Framework twelfth gen laptop computer evaluation

The Framework is a 13.5″ laptop computer physique with swappable components, which
makes it considerably future-proof and positively simply repairable,
scoring an “exceedingly uncommon” 10/10 score from ifixit.com.

There are two generations of the laptop computer’s essential board (each suitable
with the identical physique): the Intel eleventh and twelfth gen chipsets.

I’ve acquired my Framework, twelfth technology “DIY”, system in late
September 2022 and can replace this web page as I’m going alongside within the course of
of ordering, burning-in, establishing and utilizing the system over the
years.

Total, the Framework is an effective laptop computer. I just like the keyboard, the
contact pad, the growth playing cards. Clearly there’s been some good work
accomplished on industrial design, and it is probably the most repairable laptop computer I’ve
had in years. Time will inform, nevertheless it seems to be sturdy sufficient to outlive me
a few years as effectively.

That is additionally one of the highly effective gadgets I ever lay my arms
on. I’ve managed, remotely, extra highly effective servers, however that is the
quickest pc I’ve ever owned, and it suits on this tiny case. It
is a tremendous machine.

On the draw back, there is a little bit of proprietary firmware required (WiFi,
Bluetooth, some graphics) and the Framework ships with a proprietary
BIOS, with presently no Coreboot support. Anticipate to wish the
newest kernel, firmware, and hacking round a bunch of issues to get
decision and keybindings working proper.

Like others, I’ve first discovered important energy administration points,
however many points can really be solved with some configuration. Some
of the growth ports (HDMI, DP, MicroSD, and SSD) use energy when
idle, so do not count on week-long droop, or “full day” battery whereas
these are plugged in.

Lastly, the growth ports are good, however there’s solely 4 of
them. When you plan to have a two-monitor setup, you are probably going to
want a dock.

Learn on for the detailed evaluation. For context, I am shifting from the
Purism Librem 13v4 as a result of it
basically exploded on me. I
had, within the meantime, reverted again to an previous ThinkPad X220, so I
typically evaluate the Framework with that venerable laptop computer as effectively.

This weblog put up has been maturing for months now. It began in
September 2022 and I declared it accomplished in March 2023. It is the
longest single article on this complete web site, presently clocking at
about 13,000 phrases. It can take a median reader a full hour to go
via this factor, so I do not count on anybody to really do
that. This introduction ought to be ok for most individuals, learn
the primary part if you happen to intend to really purchase a Framework. Bounce
across the desk of contents as you see match for after you probably did purchase the
laptop computer, as it’d embrace some essential hints on tips on how to make it work
greatest for you, particularly on (Debian) Linux.

These are issues I want I’d have recognized earlier than shopping for:

1. take into account shopping for 4 USB-C growth playing cards, or at the very least a mixture of 4
   USB-A or USB-C playing cards, as they use much less energy than different playing cards and
   you do need to fill these growth slots in any other case they snag
   round and really feel insecure

2. you’ll probably want a dock or at the very least a USB hub if you’d like a
   two-monitor setup, in any other case you may run out of ports

3. it’s a must to do some severe tuning to get correct (10h+ idle, 10
   days droop) energy financial savings

4. specifically, beware that the HDMI, DisplayPort and
   significantly the SSD and MicroSD playing cards take a big quantity
   energy, even when sleeping, as much as 2-6W for the latter two

5. beware that the MicroSD card is what it says: Micro, regular SD
   playing cards will not match, and whereas there could be full sized one
   ultimately, it is presently solely on the prototyping stage

6. the Framework monitor has an uncommon side ratio (3:2): I
   prefer it (and it matches classic and digital photography aspect
   ratio), nevertheless it may shock you

I’ve the framework! It is setup with a contemporary new Debian bookworm
set up. I’ve ran via numerous checks and burn in.

I’ve determined to make use of the Framework as my each day driver, and had to purchase
a USB-C dock to get my two screens
linked, which was personal journey.

These are the specs of the twelfth gen, basically phrases. Your
construct will in fact range in response to your wants.

