Question W11 22000 ----> 22621

[Tech Junky]

So, being keen to keep things up to date I've been thinking about how to tackle moving to the 22H2 version for awhile now. I had been looking for a simple KB patch that would accomplish this rather than burning an ISO or wiping things and doing a fresh install. Well MSFT doesn't make it that simple for some reason. I recently installed the latest 22000 patch KB5021234 and had attempted the 22H2 patch KB5021255 which doesn't work with 21H2.

Somewhere along the line I had installed a different update to something else and on a reboot got no desktop / black screen but, CTRL / Shift / ESC still brought up task manager but couldn't access anything else. I went into safe mode and reversed the last "update" and got back to a functional desktop and able to access things. I went looking to see how I could upgrade the base kernel to the newest option.

Well, of course this ended up being a combo of WindowsPCHealthCheckSetup as required by Windows11InstallationAssistant.

Ran the install assistant and it downloaded an ISO from the size of the file being 3GB+ and then went through its dumb processes. Rebooted a few times and then came up and working. It took a couple more reboots though to get everything to settle down and function as before. I use a program called Explorer Patcher to fix the issue with moving the taskbar and making it more W10 in appearance and it spazzed out a bit post upgrade.

Anyway, now that it's up and running on 22621.963 after some additional patches it seems to be running cooler and smoother. There seems to be some additional performance improvements in 22 vs 21. Ran a gaming sessions for a couple of hours and normally the laptop takes a bit to cool back down but, the CPU/GPU seemed to be running a bit cooler than before on 21H2. The fans spun down faster and the system cooled back to idle fairly quickly. Looking at the task manager during the game being open appeared that both weren't working as hard with high utilization in comparison as well.

Not exactly sure what MSFT changed in the two releases to make things smoother but, it appears to be slightly better in terms of performance and properly handling ADL CPU's. Between this and the Chrome hack that just got enabled to drop RAM / Energy use this thing should be quiet and cool.

 

[OlyAR15]

You seem to enjoy making things much harder than they are. Upgrading to 22H2 was easy to do on all my machines.

 

[Tech Junky]

You seem to enjoy making things much harder than they are. Upgrading to 22H2 was easy to do on all my machines.

That's a possibility but, there wasn't any automatic option that popped up and it's just how I got here with Windows. I was a bit hesitant anyway with all of the chatter about certain updates breaking stuff w/ 22H2. I don't let Windows automagically do anything though because it breaks stuff. I ran updates post upgrade and it downgraded things like it usually does and had to reapply drivers to their prior version.

 

[igor_kavinski]

Is it possible that before the update, the power plan was set to High Performance and after the update, it's been reset to Balanced or Power Saving plan, and that is the actual reason why everything seems to be running cooler?

 

[Tech Junky]

Nope, I reset it to performance and it still seems to run cooler. And with the game it maxes things out anyway.

 

[igor_kavinski]

If you don't mind and have the time (less than 3 hours), would you be willing to run the following benchmarks on your laptop in Ubuntu 22.04 running inside WSL2 for comparison?

https://openbenchmarking.org/result/2205126-DANI-COMPILA82&sdt=1

It would be amazing if we can see real differences in the power usage graphs.

 

[Tech Junky]

https://openbenchmarking.org/result/2205126-DANI-COMPILA82&sdt=1

I have a feeling this won't look great under WSL just looking at the first couple of test results. On my server though running Linux natively it's doing alright. Looks like the test has a time on it of ~5hours though.

 

[igor_kavinski]

Looks like the test has a time on it of ~5hours though.

Hopefully that's just a worst case estimate.

 

[Tech Junky]

Hopefully that's just a worst case estimate.

Yeah, it's dropping as things progress. The server / 12700K is knocking them out a bit faster by comparison whether by additional P cores or native linux.

 

[Tech Junky]

@igor_kavinski
So, I stripped out all of the tests that error out and narrowed the scope to the ones that produced results instead to minimize the time taken.

phoronix-test-suite benchmark build-apache build-php build-eigen build-imagemagick build-llvm build-ffmpeg build-mplayer build2 build-godot build-nodejs build-gem5

I'm running them but, this is a bit of a PITA.

 

[igor_kavinski]

I'm running them but, this is a bit of a PITA.

Hope it's worth it in the end. What commands did you run to install the Phoronix benchmark? I always get stuck trying to get it to install.

 

[Tech Junky]

Just downloaded the Deb and dpkg it and then when executing the test names it checks for the dependencies and downloads what is needed. It misses some minor files though when triggering some of the tests which is why our results have those blank spots. Of course my Blocklists in pihole prevent some downloads and have to disable and start over to get them to dl.

 

[Tech Junky]

@igor_kavinski After messing with phoronix testing all day and not getting the same outputs to register with the both systems I'm throwing in the towel on it. I even tried the Windows version to get native results but, it doesn't have near the same tasks to run. I even blew away my 2nd drive and installed Linux natively to run them and just couldn't reproduce the same outputs. While in theory it shouldn't have been a complicated issue it seems to not have any consistency other than taking forever to complete. All I can say is with the upgrade to 22H2 it does run more efficiently and cooler. WSL is something to play with though but, I don't really see it being something I would use all that much since I run dual systems already. The tests aren't anything I would normally be doing anyway on either system.

 

[igor_kavinski]

Thanks for the effort! Yeah, my experience benchmarking with Phoronix wasn't that great also. But at the time, I thought maybe I had bad luck.

 

[Tech Junky]

I mean it gives some insight to some things but, the "tests" are scattered for what they measure and it seems more geared towards programmers compiling things. There's no real CPU / GPU / system test that's meaningful after skimming through a bunch of them. Sire, I update the kernel but, I don't bother with building them because they're prebuilt and tested on the kernel.org page. I tried messing with the 6.1 based off source but, whatever I did it just failed to boot with a panic. Sure, I might download some source apps that need to be built but, they take less than a minute typically to compile and install. I do use some NODE stuff and it doesn't take nearly as much horsepower or time to build something compared to that test.

I don't host anything so the apache / php / etc. is just a waste of time since pihole does all of that on its own. The Windows version of the test suite doesn't have any sets that are appealing either for what I do. If some idiot would just come up with a cross platform / OS agnostic benchmark then maybe there would be some true comparison in the results though.

The odd thing is that my 12700K beat some of the H tests but did slightly worse in some of the others. The other issue is the K doesn't have a GPU in it since it's a headless server but the laptop does which messes with the testing as well. I figured you were wanting to see the outputs for the CPU / fans / temps and so on at the bottom of your results and for some reason I just couldn't get that output using WSL or native. So, the rest of it was a waste of time trying to produce that output;

 

[hoangcf]

me too. I use a program called explorer patcher to fix the issue....