be9b80021 100644
--- a/rpcs3/Emu/RSX/rsx_methods.cpp
+++ b/rpcs3/Emu/RSX/rsx_methods.cpp
@@ -467,13 +467,10 @@ namespace rsx
 				u32 rcount = count;
 
 				if (const u32 max = load_pos * 4 + rcount + (index % 4);
-					max > 512 * 4)
+					max > max_vertex_program_instructions * 4)
 				{
-					// PS3 seems to allow exceeding the program buffer by upto 32 instructions before crashing
-					// Discard the "excess" instructions to not overflow our transform program buffer
-					// TODO: Check if the instructions in the overflow area are executed by PS3
-					rsx_log.warning("Program buffer overflow!");
-					rcount -= max - (512 * 4);
+					rsx_log.warning("Program buffer overflow! Attempted to write %u VP instructions.", max / 4);
+					rcount -= max - (max_vertex_program_instructions * 4);
 				}
 
 				stream_data_to_memory_swapped_u32<true>(&rsx::method_registers.transform_program[load_pos * 4 + index % 4]
diff --git a/rpcs3/Emu/RSX/rsx_methods.h b/rpcs3/Emu/RSX/rsx_methods.h
index a9845c79e5..e1e9ff9be5 100644
--- a/rpcs3/Emu/RSX/rsx_methods.h
+++ b/rpcs3/Emu/RSX/rsx_methods.h
@@ -8,6 +8,7 @@
 #include "rsx_vertex_data.h"
 #include "rsx_utils.h"
 #include "Emu/Cell/timers.hpp"
+#include "Program/program_util.h"
 
 namespace rsx
 {
@@ -492,7 +493,7 @@ namespace rsx
 		std::array<vertex_texture, 4> vertex_textures;
 
 
-		std::array<u32, 512 * 4> transform_program{};
+		std::array<u32, max_vertex_program_instructions * 4> transform_program{};
 		std::array<u32[4], 512> transform_constants{};
 
 		draw_clause current_draw_clause{};
diff --git a/rpcs3/Emu/RSX/rsx_utils.h b/rpcs3/Emu/RSX/rsx_utils.h
index 120a5aab89..df0c0e03be 100644
--- a/rpcs3/Emu/RSX/rsx_utils.h
+++ b/rpcs3/Emu/RSX/rsx_utils.h
@@ -855,18 +855,17 @@ namespace rsx
 	template <int N>
 	void unpack_bitset(const std::bitset<N>& block, u64* values)
 	{
-		constexpr int count = N / 64;
-		for (int n = 0; n < count; ++n)
+		for (int bit = 0, n = -1, shift = 0; bit < N; ++bit, ++shift)
 		{
-			int i = (n << 6);
-			values[n] = 0;
-
-			for (int bit = 0; bit < 64; ++bit, ++i)
+			if ((bit % 64) == 0)
 			{
-				if (block[i])
-				{
-					values[n] |= (1ull << bit);
-				}
+				values[++n] = 0;
+				shift = 0;
+			}
+
+			if (block[bit])
+			{
+				values[n] |= (1ull << shift);
 			}
 		}
 	}
@@ -874,18 +873,11 @@ namespace rsx
 	template <int N>
 	void pack_bitset(std::bitset<N>& block, u64* values)
 	{
-		constexpr int count = N / 64;
-		for (int n = (count - 1); n >= 0; --n)
+		for (int n = 0, shift = 0; shift < N; ++n, shift += 64)
 		{
-			if ((n + 1) < count)
-			{
-				block <<= 64;
-			}
-
-			if (values[n])
-			{
-				block |= values[n];
-			}
+			std::bitset<N> tmp = values[n];
+			tmp <<= shift;
+			block |= tmp;
 		}
 	}