diff --git a/GPU/GLES/DepalettizeShaderGLES.cpp b/GPU/GLES/DepalettizeShaderGLES.cpp
index 49e1b36f3c..77614c93df 100644
--- a/GPU/GLES/DepalettizeShaderGLES.cpp
+++ b/GPU/GLES/DepalettizeShaderGLES.cpp
@@ -95,7 +95,7 @@ GLRTexture *DepalShaderCacheGLES::GetClutTexture(GEPaletteFormat clutFormat, con
 
 	uint8_t *clutCopy = new uint8_t[1024];
 	memcpy(clutCopy, rawClut, 1024);
-	render_->TextureImage(tex->texture, 0, texturePixels, 1, components, components2, dstFmt, clutCopy, false);
+	render_->TextureImage(tex->texture, 0, texturePixels, 1, components, components2, dstFmt, clutCopy, GLRAllocType::NEW, false);
 
 	tex->lastFrame = gpuStats.numFlips;
 	texCache_[clutId] = tex;
diff --git a/GPU/GLES/DrawEngineGLES.cpp b/GPU/GLES/DrawEngineGLES.cpp
index 5988ee5fa1..bb0676719c 100644
--- a/GPU/GLES/DrawEngineGLES.cpp
+++ b/GPU/GLES/DrawEngineGLES.cpp
@@ -777,7 +777,7 @@ void DrawEngineGLES::TessellationDataTransferGLES::SendDataToShader(const float
 	uint8_t *pos_data = new uint8_t[size * sizeof(float) * 4];
 	memcpy(pos_data, pos, size * sizeof(float) * 4);
 	data_tex[0] = renderManager_->CreateTexture(GL_TEXTURE_2D);
-	renderManager_->TextureImage(data_tex[0], 0, size, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, pos_data, false);
+	renderManager_->TextureImage(data_tex[0], 0, size, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, pos_data, GLRAllocType::NEW, false);
 	renderManager_->FinalizeTexture(data_tex[0], 0, false);
 	renderManager_->BindTexture(4, data_tex[0]);
 
@@ -788,7 +788,7 @@ void DrawEngineGLES::TessellationDataTransferGLES::SendDataToShader(const float
 		uint8_t *tex_data = new uint8_t[size * sizeof(float) * 4];
 		memcpy(tex_data, pos, size * sizeof(float) * 4);
 		data_tex[1] = renderManager_->CreateTexture(GL_TEXTURE_2D);
-		renderManager_->TextureImage(data_tex[1], 0, size, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, tex_data, false);
+		renderManager_->TextureImage(data_tex[1], 0, size, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, tex_data, GLRAllocType::NEW, false);
 		renderManager_->FinalizeTexture(data_tex[1], 0, false);
 		renderManager_->BindTexture(5, data_tex[1]);
 	}
@@ -800,7 +800,7 @@ void DrawEngineGLES::TessellationDataTransferGLES::SendDataToShader(const float
 	uint8_t *col_data = new uint8_t[sizeColor * sizeof(float) * 4];
 	memcpy(col_data, col, sizeColor * sizeof(float) * 4);
 
-	renderManager_->TextureImage(data_tex[2], 0, sizeColor, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, col_data, false);
+	renderManager_->TextureImage(data_tex[2], 0, sizeColor, 1, GL_RGBA32F, GL_RGBA, GL_FLOAT, col_data, GLRAllocType::NEW, false);
 	renderManager_->FinalizeTexture(data_tex[2], 0, false);
 	renderManager_->BindTexture(6, data_tex[2]);
 }
diff --git a/GPU/GLES/FramebufferManagerGLES.cpp b/GPU/GLES/FramebufferManagerGLES.cpp
index dd0a0c2473..351579744f 100644
--- a/GPU/GLES/FramebufferManagerGLES.cpp
+++ b/GPU/GLES/FramebufferManagerGLES.cpp
@@ -367,7 +367,7 @@ void FramebufferManagerGLES::MakePixelTexture(const u8 *srcPixels, GEBufferForma
 			break;
 		}
 	}
-	render_->TextureImage(drawPixelsTex_, 0, texWidth, height, GL_RGBA, GL_RGBA, GL_UNSIGNED_BYTE, convBuf, false);
+	render_->TextureImage(drawPixelsTex_, 0, texWidth, height, GL_RGBA, GL_RGBA, GL_UNSIGNED_BYTE, convBuf, GLRAllocType::NEW, false);
 	render_->FinalizeTexture(drawPixelsTex_, 0, false);
 
 	// TODO: Return instead?
diff --git a/GPU/GLES/TextureCacheGLES.cpp b/GPU/GLES/TextureCacheGLES.cpp
index d16f14f372..1c1ff3551d 100644
--- a/GPU/GLES/TextureCacheGLES.cpp
+++ b/GPU/GLES/TextureCacheGLES.cpp
@@ -714,7 +714,7 @@ u8 *TextureCacheGLES::DecodeTextureLevelOld(GETextureFormat format, GEPaletteFor
 		decPitch = bufw * pixelSize;
 	}
 