• CPU: i5-1240P, i7-1260P, or i7-1280P (As much as 4.4-4.8 GHz, 4+8
  cores), Iris Xe graphics
• Storage: 250-4000GB NVMe (or carry your personal)
• Reminiscence: 8-64GB DDR4-3200 (or carry your personal)
• WiFi 6e (AX210, vPro elective, or carry your personal)
• 296.63mm X 228.98mm X 15.85mm, 1.3Kg
• 13.5″ show, 3:2 ratio, 2256px X 1504px, 100% sRGB, >400 nit
• 4 x USB-C user-selectable growth ports, together with
  • USB-C
  • USB-A
  • HDMI
  • DP
  • Ethernet
  • MicroSD
  • 250-1000GB SSD
• 3.5mm combo headphone jack
• Kill switches for microphone and digicam
• Battery: 55Wh
• Digicam: 1080p 60fps
• Biometrics: Fingerprint Reader
• Backlit keyboard
• Energy Adapter: 60W USB-C (or carry your personal)
• ships with a screwdriver/spludger
• 1 12 months guarantee
• base value: 1000$CAD, however would not offer you a lot, typical builds
  round 1500-2000$CAD

That is the precise construct I ordered. Quantities in CAD. (1CAD =
~0.75EUR/USD.)

Base configuration

Customization

Enlargement Playing cards

Equipment

Complete

Professionals

• simply repairable (full with QR codes pointing to repair
  guides!), the eleventh gen acquired a 10/10 score from
  ifixit.com, which they name “exceedingly uncommon”, the twelfth gen
  has an analogous {hardware} design and would in all probability fee equally

• replaceable motherboard!!! will be reused as a NUC-like system, with a
  3d-printed case, twelfth gen board will be purchased standalone and
  retrofitted into an eleventh gen case

• not a passing fad: they made a primary laptop computer with the eleventh gen Intel
  chipset in 2021, and a second motherboard with the twelfth Intel
  chipset in 2022

• 4 modular USB-C ports which might match HDMI, USB-C (pass-through,
  can present energy on each side), USB-A, DisplayPort, MicroSD,
  exterior storage (250GB, 1TB), active modding community

• good energy led indicating energy stage (charging, charged, and many others) when
  plugged

• test account on fwupd.org, “expressed curiosity to port to
  coreboot” (in response to the Fedora developer) and are testing
  firmware updates over fwupd, present on LVFS testing, however
  including for the 12th gen, newest BIOS (3.06) was shipped
  via LVFS

• excellent documentation of the (proprietary) BIOS

• express Linux help with install guides, though you may
  must stay with a little bit of proprietary firmware, and never every little thing
  works accurately

• the eleventh gen had good opinions: Ars Technica, Fedora
  developer, iFixit teardown, phoronix, wonderful
  keyboard and contact pad, in response to Linux After Dark, most
  exciting laptops I’ve ever broken (Cory Doctorow) ; extra
  essential evaluation from an OpenBSD developer

• the EC (Embedded Controller) is open source so in fact
  persons are hacking at it, some documentation on what’s
  possible (e.g. altering LED colours, fan curves, and many others), see
  also

Cons

• the eleventh gen is out of inventory, apart from the higher-end CPUs, which
  are a lot much less reasonably priced (700$+)

• the twelfth gen has compatibility points with Debian, followup within the
  DebianOn page, however principally: brightness hotkeys, power
  management, wifi, the webcam is okay regardless that the
  chipset is the infamous alder lake as a result of it does not have
  the fancy camera; most points presently appear solvable, and
  upstream is working with mainline to get their shit working

• twelfth gen may need issues with thunderbolt docks

• they used to have some issue maintaining with the orders: first
  two batches shipped, third batch offered out, fourth batch ought to have
  shipped in October 2021. they typically appear to maintain up with
  delivery. replace (august 2022): they rolled out a second line of
  laptops (twelfth gen), first batch shipped, second batch shipped
  late, September 2022 batch was usually on time, see this
  spreadsheet for a crowdsourced effort to track those
  provide chain points appear to be underneath management as of early 2023. I
  received the Ethernet growth card shipped inside every week.

• in comparison with my earlier laptop computer (Purism Librem
  13v4), it feels unusually
  bulkier and heavier; it is really lighter than the purism (1.3kg
  vs 1.4kg) and thinner (15.85mm vs 18mm) however the design of the
  Purism laptop computer (tapered edges) makes it really feel thinner

• no area for a 2.5″ drive

• fairly shiny LED round energy button, however will be dimmed within the
  BIOS (not low sufficient to my style) I received used to it

• fan quiet when idle, however will be noisy when operating, for instance if
  you max a CPU for some time

• battery described as “mediocre” by Ars Technica (above), confirmed
  poor in my checks (see under)

• no RJ-45 port, and makes an attempt at designing ones are failing
  as a result of the modular plugs are too skinny to suit (in response to Linux
  After Dark), so unlikely to have one sooner or later
  Replace: they cracked that nut and ship an 2.5 gbps Ethernet
  expansion card with a realtek chipset, with none
  firmware blob

• a bit dear for the efficiency, particularly when in comparison with the
  competitors (e.g. Dell XPS, Apple M1)

• twelfth gen Intel has glitchy graphics, looks like Intel hasn’t totally
  landed correct Linux help for that chipset but

A breeze.

