Solved Install fresh Win7 on M.2 stick - have RTFM'd yet still no success...

[IDEngineer]

I've spent all day on this and still cannot get it to work.


I'm building a second Win7 machine as insurance against my primary Win7 machine having a problem. Given the efforts to make Win7 unusable on later generations, I specifically chose the most recent generation of components that have native Win7 support. I ended up with a brand new Asus Prime Z270-A, Intel i7-6700K, 32GB of DDR4-3200 with the lowest latency I could find that was on the Asus QVL for that board, etc.


I specifically chose this motherboard because it supports two M.2 sticks in NVMe (not just SATA) mode. Might as well get the most performance possible.I have a 500GB Samsung 980 Pro installed in the M2_1 slot to hold the OS and applications, and I'll add a 2GB M.2 later for data.


The problem is I cannot get Win7's installer to see the M.2. I've read countless articles that dance around this topic Here's the sequence I followed:



* The BIOS sees the M.2 drive and properly identifies its size and details, so we're communicating with the M.2 at least at the hardware level. There are no other drives of any kind in the machine, on purpose, other than the USB stick with the Win7 ISO image.



* I initially created a USB stick with the bootable install image of Win7 Ultimate. It booted, but since this motherboard only natively exposes a single sub-USB3 connection (and Win7 doesn't natively handle USB3+) I had to grab the USB3 drivers from the Asus CD and incorporate them into the bootable image using DISM. That worked, and now I'm able to have the USB stick, and a keyboard, and a mouse all operational at the same time.


* The Win7 installer boots just fine and gets all the way to the point where it realizes it cannot see any drives on which to install the OS, so it prompts for drivers. No problem, I'm prepared, I have the "bare" drivers from Samsung on the USB stick. Except that it claims the Samsung drivers aren't signed and refuses to install them.


* No problem, I reimaged install.wim and boot.wim again and incorporated the Samsung drivers, using the "force" option to override the lack of signature. This was successful, so I tried again but the installer still cannot see the M.2 drive.


* Then I ran across a discussion of two Win7 hotfixes, KB2990941 and KB3087873, which (though vaguely described) seem to add NVMe support to Win7. I incorporated those into the ISO image using DISM and tried again. No luck.


* I created brand new versions of install.wim and boot.wim, this time with just the USB3 and hotfixes, leaving out the Samsung drivers. Make a tiny bit of progress... now the Samsung .inf file is visible (actually, the installer lists it THREE times) but when I select and try to load it the installer claims "No new devices could be found".


* I thought perhaps going back to versions of the Samsung drivers closer to the formal end of Win7 might prove helpful since 1) Samsung claims their later M.2's are backwards compatible, but I can't find drivers on any official Samsung site anywhere. Every reference to M.2 drivers says "No drivers are necessary, your M.2 will be automatically supported". Yeah, right.


* There's been some suggestion that the hotfixes could eliminate the need for separate, manufacturer-specific drivers and that the M.2 drive would become accessible solely with the hotfixes. That has not happened. I've tried .wim files both with, and without, the Samsung drivers but nothing makes the drive accessible within the installer.


I'm sort of stuck now. The Samsung drivers I have either "aren't signed" (if I try to load them manually) or I have to force past the lack of signature (if I incorporate them into the ISO image). The hotfixes magically ignore the lack of signature but don't "see" the physical device, yet I know the hardware is accessible because the BIOS accurately reports everything about the drive. Thus this is a Win7 installer problem, not a hardware problem. I'm confident my .wim file handling is OK because the USB3 and hotfix issues are solved, but incorporating the Samsung drivers the same way yields nothing.


Any advice would be GREATLY appreciated. Absolute worst case I can go back to regular SSD's but it would be major bummer if I have two nice M.2 slots sitting here that I cannot use.


Thanks!

 

[SIW2]

There's been some suggestion that the hotfixes could eliminate the need for separate, manufacturer-specific drivers and that the M.2 drive would become accessible solely with the hotfixes.

The two nvme hotfixes need to be installed into boot.wim image 2 and install.wim ( whatever image number you are going to install.


In addition after integrating the 2 nvme hotfixes into boot.wim image 2 - you need to copy the entire sources folder from boot.wim image 2 onto the installation media, overwriting the sources folder and its contents in the installation media.

If you missed that step, the updated version of setup.exe in boot.wim sources folder will not recognize the installation media and will ask for drivers.


Or you can use my win7 updater tool which will do it all for you. You need to provide installation media for the tool to update.

