Jump to content
Strawberry Orange Banana Lime Leaf Slate Sky Blueberry Grape Watermelon Chocolate Marble
Strawberry Orange Banana Lime Leaf Slate Sky Blueberry Grape Watermelon Chocolate Marble

MSFN is made available via donations, subscriptions and advertising revenue. The use of ad-blocking software hurts the site. Please disable ad-blocking software or set an exception for MSFN. Alternatively, register and become a site sponsor/subscriber and ads will be disabled automatically. 


Sign in to follow this  
bluebolt

XP Pro x64 OS Boot NVMe

Recommended Posts

I installed the Windows XP Professional x64 operating system on a Samsung 970 Pro NVME SSD drive using an ASRock X99 Extreme4 motherboard.  Really easy the second time.

Download Kai Schtrom's driver here:

https://sourceforge.net/projects/nvme-for-windows-2003-server/files/

Download KB932755 here:

https://www.microsoft.com/en-us/download/details.aspx?id=11619

Download AOMEI Backupper Standard here:

https://www.aomeitech.com/download.html

Use diskpart in Windows 7 x64 to create the aligned partition on, and format, the NVMe drive (the Windows 7 OS must have an NVMe driver installed, or diskpart will not be able to see the NVMe drive).

Use Device Manager in XP x64 to install Schtrom's driver, which makes the NVMe SSD available as a storage drive -- and the XP x64 operating system as such now includes the NVMe driver.  Run the storport hotfix.

Use AOMEI Backupper to clone the XP x64 disk to the NVMe disk (remember to align the partition).  Reboot into the BIOS and set the NVMe drive to boot first.
 

X99-x64-NVMe-Boot-OS.png

 

I don't know whether Kai Schtrom had Windows XP 64-bit in mind when he wrote his driver, but for those of you running that OS it does present the potential for significant performance improvement, to say the least.

  • Like 4

Share this post


Link to post
Share on other sites

I'm currently working to overhaul my desktop machine, which uses XP64. I've got an ASUS SABERTOOTH TUF X99 mobo I'll be using and apparently it can support NVMe SSDs, although I do see one person saying that the drives get zero airflow and heat up to 80 C / 176 F when using the mobo's native connection, which ... does not sound ideal.

Will this driver work with the SABERTOOTH X99s? Also, what would you recommend in order to TRIM? We have a thread on that for XP32, but I don't know if what software would be recommended for XP64, especially for this type of SSD.

Share this post


Link to post
Share on other sites

@TrevMUN

I've only used SATA III Intel SSDs and the Intel SSD Toolbox, so this is new to me.  Schtrom tested on a Samsung, so I thought it the best chance for success, and the 970 Pro very fast for the price.

Have to consider the driver experimental -- XP x64 can't even format the NVMe. And for some reason, when the driver is integrated into the installation CD using nLite, setup still can't see the NVMe drive (hence use of the clone method).  Your guess is as good as mine as to whether it will work on the Sabertooth, probably a lot better chance as a secondary drive...

After booting and running fine initially, I turned mine on yesterday and waited 20 minutes for the desktop, with the system barely moving after that.  And when I say barely moving, I mean glacial, which is frustrating because of the time it will take to work things out...half the morning today just to check Task Manager, Event Viewer and such.  I've tried some things outside of Windows, such as running the clone process over again, but so far that first time was the charm.

Share this post


Link to post
Share on other sites

Looks like this may be solved...

First off, in Event Viewer neither /Application nor /System showed any problems (except for the expected hangs owing to the molasses-in-wintertime mode the machine was "running" in).  Task Manager showed the usual System Idle Process 99.

I ran CrystalDiskMark, which took a very long time -- I expected terrible readings, but instead the top-line read and write marks were 1607/2325.  In other words, the read mark was about half-speed, and the write mark was up to spec.  Based on that, the machine should still have been lightning fast.

For some reason the half-speed mark made me think of hyper-threading, so I turned off hyper-threading in the BIOS -- and the machine booted up and ran normally.