Accessing the board

Unhealthy QR codes

In any case, they should level to one thing and meaning a URL, a
operating web site that may reply these requests endlessly. I guess these
will break ahead of later and in reality, so far as I can inform, they
simply do not work in any respect. I desire the strategy taken by the MNT reform
right here which designed (with the 100 rabbits folks) an actual paper
handbook (PDF).

The primary QR code that is instantly seen from the again of the
laptop computer, in an growth wire slot, is a 404. It appears to be some
serial quantity URL, however I can not really inform as a result of, effectively, the web page
is a 404.

I used to be anticipating that bar code to steer me to an introduction web page,
one thing like “tips on how to setup your Framework laptop computer”. Help really
confirmed that it ought to level a quickstart guide. However in a
weird twist, they in some way despatched me the URL with the plus (+) indicators
escaped, like this:

https://guides.body.work/Information/Framework+Laptop computer+DIY+Version+Fast+Begin+Information/57

… which Firefox instantly transforms in:

https://guides.body.work/Information/Framework/+Laptop computer/+DIY/+Version/+Fast/+Begin/+Information/57

I am puzzled as to why they’d ship the URL that approach, the right URL
is in fact:

https://guides.body.work/Information/Framework+Laptop computer+DIY+Version+Fast+Begin+Information/57

(They’ve additionally “let the workforce learn about this for suggestions and assist
resolve the issue with the hyperlink” which is a help code phrase for
“ha-ha! nope! not my downside proper now!” Belief me, I do know, my very own
code phrase is “are you able to please make a ticket?”)

Seating disks and reminiscence

Keyboard and contact pad

cat > /and many others/X11/xorg.conf.d/40-libinput.conf <<EOF
Part "InputClass"
      Identifier "libinput contact pad catchall"
      MatchIsTouchpad "on"
      MatchDevicePath "/dev/enter/occasion*"
      Driver "libinput"
      Choice "Tapping" "on"
      Choice "TappingButtonMap" "lmr"
EndSection
EOF

Energy button

Another is to make the facility button do one thing else. With
systemd-managed methods, it is really fairly simple. Add a
HandlePowerKey stanza to (say)
/and many others/systemd/logind.conf.d/power-suspends.conf:

[Login]
HandlePowerKey=droop
HandlePowerKeyLongPress=poweroff

You may need to create the listing first:

mkdir /and many others/systemd/logind.conf.d/

Then restart logind:

systemctl restart systemd-logind

And the facility button will droop! Lengthy-press to energy off would not
really work because the laptop computer instantly suspends…

Be aware that there is in all probability half a dozen different methods of doing this,
see this, this, or that.

Particular keybindings

It seems to be like these are defined in the microcontroller so it
could be attainable so as to add some. For instance, the SysRq key
is virtually sure to fn s in there.

Be aware that almost all different shortcuts like this are clearly documented
(quantity, brightness, and many others). One key that is much less apparent is
F12 that solely has the Framework emblem on it. That really
calls the keysym XF86AudioMedia which, apparently, does
completely nothing right here. By default, on Home windows, it opens your
browser to the Framework website and, on Linux, your “default
media participant”.

The keyboard backlight will be cycled with fn-space. The
dimmer model is dim sufficient, and the keybinding is simple to seek out in
the darkish.

A skinny elephant could be carried out with alt
PrtScr (above F11) KEY, so for
instance alt fn F11 b
ought to do a tough reset. This comment suggests it’s worthwhile to maintain
the fn solely if “operate lock” is on, however that is
really the alternative of my expertise.

Out of the field, a number of the fn keys do not work. Mute,
quantity up/down, brightness, monitor adjustments, and the airplane mode key
all do principally nothing. They do not ship correct keysyms to Xorg at
all.

It is a known problem and it is associated to the truth that the
laptop computer has mild sensors to regulate the brightness
routinely. By some means a few of these keys (e.g. the brightness
controls) are supposed to point out up as a special enter system, however
do not appear to work accurately. It looks like the answer is for the
Framework workforce to put in writing a driver particularly for this, however so far no
progress since July 2022.

Within the meantime, the flamboyant performance will be supposedly disabled with:

echo 'blacklist hid_sensor_hub' | sudo tee /and many others/modprobe.d/framework-als-blacklist.conf

… and a reboot. This resolution can also be documented in the upstream
guide.

Be aware that there is one other resolution flying round that fixes this by
changing permissions on the input device however I have not examined
that or seen affirmation it really works.

