diff --git a/lib/GfxRenderer/GfxRenderer.cpp b/lib/GfxRenderer/GfxRenderer.cpp
index ddd5f1e4..b024fab8 100644
--- a/lib/GfxRenderer/GfxRenderer.cpp
+++ b/lib/GfxRenderer/GfxRenderer.cpp
@@ -661,8 +661,10 @@ void GfxRenderer::drawRoundedRect(const int x, const int y, const int width, con
 }
 
 void GfxRenderer::fillRect(const int x, const int y, const int width, const int height, const bool state) const {
-  for (int fillY = y; fillY < y + height; fillY++) {
-    drawLine(x, fillY, x + width - 1, fillY, state);
+  if (state) {
+    fillRectImpl<Color::Black>(x, y, width, height);
+  } else {
+    fillRectImpl<Color::White>(x, y, width, height);
   }
 }
 
@@ -694,26 +696,194 @@ void GfxRenderer::drawPixelDither<Color::DarkGray>(const int x, const int y) con
 }
 
 void GfxRenderer::fillRectDither(const int x, const int y, const int width, const int height, Color color) const {
-  if (color == Color::Clear) {
-  } else if (color == Color::Black) {
-    fillRect(x, y, width, height, true);
-  } else if (color == Color::White) {
-    fillRect(x, y, width, height, false);
-  } else if (color == Color::LightGray) {
-    for (int fillY = y; fillY < y + height; fillY++) {
-      for (int fillX = x; fillX < x + width; fillX++) {
-        drawPixelDither<Color::LightGray>(fillX, fillY);
+  switch (color) {
+    case Color::Clear:
+      break;
+    case Color::Black:
+      fillRectImpl<Color::Black>(x, y, width, height);
+      break;
+    case Color::White:
+      fillRectImpl<Color::White>(x, y, width, height);
+      break;
+    case Color::LightGray:
+      fillRectImpl<Color::LightGray>(x, y, width, height);
+      break;
+    case Color::DarkGray:
+      fillRectImpl<Color::DarkGray>(x, y, width, height);
+      break;
+  }
+}
+
+template <Color C>
+void GfxRenderer::fillRectImpl(const int x, const int y, const int width, const int height) const {
+  if constexpr (C == Color::Clear) return;
+  if (width <= 0 || height <= 0) return;
+  if (fontCacheManager_ && fontCacheManager_->isScanning()) return;
+
+  // Clip in logical space.
+  const int screenW = getScreenWidth();
+  const int screenH = getScreenHeight();
+  const int lx0 = std::max(0, x);
+  const int ly0 = std::max(0, y);
+  const int lx1 = std::min(screenW, x + width);
+  const int ly1 = std::min(screenH, y + height);
+  if (lx0 >= lx1 || ly0 >= ly1) return;
+
+  // Rotate the two opposing logical corners into physical-framebuffer space.
+  // The bounding rect in physical space is the rect we need to fill — rotation
+  // is rigid (no shear/stretch) so the bbox of the two corners IS the rect.
+  int paX, paY, pbX, pbY;
+  rotateCoordinates(orientation, lx0, ly0, &paX, &paY, panelWidth, panelHeight);
+  rotateCoordinates(orientation, lx1 - 1, ly1 - 1, &pbX, &pbY, panelWidth, panelHeight);
+
+  const int phyX0 = std::min(paX, pbX);
+  const int phyX1 = std::max(paX, pbX);  // inclusive
+  int phyY0 = std::min(paY, pbY);
+  int phyY1 = std::max(paY, pbY);
+
+  // Strip mode: clip Y range to the active band and redirect writes.
+  uint8_t* target = getWriteTarget();
+  const int originY = getWriteOriginY();
+  const int writeRows = getWriteRows();
+  phyY0 = std::max(phyY0, originY);
+  phyY1 = std::min(phyY1, originY + writeRows - 1);
+  if (phyY0 > phyY1) return;
+
+  // Bit/byte layout: MSB-first within a byte, so phyX → bit (7 - (phyX & 7)).
+  // Head and tail masks cover only the in-rect bits of the first/last byte.
+  const int byteStart = phyX0 >> 3;
+  const int byteEnd = phyX1 >> 3;  // inclusive
+  const uint8_t headMask = static_cast<uint8_t>(0xFFu >> (phyX0 & 7));
+  const uint8_t tailMask = static_cast<uint8_t>(0xFFu << (7 - (phyX1 & 7)));
+  const int32_t panelStride = static_cast<int32_t>(panelWidthBytes);
+
+  if constexpr (C == Color::Black || C == Color::White) {
+    // Solid fill. Framebuffer: 0 = black, 1 = white.
+    const uint8_t fillByte = (C == Color::Black) ? 0x00u : 0xFFu;
+    for (int py = phyY0; py <= phyY1; ++py) {
+      uint8_t* row = target + static_cast<int32_t>(py - originY) * panelStride;
+      if (byteStart == byteEnd) {
+        const uint8_t mask = headMask & tailMask;
+        if constexpr (C == Color::Black) {
+          row[byteStart] &= static_cast<uint8_t>(~mask);
+        } else {
+          row[byteStart] |= mask;
+        }
+      } else {
+        if constexpr (C == Color::Black) {