-	uint8_t *texBuf = new uint8_t[std::max(w, bufw) * h * pixelSize];
+	uint8_t *texBuf = (uint8_t *)AllocateAlignedMemory(std::max(w, bufw) * h * pixelSize, 16);
 	DecodeTextureLevel(texBuf, decPitch, format, clutformat, texaddr, level, bufw, true, false, false);
 	return texBuf;
 }
@@ -755,7 +755,7 @@ void TextureCacheGLES::LoadTextureLevel(TexCacheEntry &entry, ReplacedTexture &r
 		PROFILE_THIS_SCOPE("replacetex");
 
 		int bpp = replaced.Format(level) == ReplacedTextureFormat::F_8888 ? 4 : 2;
-		uint8_t *rearrange = new uint8_t[w * h * bpp];
+		uint8_t *rearrange = (uint8_t *)AllocateAlignedMemory(w * h * bpp, 16);
 		replaced.Load(level, rearrange, bpp * w);
 		pixelData = rearrange;
 
@@ -785,7 +785,7 @@ void TextureCacheGLES::LoadTextureLevel(TexCacheEntry &entry, ReplacedTexture &r
 		}
 
 		if (scaleFactor > 1) {
-			uint8_t *rearrange = new uint8_t[w * scaleFactor * h * scaleFactor * 4];
+			uint8_t *rearrange = (uint8_t *)AllocateAlignedMemory(w * scaleFactor * h * scaleFactor * 4, 16);
 			scaler.ScaleAlways((u32 *)rearrange, (u32 *)pixelData, dstFmt, w, h, scaleFactor);
 			pixelData = rearrange;
 			delete [] finalBuf;
@@ -817,9 +817,9 @@ void TextureCacheGLES::LoadTextureLevel(TexCacheEntry &entry, ReplacedTexture &r
 	} else {
 		PROFILE_THIS_SCOPE("loadtex");
 		if (IsFakeMipmapChange())
-			render_->TextureImage(entry.textureName, 0, w, h, components, components2, dstFmt, pixelData);
+			render_->TextureImage(entry.textureName, 0, w, h, components, components2, dstFmt, pixelData, GLRAllocType::ALIGNED);
 		else
-			render_->TextureImage(entry.textureName, level, w, h, components, components2, dstFmt, pixelData);
+			render_->TextureImage(entry.textureName, level, w, h, components, components2, dstFmt, pixelData, GLRAllocType::ALIGNED);
 	}
 }
 
diff --git a/ext/native/thin3d/GLQueueRunner.cpp b/ext/native/thin3d/GLQueueRunner.cpp
index e37b411e69..1053752af1 100644
--- a/ext/native/thin3d/GLQueueRunner.cpp
+++ b/ext/native/thin3d/GLQueueRunner.cpp
@@ -1,3 +1,4 @@
+#include "Common/MemoryUtil.h"
 #include "Core/Reporting.h"
 #include "GLQueueRunner.h"
 #include "GLRenderManager.h"
@@ -241,7 +242,11 @@ void GLQueueRunner::RunInitSteps(const std::vector<GLRInitStep> &steps) {
 			// For things to show in RenderDoc, need to split into glTexImage2D(..., nullptr) and glTexSubImage.
 			glTexImage2D(tex->target, step.texture_image.level, step.texture_image.internalFormat, step.texture_image.width, step.texture_image.height, 0, step.texture_image.format, step.texture_image.type, step.texture_image.data);
 			allocatedTextures = true;
-			delete[] step.texture_image.data;
+			if (step.texture_image.allocType == GLRAllocType::ALIGNED) {
+				FreeAlignedMemory(step.texture_image.data);
+			} else {
+				delete[] step.texture_image.data;
+			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			tex->wrapS = GL_CLAMP_TO_EDGE;
 			tex->wrapT = GL_CLAMP_TO_EDGE;
diff --git a/ext/native/thin3d/GLQueueRunner.h b/ext/native/thin3d/GLQueueRunner.h
index 63a17503c4..9119b0d05f 100644
--- a/ext/native/thin3d/GLQueueRunner.h
+++ b/ext/native/thin3d/GLQueueRunner.h
@@ -19,6 +19,10 @@ struct GLOffset2D {
 	int x, y;
 };
 
+enum class GLRAllocType {
+	NEW,
+	ALIGNED,
+};
 
 class GLRShader;
 class GLRTexture;
@@ -242,6 +246,7 @@ struct GLRInitStep {
 			int level;
 			int width;
 			int height;
+			GLRAllocType allocType;
 			bool linearFilter;
 			uint8_t *data;  // owned, delete[]-d
 		} texture_image;
diff --git a/ext/native/thin3d/GLRenderManager.h b/ext/native/thin3d/GLRenderManager.h
index 2e50f392a5..c3b473e605 100644
--- a/ext/native/thin3d/GLRenderManager.h
+++ b/ext/native/thin3d/GLRenderManager.h
@@ -337,7 +337,7 @@ public:
 	}
 
 	// Takes ownership over the data pointer and delete[]-s it.
-	void TextureImage(GLRTexture *texture, int level, int width, int height, GLenum internalFormat, GLenum format, GLenum type, uint8_t *data, bool linearFilter = false) {
+	void TextureImage(GLRTexture *texture, int level, int width, int height, GLenum internalFormat, GLenum format, GLenum type, uint8_t *data, GLRAllocType allocType = GLRAllocType::NEW, bool linearFilter = false) {
 		GLRInitStep step{ GLRInitStepType::TEXTURE_IMAGE };
 		step.texture_image.texture = texture;
 		step.texture_image.data = data;
@@ -347,6 +347,7 @@ public:
 		step.texture_image.level = level;
 		step.texture_image.width = width;
 		step.texture_image.height = height;
+		step.texture_image.allocType = allocType;
 		step.texture_image.linearFilter = linearFilter;
 		initSteps_.push_back(step);
 	}