diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 6aceceaa4..8f90c9b00 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -103,6 +103,7 @@ add_library(citra_common STATIC
     logging/text_formatter.cpp
     logging/text_formatter.h
     logging/types.h
+    math_util.cpp
     math_util.h
     memory_detect.cpp
     memory_detect.h
diff --git a/src/common/math_util.cpp b/src/common/math_util.cpp
new file mode 100644
index 000000000..f9e7b5a35
--- /dev/null
+++ b/src/common/math_util.cpp
@@ -0,0 +1,151 @@
+// Copyright Citra Emulator Project / Azahar Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include "math_util.h"
+
+#if defined(CITRA_HAS_SSE42)
+#include <emmintrin.h>
+#include <smmintrin.h>
+#endif
+
+#if defined(__aarch64__) || defined(__ARM_NEON)
+#define CITRA_HAS_NEON
+#include <arm_neon.h>
+#endif
+
+#if defined(_MSC_VER)
+#define DISABLE_VECTORIZE __pragma(loop(no_vector))
+#elif defined(__clang__)
+#define DISABLE_VECTORIZE _Pragma("clang loop vectorize(disable)")
+#elif defined(__GNUC__)
+#define DISABLE_VECTORIZE _Pragma("GCC novector")
+#else
+#define DISABLE_VECTORIZE
+#endif
+
+namespace Common {
+std::pair<u8, u8> FindMinMax(const std::span<const u8>& data) {
+    const size_t count = data.size();
+    const u8* data_ptr = data.data();
+    u8 final_min, final_max;
+#if defined(CITRA_HAS_SSE42) || defined(CITRA_HAS_NEON)
+    u8 simd_min = 0xFF;
+    u8 simd_max = 0;
+    size_t i = 0;
+    constexpr size_t simd_line_count = 16;
+    constexpr size_t count_threshold = simd_line_count * 2;
+    if (count >= count_threshold) {
+#if defined(CITRA_HAS_SSE42)
+        __m128i vmin = _mm_set1_epi8(static_cast<char>(0xFF));
+        __m128i vmax = _mm_setzero_si128();
+        for (; i + simd_line_count <= count; i += simd_line_count) {
+            __m128i vals = _mm_loadu_si128(reinterpret_cast<const __m128i*>(data_ptr + i));
+            vmin = _mm_min_epu8(vmin, vals);
+            vmax = _mm_max_epu8(vmax, vals);
+        }
+        alignas(16) u8 tmp[simd_line_count];
+        _mm_storeu_si128(reinterpret_cast<__m128i*>(tmp), vmin);
+        simd_min = *std::min_element(tmp, tmp + simd_line_count);
+        _mm_storeu_si128(reinterpret_cast<__m128i*>(tmp), vmax);
+        simd_max = *std::max_element(tmp, tmp + simd_line_count);
+#elif defined(CITRA_HAS_NEON)
+        uint8x16_t vmin = vdupq_n_u8(0xFF);
+        uint8x16_t vmax = vdupq_n_u8(0);
+        for (; i + simd_line_count <= count; i += simd_line_count) {
+            uint8x16_t vals = vld1q_u8(data_ptr + i);
+            vmin = vminq_u8(vmin, vals);
+            vmax = vmaxq_u8(vmax, vals);
+        }
+        alignas(16) uint8_t tmp[simd_line_count];
+        vst1q_u8(tmp, vmin);
+        simd_min = *std::min_element(tmp, tmp + simd_line_count);
+        vst1q_u8(tmp, vmax);
+        simd_max = *std::max_element(tmp, tmp + simd_line_count);
+#endif // CITRA_HAS_SSE42
+    }
+    DISABLE_VECTORIZE
+    for (; i < count; ++i) {
+        const u8 val = data_ptr[i];
+        simd_min = std::min(simd_min, val);
+        simd_max = std::max(simd_max, val);
+    }
+
+    final_min = simd_min;
+    final_max = simd_max;
+
+#else
+    // Scalar fallback
+    for (size_t i = 0; i < count; ++i) {
+        const u8 val = data_ptr[i];
+        final_min = std::min(final_min, val);
+        final_max = std::max(final_max, val);
+    }
+#endif // CITRA_HAS_SSE42 || CITRA_HAS_NEON
+
+    return {final_min, final_max};
+}
+
+std::pair<u16, u16> FindMinMax(const std::span<const u16>& data) {
+    const size_t count = data.size();
+    const u16* data_ptr = data.data();
+    u16 final_min, final_max;
+
+#if defined(CITRA_HAS_SSE42) || defined(CITRA_HAS_NEON)
+    u16 simd_min = 0xFFFF;
+    u16 simd_max = 0;
+    size_t i = 0;
+    constexpr size_t simd_line_count = 8;
+    constexpr size_t count_threshold = simd_line_count * 2;
+    if (count >= count_threshold) {
+#if defined(CITRA_HAS_SSE42)
+        __m128i vmin = _mm_set1_epi16(static_cast<short>(0xFFFF));
+        __m128i vmax = _mm_setzero_si128();
+        for (; i + simd_line_count <= count; i += simd_line_count) {
+            __m128i vals = _mm_loadu_si128(reinterpret_cast<const __m128i*>(data_ptr + i));
+            vmin = _mm_min_epu16(vmin, vals);
+            vmax = _mm_max_epu16(vmax, vals);
+        }
+        alignas(16) u16 tmp[simd_line_count];
+        _mm_storeu_si128(reinterpret_cast<__m128i*>(tmp), vmin);
+        simd_min = *std::min_element(tmp, tmp + simd_line_count);
+        _mm_storeu_si128(reinterpret_cast<__m128i*>(tmp), vmax);
+        simd_max = *std::max_element(tmp, tmp + simd_line_count);
+#elif defined(CITRA_HAS_NEON)
+        uint16x8_t vmin = vdupq_n_u16(static_cast<u16>(0xFFFF));
+        uint16x8_t vmax = vdupq_n_u16(0);
+        for (; i + simd_line_count <= count; i += simd_line_count) {
+            uint16x8_t vals = vld1q_u16(data_ptr + i);
+            vmin = vminq_u16(vmin, vals);
+            vmax = vmaxq_u16(vmax, vals);
+        }
+        alignas(16) uint16_t tmp[simd_line_count];
+        vst1q_u16(tmp, vmin);
+        simd_min = *std::min_element(tmp, tmp + simd_line_count);
+        vst1q_u16(tmp, vmax);
+        simd_max = *std::max_element(tmp, tmp + simd_line_count);
+#endif // CITRA_HAS_SSE42
+    }
+    DISABLE_VECTORIZE
+    for (; i < count; ++i) {
+        const u16 val = data_ptr[i];
+        simd_min = std::min(simd_min, val);
+        simd_max = std::max(simd_max, val);
+    }
+
+    final_min = simd_min;
+    final_max = simd_max;
+
+#else
+    // Scalar fallback
+    for (u32 i = 0; i < count; ++i) {
+        const u16 val = data_ptr[i];
+        final_min = std::min(final_min, val);
+        final_max = std::max(final_max, val);
+    }
+#endif // CITRA_HAS_SSE42 || CITRA_HAS_NEON
+
+    return {final_min, final_max};
+}
+} // namespace Common
diff --git a/src/common/math_util.h b/src/common/math_util.h
index 7b5513788..9587fba93 100644
--- a/src/common/math_util.h
+++ b/src/common/math_util.h
@@ -9,7 +9,10 @@
 #pragma once
 