Update your Win 7 installation media

 

[IDEngineer]

The two nvme hotfixes need to be installed into boot.wim image 2 and install.wim ( whatever image number you are going to install. In addition after integrating the 2 nvme hotfixes into boot.wim image 2 - you need to copy the entire sources folder from boot.wim image 2 onto the installation media, overwriting the sources folder and its contents in the installation media. If you missed that step, the updated version of setup.exe in boot.wim sources folder will not recognize the installation media and will ask for drivers.

As I tried to explain in my original post above, I did those things. I used DISM /add-package to install each of the hotfixes KB2990941 and KB3087873 to both boot.wim (image:2) and install.wim (image:1). Also used DISM to /add-driver the USB3 drivers. DISM reported success with all steps. I then copied the modified .wim files to the USB stick I'm using to install Win7. It boots the machine and doesn't complain until it can't find a drive to which it can install the OS.

I'm reading your comments more closely: "...you need to copy the entire sources folder from boot.wim image 2 onto the installation media, overwriting the sources folder and its contents in the installation media". Strictly speaking, this sounds like I should 1) take a copy of the entire /sources folder, 2) modify the .wim files, and then 3) freshly overwrite the entire /sources folder with the copy + updated .wim files. Are you actually saying that? It was my understanding that DISM only changes the .wim file(s) that you mount for it to modify. It sounds like you're suggesting it also modifies other files in /sources. That would be news to me, because when I'm running DISM I copy the two .wim files to my hard drive and then mount/dismount them from there. In other words, when DISM is running its command line has nothing about the USB stick. Are you saying DISM modifies other files besides the .wim's, and that you really do need to copy and overwrite the entire contents of the /sources directory after using DISM to modify the .wim files?


Thanks for your help, I sincerely appreciate it!

 

[SIW2]

If you have already integrated the nvme hotfixes into boot.wim and install.wim, then you can do the copy with 7-zip.

 

[IDEngineer]

Thank you for the very clear screenshots, but I notice that the first file that's going to be copied is the same file (note timestamp).


This suggests that DISM should be allowed to operate on the install.wim and boot.wim in directory that contains all of the other files from /sources. Yet when I DON'T do that, DISM doesn't complain about lack of access to other files. Does it actually alter anything besides install.wim and boot.wim? If it doesn't, it seems like there's no reason to duplicate everything.


I'll give it a try, but I'm trying to understand why copying the same files over existing copies would make a difference.

 

[SIW2]

but I notice that the first file that's going to be copied is the same file (note timestamp).

It doesn't matter if you overwrite an existing file with one that is the same.

It is simpler for you to copy the whole lot rather than going through them individually to find out which are the updated files.

Is that hard to understand?

 

[IDEngineer]

No, what is difficult to understand is how DISM 1) changes additional files without knowing where they are, or 2) doesn't report an error if it can't find and alter those additional files.

Just trying to understand the how and why. I'll give it a test and report back. Thanks!

 

[IDEngineer]

Credit where credit is due: My M.2 drive is now visible and Win7 is "Expanding Windows files" as I type this! THANK YOU!


There's still a logical disconnect in here for me. When I installed the USB3 drivers I only copied the modified .wim files back to the USB stick. DISM didn't have access to the USB stick, and only the .wim files were copied to my hard drive for DISM work and then back to the USB stick afterwards. Yet the USB3 installation worked perfectly - my keyboard and mouse started working through USB3 ports.


When I tried that exact same process for the NVMe update, it didn't work and obviously your comments about having to copy the (new) contents of boot.wim/2/sources over the top of the contents of USBstick/sources was the difference.


The only thing I can think of is that perhaps DISM /add-driver doesn't modify the /sources files, while DISM /add-package does.


In any case, we are definitely making progress at last. A huge THANK YOU for your advice, you really know your Win7. Looking forward to reporting back once the install is finished and successful.

 

[IDEngineer]

I haven't yet installed the Asus-specific drivers for the hardware on this X270 board, but the preliminary mass storage numbers are pretty impressive for PCIe 4.0 SSD (Samsung 980 Pro) running in an PCIe 3.0 environment.


Using CrystalDiskMark, M.2 read/write speeds are ~3400/~3400 MBps sequential, 1580/622 MBps random. This compares with 525/198 sequential and 495/200 random on the SSD in my main machine, so we're seeing a substantial uptick.


Online reports suggest that PCIe 4.0 could yield another ~30% in effective throughput but compatibility with Win7 was the primary requirement so I'll have to leave that on the table. I don't know if I can expect any further increase once the board-specific drivers are installed.


Again, my thanks to user SIW2 for guidance on getting NVMe support in Win7! I've marked this thread "Solved" and ticked your positive reputation.

 

[SIW2]