Kill switches

The Framework has two “kill switches“: one for the digicam and the
different for the microphone. The digicam one really disconnects the USB
system when turned off, and the mic one appears to chop the circuit. It
would not present up as muted, it simply stops feeding the sound.

Each kill switches are round the primary digicam, on high of the monitor,
and fairly discreet. Then flip “pink” when enabled (i.e. “pink” means
“turned off”).

Monitor

Display screen backlight

sep 29 22:09:14 angela i3[5661]: No outputs have backlight property

Part "System"
    Identifier  "Card0"
    Driver      "intel"
    Choice      "Backlight"  "intel_backlight"
EndSection

A number of monitor help

However nonetheless, this can be a downside. I am going to in all probability want a powered USB-C dock
and higher screens, presumably with some Thunderbolt chaining, to
save but extra ports.

However meaning more cash into this setup, argh. And determining my
monitor state of affairs is the sort of factor I am not that huge
of a fan of. And neither is looking for USB-C (or is it Thunderbolt?)
hubs.

My regular autorandr setup would not work: I’ve tried saving a
profile and it would not get autodetected, so I additionally first must do:

autorandr -l framework-external-dual-lg-acer

The magic:

autorandr -l horizontal

… additionally works effectively.

The worst downside with these screens proper now’s that they’ve a
radically smaller decision than the primary display on the laptop computer, which
means I must reset the font scaling to regular each time I change
backwards and forwards between these screens and the laptop computer, which suggests I
really want to do that:

autorandr -l horizontal &&
eho Xft.dpi: 96 | xrdb -merge &&
systemctl restart terminal xcolortaillog background-image emacs &&
i3-msg restart

Sort of disruptive.

Enlargement ports

There’s solely 4 growth ports. After getting two screens, the
drive, and energy plugged in, bam, you are out of ports; there’s nowhere
to plug my Yubikey. So if that is going to be my each day driver, with a
twin monitor setup, I’ll want a dock, which suggests extra crap firmware
and uncertainty, which is not nice. There are literally plans to make a
dual-USB card, however that’s blocked on designing an actual
board for this.

I can not wait to see extra growth ports produced. There is a ethernet
expansion card which rapidly went out of inventory principally the day
it was introduced, however was ultimately restocked.

I want to see a correct SD-card reader. There is a MicroSD card
reader, however that clearly would not work for regular SD playing cards, which
could be extra broadly suitable anyhow (as a result of you possibly can have a
MicroSD to SD card adapter, however I’ve by no means heard of the
reverse). Somebody really found a SD card reader that fits and
then someone else managed to cram it in a 3D printed case, which
is sort of wonderful.

Nonetheless, I actually like that concept that I can carry all these little
adapters in a pouch once I journey and might principally do something I
need. It does imply I must shuffle via them to seek out the best
one which is just a little annoying. I’ve an elastic band to maintain them
lined up so that every one the ports present the identical facet, to make it simpler
to seek out the best one. However that rapidly will get undone and as an alternative I’ve
a pouch stuffed with growth playing cards.

One other superior factor with the growth playing cards is that they do not simply
work on the laptop computer: something that takes USB-C can take these playing cards,
which suggests you should utilize it to attach an SD card to your cellphone, for
backups, for instance. Heck, you can even join an exterior show
to your cellphone that approach, assuming that is supported by your cellphone of
course (and it in all probability is not).

The growth ports do take up some energy, even when idle. See the
power management part under, and significantly the power usage
tests for particulars.

USB-C charging

I couldn’t find documentation about this, however the forum
answered.

That is one thing of a recurring theme with the Framework. Whereas it
has a very good knowledge base and repair/setup guides (and the
forum is awesome) nevertheless it would not have a very good “proprietor guide” that
reveals you the totally different components of the laptop computer and what they do. Once more,
one thing the MNT reform did well.

One other factor that persons are asking about is an external sleep
indicator: as a result of the facility LED is on the primary keyboard meeting,
you do not really see whether or not the system is lively or not when the
lid is closed.

Lastly, I questioned what happens when you plug in multiple power
sources and it seems the cost controller is definitely fairly
sensible: it is going to choose the very best energy supply and use it. The one
draw back is it could’t use a number of energy sources, however that looks like
a bit a lot to ask.

These issues additionally work:

• webcam: splendid, greatest webcam I’ve ever had (however my requirements are
  actually low)
• onboard mic: works effectively, good achieve (possibly a bit a lot)
• onboard audio system: sound okay, just a little metal-ish, loud sufficient to be
  annoying, see this thread for benchmarks, apparently fairly
  good audio system
• combo jack: works, with slight hiss, see under

There’s additionally a lightweight sensor, nevertheless it conflicts with the keyboard
brightness controls (see above).

