diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 58f3f88..6e0c136 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -38,7 +38,6 @@ jobs:
           - "ubuntu-22.04"
           - "windows-2022"
         wheel-selector:
-          - "cp38-*"
           - "cp39-*"
           - "cp310-*"
           - "cp311-*"
@@ -57,7 +56,7 @@ jobs:
           fetch-depth: 0
 
       - name: Build wheels
-        uses: pypa/cibuildwheel@v2.20.0
+        uses: pypa/cibuildwheel@v2.21.3
         with:
           output-dir: wheelhouse
         env:
@@ -71,6 +70,11 @@ jobs:
           CIBW_ARCHS_MACOS: "x86_64 universal2 arm64"
           # Skip testing arm64 builds on Intel Macs.
           CIBW_TEST_SKIP: "*-macosx_arm64 *-macosx_universal2:arm64"
+          # Use newer manylinux so we get newer gcc and can take advantage of
+          # speed boost from x86-64-v2, at the cost of not supporting some old
+          # patch releases of Python 3.9 if they haven't updated pip in their
+          # virtualenv.
+          CIBW_MANYLINUX_X86_64_IMAGE: "manylinux_2_28"
       - uses: actions/upload-artifact@v3
         name: Upload artifacts
         with:
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 852ee25..2a51799 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -21,13 +21,11 @@ jobs:
           - "ubuntu-22.04"
           - "windows-2022"
         python-version:
-          - "3.8"
           - "3.9"
           - "3.10"
           - "3.11"
           - "3.12"
-          - "3.13-dev"
-          - "pypy-3.8"
+          - "3.13"
           - "pypy-3.9"
           - "pypy-3.10"
 
diff --git a/setup.py b/setup.py
index 27695b6..8c1789d 100644
--- a/setup.py
+++ b/setup.py
@@ -1,8 +1,8 @@
 from setuptools import setup
 from setuptools.extension import Extension
 
+import platform
 import sys
-import os
 import versioneer
 
 DEBUGMODE = False
@@ -24,6 +24,11 @@
         break
 
 extra_compile_args.append("-std=c99")
+if platform.machine() == "x86_64" and platform.system().lower().startswith("linux"):
+    # Only support CPUs starting from 2008/2009 or so, and run twice as fast as
+    # a result! This requires sufficiently new gcc, 11 maybe?
+    extra_compile_args.append("-march=x86-64-v2")
+
 if DEBUGMODE:
     extra_compile_args.append("-O0")
     extra_compile_args.append("-g")