Finally a solution to the not supported AMD APU

[Menard]

and with no customization nor settings to add ...

With this quadrocore APU the AMD Turbo core was not active, even with AMD Turbo Core set as "enabled" in the Bios, and so instead of 3.8 GHz I add only 3.1 Ghz as maximum

The solution is to add a graphic card with a proprietary driver (Nvidia) and disable the APU Graphic part in the bios and then with Auto setting for the Turbo Core, it uses it when required
I ve seen it by controlling the hardinfo benchmarks ... that increase by around 20% for the most ones, and 8 to 200% for the rest, power consumption on 6 days increased by 36%

but a problem, why the 3D benchmarks are so bad compared with the APU that is as old as the grapĥic card ? (driver Nvidia v340) most are 60 to 75 % bader and particularly it crashes the glmark2 benchmark and is not supported by Unigine Superposition benchmark

NB : to people who could say that it is not a Linux issue but a issue due to my motherboard A88XM v2 MSI , no this no support of some APUs is well known and was the subject of topics here and there

[djph]

If you installed a graphics card that is WORSE than the card on-die with the APU, your benchmarks make sense.

Seems a lot of trouble to convince the motherboard "yes, really, give me the last 700 MHz turbo-boost ... " Honestly, even with my i7's "turbo" feature (and the governor set to allow it), I find that so few things are cpu bound that the CPU nearly always is hanging out in one of the 1.x GHz speed steps on 3 out of the 4 cores.

[Menard]

Yes yoû are right on this part https://gpu.userbenchmark.com/Compare/N ... 899vsm8791 I had not integrated this enough ... and because in part at first glance it is difficult to see so huge difference for 2 devices of the same year, a "2" factor in speed

Few things CPU bound, ok ... why not ? I don't really understand the following part, you experience CPU hanging out in some task in spite of Turbo Boost ? perhaps a thermal issue ...

[djph]

Few things CPU bound, ok ... why not ?

"CPU Bound" means that the process in question is "limited" by the CPU. These days, that is rarely the case for "normal desktop use" -- stuff like editing a paper, or internet browsing, zoom meetings, or anything else that the "majority" of computer users tend to do (there are certainly things that still are limited by the CPU speed of course; but "most people" won't run into those limits).

I don't really understand the following part, you experience CPU hanging out in some task in spite of Turbo Boost ? perhaps a thermal issue ...

No, sorry. What I mean is that the CPU (i7-1165G7), with the governor set to run it hard (4.7 GHz) results in general usage of one core at 2.6-3.2 GHz; and the other three hanging out bored at between 1.2 and 1.8 GHz. Allegedly the governor can spin down to as little as 400 MHz / core; but that might be on "powersave" mode rather that "performance".

Honestly, there are a lot of days where I'll shut down 3 cores, and knock the remaining core down to a maximum of 700 Mhz. It still works absolutely fine for the majority of "general usage" applications (except teleconferences or games, but spinning the CPU cores back online only takes a few seconds before starting the application in question).

[Menard]

Ok, I understand that you have a good practice, experience of these settings ... and I could try to install what is necessary to control these cores
The fact is also that I don't see a clear difference since my APU's Turbo is unlocked ... except on the power consumption
Now their clocks are 1397 1397 1596 1805 MHz with one internet browser on this page

It could be a good idea to be able to control the cores from the desktop without to have to reboot

[djph]

That looks fine. CPU downclocking when it's not busy has been a feature for at least the last decade.

Go do something to make it busy (e.g. compile something big), and you'll see the cores spin right up.

[Menard]

I have some doubts about this though because I ve watched the power consumption on Windows XP (supposed not supporting quadricores) with all cores at the maximum speed (by default) and it was clearly less than with Linux and all cores at minimum speed

[djph]

I have some doubts about this though because I ve watched the power consumption on Windows XP (supposed not supporting quadricores) with all cores at the maximum speed (by default) and it was clearly less than with Linux and all cores at minimum speed

What do you mean "watched power consumption" ?

[Menard]

What do you mean "watched power consumption" ?

I have a device for this, to plug on the power and I had plugged the wire of the tower on it, but this morning it measures the fridge

[djph]

You'd need the laptop in exactly the same battery charge state, and doing (more or less) exactly the same thing.

If you're comparing mint (charging battery) to windows (fully charged); the power consumption will be significantly higher in mint, thanks to the battery charge circuitry kicking in.

Likewise, if you're comparing mint running flat out to windows idling ...

[Menard]

It is not a laptop, it is a mini ATX tower that stands 7 cm upon the ground (I hesitate to call this a desktop) but the screen is clearly on a desk, and so I am only a little concerned by the consumption,
but power consumption is relied to power supply and its limits, and also it can heat the room ... in this season it may be a problem