There’s additionally an accelerometer, nevertheless it’s off by default and
might be removed from future builds.

Combo jack mic checks

The Framework laptop computer ships with a combo jack on the left facet, which
permits you to plug in a CTIA (source) headset. In human
phrases, it is a system that has each a stereo output and a mono enter,
sometimes a headset or ear buds with a microphone someplace.

It works, which is best than the Purism (which solely had audio
out), however is on par for the course for that sort of onboard
{hardware}. Due to electrical interference, such sound playing cards very
typically get a lot of noise from the board.

With a Jabra Evolve 40, the built-in USB sound card generates
principally zero noise on silence (invisible all the way down to -60dB in Audacity)
whereas plugging it in straight generates a stable -30dB hiss. There is
a noise-reduction system in that sound card, however the distinction is
nonetheless fairly putting.

On a comparable setup (curie, a 2017 Intel NUC), there’s
additionally a his with the Jabra headset, nevertheless it’s quieter, extra within the order
of -40/-50 dB, a noticeable distinction. Curiously, testing with my
Mee Audio Professional M6 earbuds results in just a little extra hiss on curie, extra on
the -35/-40 dB vary, near the Framework.

Additionally word that one other sound card, the Antlion USB adapter that comes
with the ModMic 4, additionally offers me fairly near silence on a quiet
recording, choosing up lower than -50dB of background noise. It is
really in all probability choosing up the followers within the workplace, which do make
audible noises.

In different phrases, the hiss of the sound card constructed within the Framework
laptop computer is so loud that it makes extra noise than the quiet followers within the
workplace. Or, one other strategy to put it’s that two USB sound playing cards (the
Jabra and the Antlion) are in a position to choose up ambient noise in my workplace
however not the Framework laptop computer.

See additionally my audio web page.

Efficiency checks

Compiling Linux 5.19.11

5411.85user 673.33system 1:37:46elapsed 103percentCPU (0avgtext+0avgdata 831700maxresident)ok
10594704inputs+87448000outputs (9131major+410636783minor)pagefaults 0swaps

19251.06user 2467.47system 24:13.07elapsed 1494percentCPU (0avgtext+0avgdata 831676maxresident)ok
8321856inputs+87427848outputs (30792major+409145263minor)pagefaults 0swaps

  Time    Person  Good   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Common  87.9   0.0  10.7   1.4   0.1 17.8 6583.6 5054.3 233.0 223.9 233.1  55.96
 GeoMean  87.9   0.0  10.6   1.2   0.0 17.6 6427.8 5048.1 227.6 218.7 227.7  55.96
  StdDev   1.4   0.0   1.2   0.6   0.2  3.0 1436.8  255.5 50.0 47.5 49.7   0.20
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimal  85.0   0.0   7.8   0.5   0.0 13.0 3594.0 4638.0 117.0 111.0 120.0  55.52
 Most  90.8   0.0  12.9   3.5   0.8 38.0 10174.0 5901.0 374.0 362.0 375.0  56.41
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Abstract:
CPU:  55.96 Watts on common with commonplace deviation 0.20
Be aware: energy learn from RAPL domains: package-0, uncore, package-0, core, psys.
These readings don't cowl all of the {hardware} on this system.

memtest86+

Memtest86+ v6.00b3      | twelfth Gen Intel(R) Core(TM) i5-1240P
CLK/Temp: 2112MHz    78/78°C | Cross  2% #
L1 Cache:   48KB    414 GB/s | Take a look at 46% ##################
L2 Cache: 1.25MB    118 GB/s | Take a look at #3 [Moving inversions, 1s & 0s] 
L3 Cache:   12MB     43 GB/s | Testing: 16GB - 18GB [1GB of 15.7GB]
Reminiscence  :  15.7GB  14.9 GB/s | Sample: 
--------------------------------------------------------------------------------
CPU: 4P+8E-Cores (16T)    SMP: 8T (PAR))  | Time:  0:27:23  Standing: Cross     
RAM: 1600MHz (DDR4-3200) CAS 22-22-22-51  | Cross:  1        Errors: 0
--------------------------------------------------------------------------------

Reminiscence SPD Info
----------------------
 - Slot 2: 16GB DDR-4-3200 - Essential CT16G4SFRA32A.C16FP (2022-W23)







                          Framework FRANMACP04
 <ESC> Exit  <F1> Configuration  <House> Scroll Lock            6.00.unknown.x64

Safe boot and EFI

First, I could not boot off of the NVMe drive I transferred from the
earlier laptop computer (the Purism) and the
BIOS was not very useful: it was simply complaining about not discovering
any boot system, with out dropping me in the true BIOS.

