Windows 10 runs hotter than windows 7?

[jaclaz]

Windows 8/10 do a maintenance task during idle, which does Defrag, WinSxS compression and NGEN which Win7 doesn't do and this causes extra CPU usage, so the idle temperature is a bit higher.

Well, that what was expected in theory Trip had the opposite experimental results in practice, that is what is perplexing...

 

jaclaz

[NoelC]

What Andre is trying to say is that maybe there's more heat already built up during non-intensive operations, so you see the temps go higher during intensive things such as video rendering.  Heat transfer is a dynamic thing, and can sometimes be non-intuitive (think Chernobyl, for example).

 

Also, two different systems may have completely different cooling characteristics, with one having more reserve capacity than another.  A difference in observed temperatures could be something as simple as the fan speed being increased more on one than the other after the temperature crosses some threshold.

 

-Noel

Edited May 12, 2015 by NoelC

[sdt]

What Andre is trying to say is ....

-Noel

 

 

noel awaiting your results.

[jaclaz]

Sure , but here we have only one data set, recorded by a reliable, experienced member that evidence the opposite.

 

Of course and as said the data is too little to be meaningful, or "common" or "universal" still the (limited) data available tell another story.

 

If we do (arbitrarily BTW) an average of the three different tools we have:

4k Windows 7PRO x64 minCPU Average 47.33

4k Windows 8.1PRO x64 minCPU Average 47.33

4k Windows 10 10074 x64 minCPU Average 49.33

Which puts the "base" to exactly the same level for 7 and 8.1 and only very slightly higher for 10 as (49.33-47.33)=2/47.33=4.22%

 

But if we do the same on the higher temperatures we have:

 

4k Windows 7PRO x64 maxCPU Average 73.00

4k Windows 8.1PRO x64 maxCPU Average 66.33

4k Windows 10 10074 x64 maxCPU Average 64.33

 

so the reduction of temperature is 73.00-66.33=6.67/73=9.14% for Windows 8.1 over 7 and 73.00-64.33=8.67/73=11,87% for 10 over 7.

 

While differences within 5% may well be both "normal" and "random", when they go around 10% they start getting noteworthy.

 

Looking at it in another way, the IE10 on 7 raised the temperature from an "idle" of 44 to a "full throttle" of 70 with an increase of 26 degrees while the IE11 on 10 raised it from the "idle" of 49 (i.e. 10% higher) to a "full throttle" of 58 with an increase of only 9 degrees, almost 1/3 of the raise of temperature in Windows 7.

 

We definitely need more data, but looking at what is available till now and until the above data is disproved or rectified the differences are very noticeable.

 

jaclaz

[NoelC]

 

noel awaiting your results.

 

 

My two stumbling blocks at this point, since I have just one hardware system that dual boots I can test on, are:

 

1.  Acquisition of a complex enough UHD video to test with that shows with the Win 7 standard codecs.  I can't justify risking installation of a possibly questionable codec pack or much other software just to do this test.  It's a live system at this point that just happens to have a second disk drive that will boot up Windows 10 build 10074.

 

2.  Time to take the machine offline, swap the boot order, and perform the tests.  Actually, this afternoon wouldn't be a bad time...

 

I think maybe I have an alternative...  I'll try running Passmark PerformanceTest and record the temperatures.  It kind of measures something different from what is being described in this thread, but it would be interesting nonetheless...

 

-Noel

[NoelC]

noel awaiting your results.

 

OK, I have made some measurements.  I'm afraid they may not help clear much up...

 

I found a YouTube video (Avengers 2 trailer) that will play on both boot setups (Win 7 and Win 10 build 10074).

 

Win 7 package temperature:  35C - 39C

Win 10 package temperature:  45C - 49C

 

Here are the screen grabs of the measurements.  Some of the other things, such as power levels, are quite interesting.

 

 

 

 

Here are the measured temperatures across a run of the PassMark PerformanceTest...  Note that the CPU package got just about as hot, but that Win 7 seems to have a lower base temperature (Andre's comment was likely on the mark).

 

 

 

 

The minimum clock speed numbers are interesting.  I do have the power plan for Win 10 set to run the processor at 5% if it is not needed.  Either the system thinks it's needed or that's not functioning properly.

 

-Noel

Edited May 12, 2015 by NoelC

[Tripredacus]

On the side, regarding temperature in Windows 10 altogether...

I found that the idle temp is lower in Windows 10 while in Audit Mode. It may have been an error in my part as I had done all my tests while in Audit Mode of each OS. I had made an image of the Windows 10 OS after doing the video test, and then did a deployment test to make sure the image worked... because of weird usernames and goofy fonts reasons. After going through OOBE the fans kicked up to an audible level, same type increase I heard when doing HTML5 on Chrome.

[NoelC]

Curious.  What does HWMonitor show now?

 

Notwithstanding what Andre said, I've watched Process Hacker, et. al. and I just don't see whatever background activity it is using the CPU.  I wonder if it really just is a shortcoming in the power management logic that manifests when it's sitting idle.  I guess I'll have to fire up the performance toolkit at some point and make some traces.

 

-Noel

[sdt]

@noel thanks for posting the data. I notice the minimum cpu voltage in windows 10 at 1.083V where as it is 0.6 in windows 7. Something is off in that regard. possibly why the temps are being reported higher in win10.

[Tripredacus]

Curious.  What does HWMonitor show now?

Audit Mode:

"Normal" Mode:

The average temp is a little higher, but while in "Normal" mode, it will spike occasionally. It does not do this in Audit Mode. I will say that this "Normal" test was done without having anything open and having no internet connection.

[NoelC]

I wish you had chosen to expand the Clocks field in the first screen grab.

 

It's possible all of this is because the power management logic has changed a good bit.  Were there cases where a powerful machine refused to speed up with intensive operations with the older OS?  I don't recall coming across any personally, but it's the kind of thing that's hard to sense.  If it takes a few extra seconds to start something that normally starts up instantly, we all just sort of wait it out.

 

-Noel

[Tripredacus]

I wish you had chosen to expand the Clocks field in the first screen grab.

I didn't change anything each time I open the program. I don't know why it did that.

[sdt]

I wish you had chosen to expand the Clocks field in the first screen grab.

 

-Noel

 

 

id guess it would be around 800 mhz since the voltage is similar.

[NoelC]

 

I didn't change anything each time I open the program. I don't know why it did that.

 

 

Probably a minor bug.  I noticed that when I was doing my captures I had to expand some of the subheadings.

 

-Noel