Re: Before Dimdows 10 Goes EOL, I'm Testing Linux To Save My Laptop From The Landfill

On Fri, 6/27/2025 4:11 AM, Lawrence D'Oliveiro wrote:

On Fri, 27 Jun 2025 03:21:09 -0400, Paul wrote:
 

The point is, the machine is on the list ...

 
Not the X220.
 

The processor is in the microsoft-maintained Windows 11 list.

This means the Health check (previous referred to as an
Upgrade Advisor), will give a passing grade to the unit.
No matter what the manufacturer may inadvertently or otherwise,
list for the unit.

The manufacturer responsibility, revolves around "driver provisioning"
and "tech support" for the item. That's why they make certain
assertions on the site.

The company that is dynamically altering the DVD images and
provisioning the Upgrade Advisor info, ultimately determines
whether the OS installation is "friction free" or not.
To the best of my ability, this unit will be a "friction free" one.

*******

On a previous post, I had planned to add a performance section.
The performance info was not ready in time, so I had to dump
that part, and save it for later.

This is the current performance (single thread measurement, to see
to what extent the OS is messing with performance). I had trouble
getting the SuperPI menu to render on Win11, but using the cursor
keys instead of the mouse, solved the problem, and the testing continued.

I tried to isolate for the impact of subsystems in windows, not
by "generating a table of subsystems", as I don't know if there
is any easy readout that actually sniffs and determines they are
engaged.

SuperPI 1.5 XS. Windows 7 is the OS to beat. It holds the record for
best performance, of all the OSes, by a tiny margin.

Times are in minutes and seconds (the way SuperPI displays them).
A lower time is better.

       16M digits   32M digits

win7   3m 54.562s   8m 36.096s
                    8m 33.553s  (repeat run, after the slogging ones below had run)

win10              10m 12.453s  WD ON                    22H2       [Just awful results]
                   10m 35.453s  WD OFF  (yikes!!! WTF, slower ???)
                    9m 56.922s  WD OFF, Affinity Core#2
                    9m 23.094s  WD OFF, Affinity Core#2, truncatememory (2048MB)

win11               9m 32.164s  WD ON                    24H2
       4m 15.237s   9m 26.603s  WD OFF
       4m  1.533s   8m 44.491s  WD OFF, truncatememory 0xC0000000 (3072MB)

You can see there, I had considerably difficulty with the latest OSes.
It wasn't quite this bad, the last time I tested. It's really gotten
quite out of hand. In particular, notice that Windows 10 is worse
than Windows 11, and how is that even remotely possible ???

Windows Defender was turned off for most of the runs.

Setting Affinity to a single core, is to see to what extent
process migration (continuous movement) is affecting performance.
I did not run a full matrix of "Affinity", because there is
no intention on my part, to be suggesting to user that they
use Affinity on a daily basis. It was just an attempt to try
to figure out why Win10 is so bad at the moment.

Because its a rolling release, I would have to submit the topic
to continuous testing, and because I'm not getting paid to do this,
I'm not going to do it. I test things, as part of problem solving
for users needing help.

The purpose of truncating memory, and making the OS boot with
2GB of RAM, or 3GB of RAM, is to cause security-related subsystems to shut down
during boot detection. For example, providing insufficient room
for the Sandbox Image, should cause Sandboxing to stop. Well,
I was surprised to find that on Windows 11 (unlike some
previous tests), it now seems to have some "mandatory" RAM
requirements that were not previously present. I had to bump
win11 up to 3GB, so I could run SuperPI (which uses 270MB of memory
or so). You should have been able to get Windows 11 consumption
to drop below 1GB, by starving it like that. Instead, some part of
the display subsystem died, at 2048MB + SuperPI running.

Summary: You can see that given sufficient whip, Windows 11
         starts to head in the right direction. Whereas I can't
         say what disease currently afflicts Windows 10. Undoubtedly
         a "prize" in the Windows 10 Crackerjack box near EOL. The above
         collected on a 4th gen 4930K processor. When Windows 10 came out,
         it was very close to the Windows 7 numbers of the time.

         I cannot test this on one of my Zen3 systems, because Win7
         is not qualified for anything after SkyLake. And then I couldn't
         run a Win7 baseline, and know it was functioning properly.

         The 4930K is currently the box that covers the widest range
         of OSes. The previous box for doing that, the motherboard failed,
         and the Test Machine is the successor.

   Paul