At first, I assumed it was an issue with my NVMe drive, as a result of it is
not listed within the compatible SSD drives from upstream. However I
found out how to enter BIOS (press F2 manically, of
course), which confirmed the NVMe drive was really detected. It simply
did not boot, as a result of it was an previous (2010!!) Debian set up with out
EFI.

So from there, I disabled safe boot, and booted a grml image to
attempt to recuperate. And by “boot” I imply, I managed to get to the grml
boot loader which promptly did not load its personal root file system
in some way. I nonetheless have to research precisely what occurred there, however
it failed a while after the initrd load with:

Unable to seek out medium containing a stay file system

This, it seems, was fastened in Debian currently, so a daily GRML
build is not going to have this issues. The upcoming 2022 launch
(probably 2022.10 or 2022.11) may even get the repair.

I did handle in addition the development version of the Debian
installer which was a surprisingly good expertise: it mounted the
encrypted drives and did every little thing fairly easily. It even supplied
me to reinstall the boot loader, however that finally (and accurately, as
it seems) failed as a result of I did not have a /boot/efi partition.

At this level, I noticed there was no simple approach out of this, and I
simply proceeded to utterly reinstall Debian. I had a spare NVMe
drive mendacity round (backups FTW!) so I simply swapped that in, rebooted
within the Debian installer, and did a clear set up. I wished to change
to bookworm anyhow, so I suppose that is accomplished too.

Storage limitations

It may be recoverable, however simply restoring the partition desk
did not work both, so I am unsure how I recuperate the info
there. Usually, every little thing on my laptops and workstations is designed
to be disposable, in order that wasn’t that huge of an issue. I did handle
to recuperate many of the knowledge due to git-annex reinit, however
that was just a little furry.

Bootstrapping Puppet

/and many others/ssh
/var/lib/puppet

In order that I’d preserve the earlier machine’s id. That approach I may
contact the Puppet server and set up no matter was lacking. I used my
Puppet optimization
trick to do a batch
set up after which I had a very good base setup, though not precisely because it
was earlier than. 1700 packages had been put in manually on angela earlier than
the reinstall, and never in Puppet.

I didn’t examine every one individually, however I did undergo /and many others
and copied over extra SSH keys, for backups and SMTP over SSH.

LVFS help

apt set up fwupd-amd64-signed
fwupdmgr refresh
fwupdmgr get-updates
fwupdmgr replace

Good. The twelfth gen BIOS updates, presently (January 2023) beta,
will be deployed via LVFS with:

fwupdmgr enable-remote lvfs-testing
echo 'DisableCapsuleUpdateOnDisk=true' >> /and many others/fwupd/uefi_capsule.conf 
fwupdmgr replace

These directions come from the beta forum post. I carried out the
BIOS replace on 2023-01-16T16:00-0500.

Decision tweaks

The Debian wiki page has a couple of tips for this.

Console

cat >> /and many others/default/console-setup <<EOF
FONTFACE="Terminus"
FONTSIZE=32x16
EOF
echo GRUB_GFXMODE=1024x768 >> /and many others/default/grub
update-grub

Xorg

! 1.5*96
Xft.dpi: 144

! These may also be helpful relying in your monitor and private desire:
Xft.autohint: 0
Xft.lcdfilter:  lcddefault
Xft.hintstyle:  hintfull
Xft.hinting: 1
Xft.antialias: 1
Xft.rgba: rgb

The Debian Wiki has a page on HiDPI, nevertheless it’s not so good as the
Arch Wiki, the place the above blurb comes from. I’m not utilizing the
latter as a result of I believe it is inflicting a number of the “fuzziness”.

TODO: discover the equal of this GNOME hack in i3? (gsettings set
org.gnome.mutter experimental-features
"['scale-monitor-framebuffer']"), taken from this Framework
guide

BIOS configuration

The current fix (as of BIOS 3.03) is to re-enable each Fast
boot and Quiet boot. Presumably that is one thing that may get
fastened in a future BIOS replace.

Be aware that the next keybindings are lively within the BIOS POST
check:

WiFi compatibility points

I could not make WiFi work at first. Clearly, the default Debian
installer would not ship with proprietary firmware (though that might
change soon) so the WiFi card did not work out of the field. However even
after copying the firmware via a USB stick, I could not fairly
handle to seek out the best mixture of ip/iw/wpa-supplicant
(sure, after repeatedly copying a bunch extra packages over to get
these bootstrapped). (Subsequent time I ought to in all probability strive one thing like
this post.)