+          row[byteStart] &= static_cast<uint8_t>(~headMask);
+          if (byteEnd > byteStart + 1) {
+            memset(row + byteStart + 1, fillByte, byteEnd - byteStart - 1);
+          }
+          row[byteEnd] &= static_cast<uint8_t>(~tailMask);
+        } else {
+          row[byteStart] |= headMask;
+          if (byteEnd > byteStart + 1) {
+            memset(row + byteStart + 1, fillByte, byteEnd - byteStart - 1);
+          }
+          row[byteEnd] |= tailMask;
+        }
       }
     }
-  } else if (color == Color::DarkGray) {
-    for (int fillY = y; fillY < y + height; fillY++) {
-      for (int fillX = x; fillX < x + width; fillX++) {
-        drawPixelDither<Color::DarkGray>(fillX, fillY);
+  } else {
+    // Dither (LightGray / DarkGray). Both patterns have period 2 in logical
+    // (x, y), so per physical row we precompute one byte that represents the
+    // pattern across an 8-pixel stretch — every full byte in the row uses
+    // that same value.
+    //
+    // dlxPerPhyX / dlyPerPhyX: how logical (x, y) change as phyX increments
+    // along a physical row. Derived from inverting rotateCoordinates.
+    int dlxPerPhyX = 0, dlyPerPhyX = 0;
+    switch (orientation) {
+      case Portrait:
+        dlxPerPhyX = 0;
+        dlyPerPhyX = 1;
+        break;
+      case PortraitInverted:
+        dlxPerPhyX = 0;
+        dlyPerPhyX = -1;
+        break;
+      case LandscapeClockwise:
+        dlxPerPhyX = -1;
+        dlyPerPhyX = 0;
+        break;
+      case LandscapeCounterClockwise:
+        dlxPerPhyX = 1;
+        dlyPerPhyX = 0;
+        break;
+    }
+
+    // The dither pattern has period 2 in logical space, and each orientation
+    // maps py to logical coords with a fixed parity relationship. The
+    // blackMask byte therefore repeats with period 2 in py. Precompute both
+    // variants outside the row loop to eliminate the per-row switch + 8-bit
+    // construction loop.
+    uint8_t blackMasks[2];
+    for (int parityIdx = 0; parityIdx < 2; ++parityIdx) {
+      const int samplePy = phyY0 + parityIdx;
+      int lxBase = 0, lyBase = 0;
+      switch (orientation) {
+        case Portrait:
+          lxBase = panelHeight - 1 - samplePy;
+          lyBase = byteStart * 8;
+          break;
+        case PortraitInverted:
+          lxBase = samplePy;
+          lyBase = panelWidth - 1 - byteStart * 8;
+          break;
+        case LandscapeClockwise:
+          lxBase = panelWidth - 1 - byteStart * 8;
+          lyBase = panelHeight - 1 - samplePy;
+          break;
+        case LandscapeCounterClockwise:
+          lxBase = byteStart * 8;
+          lyBase = samplePy;
+          break;
+      }
+      uint8_t mask = 0;
+      for (int b = 0; b < 8; ++b) {
+        const int lx = lxBase + b * dlxPerPhyX;
+        const int ly = lyBase + b * dlyPerPhyX;
+        bool isBlack;
+        if constexpr (C == Color::LightGray) {
+          isBlack = ((lx & 1) == 0) && ((ly & 1) == 0);
+        } else {  // DarkGray
+          isBlack = (((lx + ly) & 1) == 0);
+        }
+        if (isBlack) mask |= static_cast<uint8_t>(1u << (7 - b));
+      }
+      blackMasks[samplePy & 1] = mask;
+    }
+
+    for (int py = phyY0; py <= phyY1; ++py) {
+      const uint8_t blackMask = blackMasks[py & 1];
+      const uint8_t whiteMask = static_cast<uint8_t>(~blackMask);
+
+      // Dither writes BOTH inks (the slow path called drawPixel for every
+      // pixel — setting or clearing — so we must do the same). Inside the
+      // rect mask: write whiteMask (1s where white, 0s where black). Outside
+      // the rect mask: leave the framebuffer untouched.
+      uint8_t* row = target + static_cast<int32_t>(py - originY) * panelStride;
+      if (byteStart == byteEnd) {
+        const uint8_t rectMask = headMask & tailMask;
+        row[byteStart] = static_cast<uint8_t>((row[byteStart] & ~rectMask) | (rectMask & whiteMask));
+      } else {
+        row[byteStart] = static_cast<uint8_t>((row[byteStart] & ~headMask) | (headMask & whiteMask));
+        if (byteEnd > byteStart + 1) {
+          // Period 2, so every full byte in this row is exactly whiteMask.
+          memset(row + byteStart + 1, whiteMask, byteEnd - byteStart - 1);
+        }
+        row[byteEnd] = static_cast<uint8_t>((row[byteEnd] & ~tailMask) | (tailMask & whiteMask));
       }
     }
   }
 }
 
