Framework 13 With AMD Ryzen 7040 Series Makes For A Great Linux Laptop

[autonomous_cowherd]

3:2 2256 x 1504

[Mitch]

Two questions:

• How has the battery life measured under regular workloads?
• Does the fan still get noticeably loud under regular usage?

I have the same machine and similar config and I run on Fedora KDE Wayland and haven't really done much tuning. Idles are pretty good. I think I got close to 3W with medium brightness (the screen is probably most of that) and a few things open, like Firefox, but not being used. However, it can draw 88W from the wall when it's being hammered with tasks and charging the battery at the same time on Balanced. I haven't extensively used or tested it yet. Setting to the Powersave profile lowered wattage down to about 75W in the above scenario.

Fans are normally silent or near-silent unless you push the CPU or GPU hard for a long time. Installing a crap ton of Steam games at the same time and compiling shaders will get the fans spinning. And they do get loud if you let the CPU hit its peak for long enough. Fortunately, the fans are not shrill when they're topping out. It's lower pitch, which is far less unpleasant to my ears. My lenovo has a shrill fan, and it's like nails on a chalkboard.

I have a lot of stuff I want to try with 6.6, which will introduce more tunables for CPU behavior and efficiency. It also has tons of other improvements I'd love to see like EEVDF, which should keep things more responsive than otherwise at lower clockspeeds (I like capping my speeds on battery).

[rcalixte]

Fans are normally silent or near-silent unless you push the CPU or GPU hard for a long time. Installing a crap ton of Steam games at the same time and compiling shaders will get the fans spinning. And they do get loud if you let the CPU hit its peak for long enough. Fortunately, the fans are not shrill when they're topping out. It's lower pitch, which is far less unpleasant to my ears. My lenovo has a shrill fan, and it's like nails on a chalkboard.

Thanks for this. I often compile software and other intensive tasks but unless I switch to the power-saving profile, the fans get noisy rather quickly. I'm currently on the 12th Gen Intel but antsy to get something better if it exists. Even running a flatpak update will get the fans churning under a balanced profile after a few seconds.

[OlafLostViking]

I (BIOS 3.03) do have the issue that it wakes up from sleep when you
* open the lid
* close(!) the lid
* add power source
* remove power source

Therefore you cannot close your laptop, disconnect it and put it in your bag. Or put it manually to sleep (which I prefer due to bad experiences in the past where the machine didn't suspend on closing the lid and was heating up the backpack) and then close it. Or disconnect it from the charger when it's sleeping. Or connect it to a charger while it's sleeping if you want to keep it sleeping. And since I have one big power adapter where I load all devices over night, the machine gets woken up during the night as soon as a device finished loading and the adapter renegotiates voltages.

It's known and discussed in the forum, but there is no solution yet. Ideally it would be configurable in a later BIOS update. Let's see if this can be worked around with some udev rules. But I'm not sure I should trust that. ;-)

[thongpv87]

Alder Lake totally sucks in MT scenarios. Hopefully Meteor Lake will turn out to be actually power efficient.



Bad BIOS. I use ryzenadj to bring idle power consumption to around 0.7W. Too bad this means I had to disable Secure Boot.

There's a long discussion about HP fuckery here: https://h30434.www3.hp.com/t5/Notebo...d/td-p/8731252

Wow, that's impressive. That's nearly equal to my power consumption under suspension. Would you mind sharing your tunning process?
I also wonder whether the CPU can be undervolt?

[Mitch]

I (BIOS 3.03) do have the issue that it wakes up from sleep when you
* open the lid
* close(!) the lid
* add power source
* remove power source

Therefore you cannot close your laptop, disconnect it and put it in your bag. Or put it manually to sleep (which I prefer due to bad experiences in the past where the machine didn't suspend on closing the lid and was heating up the backpack) and then close it. Or disconnect it from the charger when it's sleeping. Or connect it to a charger while it's sleeping if you want to keep it sleeping. And since I have one big power adapter where I load all devices over night, the machine gets woken up during the night as soon as a device finished loading and the adapter renegotiates voltages.

It's known and discussed in the forum, but there is no solution yet. Ideally it would be configurable in a later BIOS update. Let's see if this can be worked around with some udev rules. But I'm not sure I should trust that. ;-)

I haven't had that problem so far with the FW13 with 7840u, but I've only had the machine for a day. My Lenovo with a 6850u does have that problem, but strangely didn't have it before I reinstalled Fedora.
Both machines use Fedora KDE Wayland.

EDIT: It may be because my new Fedora install has some different sleep and hibernate settings than before. I can't recall what they were before the wipe, but I imagine something there is the culprit, especially after seeing a patch about issues during Suspend + Hibernate that has yet to hit the mainline kernel.

[ThomasD]

Is the embedded controller somehow protected by secure boot? Is it possible to detect changes to the EC?
Imaging an attacker has access to your laptop and changes the EC, what can he do with this?
Since the EC is persistent, there is a lot of risk, as I understand it.

According to google, the EC even has modules which interact with secure boot [1].

[1] https://chromium.googlesource.com/ch...HEAD/README.md

[JMurph2015]

Can someone explain two things to me:

1. Why was the Framework 13 tested with a different CPU governor than the Thinkpad P14s which uses the exact same CPU SKU?
2. Why, when given less and slower memory (DDR5-5600 vs LPDDR5X-6400), does the Framework appear to outperform the Thinkpad P14s?
3. Are there any graphics tests available? I've heard Phoenix APUs are sensitive to the memory latency and throughput for iGPU performance.

Overall, just a confusing performance there from these two. The power draw advantage goes back and forth, and the results similarly are much more mixed than expected.

Well for one, I've found that theoretically, the Framework's cTDP is 51W/35W while the Thinkpad's cTDP is supposedly 30W/24W. That means it's a tradeoff between a lot more available power consumption vs faster memory. But I'm not sure if that's the whole story.

[JanW]

Thanks for the great artlice! I've been anxiously waiting to see these benchmarks. Now benching my framework to compare...

One thing I don't know how to do is to get power consumption numbers. I did install powercap and raplcap but their command line utilities don't give me current power:

localhost:~> sudo rapl-info -v
Zone 0
 name: package-0
 enabled: 0
 max_energy_range_uj: 65532610987
 energy_uj: 27147948072
 max_power_range_uw: No such file or directory
 power_uw: No such file or directory
 Subzone 0
   name: core
   enabled: 0
   max_energy_range_uj: 65532610987
   energy_uj: 1988246513
   max_power_range_uw: No such file or directory
   power_uw: No such file or directory
​

Or is energy_uj sufficient (I guess if polling this regularly and with precise timing, power can be derived).
The Git of these tools cites Intel processors as a prerequisite, though...

[mmcdonnell]

I would love to see a gaming benchmark. The GPU should be powerful enough to play less demanding games.