Fortunately, I had just a little USB-C dongle with a RJ-45 jack mendacity
round. That additionally required a firmware blob, nevertheless it was a single
bundle to repeat over, and with that loaded, I had community.

Finally, I did managed to make WiFi work; the issue was extra on
the facet of “I forgot tips on how to configure a WPA community by hand from the
commandline” than the rest. NetworkManager labored wonderful and received
WiFi working accurately.

Be aware that that is with Debian bookworm, which has the 5.19 Linux
kernel, and with the firmware-nonfree (firmware-iwlwifi,
particularly) bundle.

Battery life

Time    Person  Good   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Common   1.7   0.0   0.5  97.6   0.2  1.2 4684.9 1985.2 126.6 39.1 128.0   7.57
 GeoMean   1.4   0.0   0.4  97.6   0.1  1.2 4416.6 1734.5 111.6 27.9 113.3   7.54
  StdDev   1.0   0.2   0.2   1.2   0.0  0.5 1584.7 1058.3 82.1 44.0 80.2   0.71
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimal   0.2   0.0   0.2  94.9   0.1  1.0 2242.0  698.2 82.0 17.0 82.0   6.36
 Most   4.1   1.1   1.0  99.4   0.2  3.0 8687.4 4445.1 463.0 249.0 449.0   9.10
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Abstract:
System:   7.57 Watts on common with commonplace deviation 0.71

Curiously, enjoying a video in a (720p) window in a window takes up
extra energy (10.5W) than in full display (9.5W) however I blame that on my
desktop setup (i3 + compton)… Undecided if mpv hits the
VA-API, possibly not in windowed mode. Related outcomes with 1080p,
apparently, besides the window struggles to maintain up altogether. Full
display playback takes a comparatively comfy 9.5W, which suggests a
stable 5h+ of playback, which is ok by me.

Fooling across the internet, small edits, youtube-dl, and I am at round 80%
battery after about an hour, with an estimated 5h left, which is a
little disappointing. I had a 7h remaining estimate earlier than I began
goofing round Discourse, so I believe the web site is a fairly
huge battery drain, really. I see about 10-12 W, whereas I used to be in all probability at
half that (6-8W) simply enjoying music with mpv within the background…

In different phrases, it seems to be like modifying posts in Discourse with Firefox
takes a stable 4-6W of energy. Wonderful and gross.

(When writing about abusive energy utilization generates extra energy utilization, is
that an heisenbug? Or schrödinbug?)

Energy administration

• TODO: evaluation Arch Linux’s tips on energy saving
• TODO: i915 driver has numerous parameters, together with some about
  energy saving, see, once more, the arch wiki, and significantly
  enable_fbc=1

TL:DR; energy administration on the laptop computer is a matter, however there’s numerous
tweaks you can also make to enhance it. Strive:

• powertop --auto-tune
• apt set up tlp && systemctl allow tlp
• nvme.noacpi=1 mem_sleep_default=deep on the kernel command line
  might assist with standby energy utilization
• preserve solely USB-C growth playing cards plugged in, all others suck energy
  even when idle
• take into account upgrading the BIOS to newest beta (3.06 on the time of
  writing), unverified energy financial savings
• newest Linux kernels (6.2) promise energy financial savings as effectively
  (unverified)

Background on CPU structure

There have been energy issues within the eleventh gen Framework laptop computer, in accordance
to this report from Linux After Dark, so the problems with energy
administration on the Framework aren’t new.

The twelfth technology Intel CPU (AKA “Alder Lake“) is a big-little
architecture with “power-saving” and “efficiency” cores. There
was performance problems introduced by the scheduler in Linux
5.16 however these had been ultimately fixed in 5.18, which makes use of
Intel’s {hardware} as an “clever, low-latency hardware-assisted
scheduler”. In accordance to Phoronix, the 5.19 launch improved the
energy saving, at the price of some penalty value. There have been additionally patch
series to make the scheduler configurable, nevertheless it would not look
these have been merged as of 5.19. There was additionally a session about this
on the 2022 Linux Plumbers, however they stopped short of
speaking extra concerning the particular issues Linux is going through in Alder
lake:

Particularly, the kernel’s energy-aware scheduling heuristics do not
work effectively on these CPUs. Plenty of options current there
complicate the power image; these embrace SMT, Intel’s “turbo
increase” mode, and the CPU’s inner power-management mechanisms. For
many workloads, operating on an ostensibly extra power-hungry Pcore can
be extra environment friendly than utilizing an Ecore. Time for dialogue of the
downside was missing, although, and the session got here to an in depth.