+template void GfxRenderer::fillRectImpl<Color::Black>(int, int, int, int) const;
+template void GfxRenderer::fillRectImpl<Color::White>(int, int, int, int) const;
+template void GfxRenderer::fillRectImpl<Color::LightGray>(int, int, int, int) const;
+template void GfxRenderer::fillRectImpl<Color::DarkGray>(int, int, int, int) const;
+
 void GfxRenderer::maskRoundedRectOutsideCorners(const int x, const int y, const int width, const int height,
                                                 const int radius, const Color color) const {
   if (radius <= 0 || color == Color::Clear) {
diff --git a/lib/GfxRenderer/GfxRenderer.h b/lib/GfxRenderer/GfxRenderer.h
index 471c0a8f..ae000bb2 100644
--- a/lib/GfxRenderer/GfxRenderer.h
+++ b/lib/GfxRenderer/GfxRenderer.h
@@ -81,6 +81,12 @@ class GfxRenderer {
   void drawPixelDither(int x, int y) const;
   template <Color color>
   void fillArc(int maxRadius, int cx, int cy, int xDir, int yDir) const;
+  // Byte-aligned, orientation-specialized rectangle fill. Rotates the rect's
+  // two opposing corners into physical-framebuffer space once, then walks each
+  // physical row with head-mask / middle memset / tail-mask byte writes — no
+  // per-pixel rotation, no per-pixel RMW.
+  template <Color color>
+  void fillRectImpl(int x, int y, int width, int height) const;
 
  public:
   explicit GfxRenderer(HalDisplay& halDisplay)