 #include <cstdlib>
+#include <span>
 #include <type_traits>
+#include <utility>
+#include "common_types.h"
 
 namespace Common {
 
@@ -73,4 +76,7 @@ struct Rectangle {
 template <typename T>
 Rectangle(T, T, T, T) -> Rectangle<T>;
 
+std::pair<u8, u8> FindMinMax(const std::span<const u8>& data);
+std::pair<u16, u16> FindMinMax(const std::span<const u16>& data);
+
 } // namespace Common
diff --git a/src/video_core/rasterizer_accelerated.cpp b/src/video_core/rasterizer_accelerated.cpp
index c6ebfdb70..d7f018405 100644
--- a/src/video_core/rasterizer_accelerated.cpp
+++ b/src/video_core/rasterizer_accelerated.cpp
@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "common/alignment.h"
+#include "common/math_util.h"
 #include "core/memory.h"
 #include "video_core/pica/pica_core.h"
 #include "video_core/rasterizer_accelerated.h"
@@ -103,12 +104,19 @@ RasterizerAccelerated::VertexArrayInfo RasterizerAccelerated::AnalyzeVertexArray
 
         vertex_min = 0xFFFF;
         vertex_max = 0;
-        const u32 size = regs.pipeline.num_vertices * (index_u16 ? 2 : 1);
+        const u32 count = regs.pipeline.num_vertices;
+        const u32 index_size = index_u16 ? 2 : 1;
+        const u32 size = count * index_size;
         FlushRegion(address, size);
-        for (u32 index = 0; index < regs.pipeline.num_vertices; ++index) {
-            const u32 vertex = index_u16 ? index_address_16[index] : index_address_8[index];
-            vertex_min = std::min(vertex_min, vertex);
-            vertex_max = std::max(vertex_max, vertex);
+
+        if (index_u16) {
+            const auto res = Common::FindMinMax({index_address_16, static_cast<size_t>(count)});
+            vertex_min = static_cast<u32>(res.first);
+            vertex_max = static_cast<u32>(res.second);
+        } else {
+            const auto res = Common::FindMinMax({index_address_8, static_cast<size_t>(count)});
+            vertex_min = static_cast<u32>(res.first);
+            vertex_max = static_cast<u32>(res.second);
         }
     } else {
         vertex_min = regs.pipeline.vertex_offset;