diff --git a/libraries/gpu-gl/src/gpu/gl41/GL41BackendTexture.cpp b/libraries/gpu-gl/src/gpu/gl41/GL41BackendTexture.cpp
index 9fe6bbb772..a9d1ddb914 100644
--- a/libraries/gpu-gl/src/gpu/gl41/GL41BackendTexture.cpp
+++ b/libraries/gpu-gl/src/gpu/gl41/GL41BackendTexture.cpp
@@ -427,15 +427,14 @@ void GL41VariableAllocationTexture::populateTransferQueue() {
 
             // break down the transfers into chunks so that no single transfer is 
             // consuming more than X bandwidth
+            // For compressed format, regions must be a multiple of the 4x4 tiles, so enforce 4 lines as the minimum block
             auto mipSize = _gpuObject.getStoredMipFaceSize(sourceMip, face);
             const auto lines = mipDimensions.y;
-            const uint32_t CHUNK_NUM_LINES { 4 };
-            const auto numChunks = (lines + (CHUNK_NUM_LINES - 1)) / CHUNK_NUM_LINES;
-            auto bytesPerChunk = mipSize / numChunks;
-            //auto bytesPerLine = mipSize / lines;
+            const uint32_t BLOCK_NUM_LINES { 4 };
+            const auto numBlocks = (lines + (BLOCK_NUM_LINES - 1)) / BLOCK_NUM_LINES;
+            auto bytesPerBlock = mipSize / numBlocks;
             Q_ASSERT(0 == (mipSize % lines));
-            //   uint32_t linesPerTransfer = (uint32_t)(MAX_TRANSFER_SIZE / bytesPerLine);
-            uint32_t linesPerTransfer = CHUNK_NUM_LINES * (uint32_t)(MAX_TRANSFER_SIZE / bytesPerChunk);
+            uint32_t linesPerTransfer = BLOCK_NUM_LINES * (uint32_t)(MAX_TRANSFER_SIZE / bytesPerBlock);
             uint32_t lineOffset = 0;
             while (lineOffset < lines) {
                 uint32_t linesToCopy = std::min<uint32_t>(lines - lineOffset, linesPerTransfer);
diff --git a/libraries/gpu-gl/src/gpu/gl45/GL45BackendVariableTexture.cpp b/libraries/gpu-gl/src/gpu/gl45/GL45BackendVariableTexture.cpp
index 192b7f3088..d46f38d88f 100644
--- a/libraries/gpu-gl/src/gpu/gl45/GL45BackendVariableTexture.cpp
+++ b/libraries/gpu-gl/src/gpu/gl45/GL45BackendVariableTexture.cpp
@@ -195,13 +195,14 @@ void GL45ResourceTexture::populateTransferQueue() {
 
             // break down the transfers into chunks so that no single transfer is 
             // consuming more than X bandwidth
+            // For compressed format, regions must be a multiple of the 4x4 tiles, so enforce 4 lines as the minimum block
             auto mipSize = _gpuObject.getStoredMipFaceSize(sourceMip, face);
             const auto lines = mipDimensions.y;
-            const uint32_t CHUNK_NUM_LINES { 4 };
-            const auto numChunks = (lines + (CHUNK_NUM_LINES - 1)) / CHUNK_NUM_LINES;
-            auto bytesPerChunk = mipSize / numChunks;
+            const uint32_t BLOCK_NUM_LINES { 4 };
+            const auto numBlocks = (lines + (BLOCK_NUM_LINES - 1)) / BLOCK_NUM_LINES;
+            auto bytesPerBlock = mipSize / numBlocks;
             Q_ASSERT(0 == (mipSize % lines));
-            uint32_t linesPerTransfer = CHUNK_NUM_LINES * (uint32_t)(MAX_TRANSFER_SIZE / bytesPerChunk);
+            uint32_t linesPerTransfer = BLOCK_NUM_LINES * (uint32_t)(MAX_TRANSFER_SIZE / bytesPerBlock);
             uint32_t lineOffset = 0;
             while (lineOffset < lines) {
                 uint32_t linesToCopy = std::min<uint32_t>(lines - lineOffset, linesPerTransfer);