The internet said if you don't install Server 2003 with hyper-threading turned on, it won't run right if you turn it on after the fact.  I never knew of such a thing.

This is really my Windows 2000 Professional machine (where I don't use hyper-threading), so I think I had hyper-threading off when I installed XP x64, then enabled it later without realizing the consequences.

So the next test will be a new installation of XP x64 with hyper-threading enabled during OS installation.

Share this post


Link to post
Share on other sites

I ran a new test with hyperthreading enabled during OS installation, but to no effect:  it remains that this driver/NVMe combination boots normally with hyperthreading disabled, but unreliably with hyperthreading enabled.

The NVMe driver was built for Windows Server 2003, so maybe this characteristic arises because...

Quote

"Windows Server 2003 does not have the functionality to detect hyperthreading processors and multicore processors."

https://support.microsoft.com/en-us/help/932370/the-number-of-physical-hyperthreading-enabled-processors-or-the-number

...not that it matters:

X99-x64-NVMe-Boot-OS-2.png

Share this post


Link to post
Share on other sites

Sometimes I get the OS to boot normally with hyperthreading enabled by introducing some change in the BIOS settings (such as enabling Fast Boot), so in one such instance I took the opportunity to benchmark.

Hyperthreading enabled:

NVMe-hyperthread-Crystal-Mark-03012019.p

As usual, XP performed better with HT enabled; too bad it won't stick.

I haven't been able to get a grip on the inconsistency--seems like with hyperthreading enabled it should either boot normally, or not.

To be clear, the setup always boots--I have not experienced a single blue-screen or failed boot in weeks of testing.  It's just a matter of whether hyperthread mode boots and runs at normal speed (which is to say, very fast), or hundreds of times slower.

I hope to test a new setup with a different NVMe drive in a few weeks.

Share this post


Link to post
Share on other sites

I tested a Western Digital WD Black SN750 NVMe SSD WDS100T3X0C.  It would not boot reliably even with hyper-threading disabled, and not at all with HT enabled.  You can tell there will be trouble with the boot speed when this screen, rather than passing within a few moments, just scans and scans:

first-sign-or-trouble.jpg

So I went back into the BIOS and disabled "Hardware Prefetcher" and there was a great increase in boot reliability (I also disabled "No-Execute Memory Protection" as that seemed to help, too):

BIOS-CPU-Configuration.png

So right now, the machine is booting much better with hyper-threading enabled, though either mode remains hit-and-miss on the WD NVMe drive, performing normally here:

WD-NVMe-HT-bench.png

The Samsung 970 NVMe drive is no longer available for testing, so I can't re-check that, but disabling Hardware Prefetcher has helped a lot.  I plan to test a different NVMe drive in a week or two.

Share this post


Link to post
Share on other sites

The system now consistently boots timely:  the solution was to enable Intel SpeedStep in the BIOS.

I disable SpeedStep for Windows 2000 and am nonplussed that the combination of SpeedStep/Disabled and XP x64/NVMe would affect boot reliability, but it does in this setup.

Disabling Hardware Prefetcher and Adjacent Line Prefetcher didn't sit well with me because of the performance hit it caused, so it's nice to be able to run with those features enabled now.  Current tests are with the Intel 760p Series 1TB NVMe (SSDPEKKW010T8X1).

Default voltage for the i7-6950x CPU is 0.976 V.  This motherboard has integrated overclock settings up to 4.0 GHz, at which ASRock sets voltage at 1.241 V (according to CPU-Z).  I was able to manually raise the clock to 4.3 with no other changes, and low temps.  So it seemed a couple of more clicks would be easily attainable with increased voltage, but alas (in my admittedly limited testing), 4.3 was a hard wall.

Top-line CrystalMark scores include one lucky shot that cracked the million barrier, establishing, at least, that the gauge doesn't break there:

OVERCLOCK-XPx64.png

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×