Project Not Cumulative

pseymour said:

This may help, or steer you toward more topics.

Windows Updates using forward and reverse differentials

A technique to produce compact software updates optimized for any origin and destination revision pair.

learn.microsoft.com

And by the way, I encourage exploration and salute your efforts. Also, I have been told by my employees that I ask an annoying amount of questions, so there's that.

Just to add on to what you already confirmed, I installed the 24H2 March 2026 ISO. Confirmed revision was 26100.8037 (March 10). Setup did its normal OOBE updates, but nothing else, as I had updates paused during a late stage of setup.

I reset my router's stats and installed a bandwidth monitor on the PC. I then installed all Windows Updates. Confirmed revision was 26100.8457, the May 12, 2026 revision.

The router and monitor app showed I had downloaded basically the same amount, about 2,060 MB. These are the updates I downloaded, along with their sizes from the catalog.

KB5089549 - 5,232.8 MB
KB5087054 - 73.6 MB

Defender:

KB4052623 - ~80 MB
KB2267602 - ~3,200 MB
KB5007651 - ~44 MB

KB890830 - ~82 MB

For a total of about 8,712 MB. So 6,600 MB+ just didn't happen.

Click to expand...

I know for a fact that Microsoft wasn't always using this method. Previously, Windows Update would download roughly 2 GB of update data, including the PSF files, and then extract everything locally. It did not require downloading the full MSU package, but it still downloaded and processed the complete update payload. That was still the case up until around August of last year.

When Microsoft switched to this newer approach, I don't know. However, it does appear that Windows Update is now downloading significantly less data than I originally thought, so I was clearly wrong about that part.

What still doesn't make sense to me is the size discrepancy. How do 26,000 extracted files end up consuming around 4.6 GB, while 80,000 extracted files total only 1.4 GB? Even if compression or reconstruction is involved, the numbers still seem strange. For comparison, the entire Windows folder in a Windows 10 installation is only around 7 GB when uncompressed. Seeing a small cumulative update produce 4.6 GB of extracted data seems excessive.
So while I now accept that Windows Update is downloading much less data than I assumed, I still don't have a good explanation for the huge difference in extracted size. That's the part that continues to puzzle me.

I also hope the feedback I provided over the years helped Microsoft move toward this more efficient delivery method.

That said, Windows Update is still significantly slower on my systems than installing the update directly with DISM. Even if the download is smaller, the overall installation process takes noticeably longer in my testing.

Hi Jen, you have the file for Build 26200.8655? Thanks...