All this to say that the 12gen Intel line shipped with this Framework
sequence ought to have higher energy administration due to its
power-saving cores. And Linux has had the scheduler adjustments to make
use of this (however possibly continues to be having hassle). In any case, this
may not be the supply of energy administration issues on my laptop computer,
fairly the alternative.

Additionally word that the firmware updates for various chipsets are
supposed to enhance issues ultimately.

Then again, The Verge merely declared the entire P-series
a mistake…

Makes an attempt at bettering energy utilization

I did attempt to comply with a number of the ideas in this forum post. The
tips powertop --auto-tune and tlp‘s
PCIE_ASPM_ON_BAT=powersupersave principally did nothing: I used to be caught
at 10W energy utilization in powertop (600+mA in tlp-stat).

Apparently, I ought to be capable to attain the C8 CPU energy state (or
even C9, C10) in powertop, however I appear to be inventory at
C7. (Though I am unsure tips on how to learn that tab in powertop: within the
Core(HW) column there’s solely C3/C6/C7 states, and most cores are 85%
in C7 or possibly C6. However the subsequent column over does present many CPUs in
C10 states…

Because it seems, the graphics card really takes up a very good chunk of
energy until correct energy administration is enabled (see under). After
tweaking this, I did handle to get all the way down to round 7W energy utilization in
powertop.

Enlargement playing cards really do take up energy, and so does the display,
clearly. The fully-lit display takes a stable 2-3W of energy in contrast
to the totally dimmed display. When eradicating all growth playing cards and
making the laptop computer idle, I can spin it all the way down to 4 watts energy utilization at
the second, and a tremendous 2 watts when the display turned off.

Caveats

Idle energy utilization checks

Some growth playing cards waste power, even when unused. Here’s a abstract
of the findings from the powerstat web page. I additionally embrace different
gadgets examined on this web page for completeness:

So it seems to be like all growth playing cards however the USB-C ones are lively,
i.e. they draw energy with idle. The USB-A playing cards are the least concern,
sucking out 10mW, just about inside the margin of error. However each the
DisplayPort and HDMI do take a couple of hundred miliwatts. It seems to be like
USB-A connectors have this elementary flaw that they essentially draw
some powers as a result of they lack the facility negotiation options of
USB-C. Not less than according to this post:

It appears the USB A should have energy going to it on a regular basis, that
the previous USB 2 and three protocols, the USB C solely offers energy when
there’s a connection. Outdated versus new.

Apparently, this can be a downside particular to the USB-C to USB-A
adapter that ships with the Framework. Some folks have actually
changed their orders to all USB-C due to this downside, however I am
unsure the issue is as severe as claimed within the boards. I
could not reproduce the “one watt” energy drains prompt elsewhere,
at the very least not repeatedly. (A earlier model of this put up did present
such an influence drain, nevertheless it was in a much less managed check
surroundings than the sequence of extra rigorous checks above.)

The worst offenders are the storage playing cards: the SSD drive takes at
least one watt of energy and the MicroSD card appears to need to take all
the best way as much as 6 watts of energy, each simply sitting there doing
nothing. This confirms claims of 1.4W for the SSD (however not
5W) energy utilization discovered elsewhere. The former post has
directions on tips on how to disable the cardboard in software program. The MicroSD card
has been reported as using 2 watts, however I’ve seen it as excessive as 6
watts, which is fairly damning.

The Framework workforce has a beta update for the DisplayPort adapter
however presently just for Home windows (LVFS technically possible, “under
investigation”). A USB-A firmware replace is additionally under
investigation. It’s subsequently probably at the very least a few of these energy
administration points will ultimately be fastened.

Be aware that the upcoming Ethernet card has a reported 2-8W power usage,
depending on traffic. I did my very own energy utilization checks in
powerstat-wayland and so they appear decrease than 2W.

The upcoming 6.2 Linux kernel may also enhance battery utilization when
idle, see this Phoronix article for details, probably in early
2023.

Idle energy utilization checks underneath Wayland

Replace: I redid these checks underneath Wayland, see powerstat-wayland
for particulars. The TL;DR: is that energy consumption is both smaller or
related.

Idle energy utilization checks, 3.06 beta BIOS

I redid the idle checks after the 3.06 beta BIOS update and ended
up with this outcomes:

Be aware that the desk abstract is totally different than the earlier desk: right here
we present absolutely the numbers whereas the earlier desk was doing a
complicated try at exhibiting relative (to the baseline) numbers.

Conclusion: the three.06 BIOS replace didn’t considerably change idle
energy utilization stats besides for the MicroSD card which has
considerably improved.

The brand new “exterior USB” check can also be fascinating: it reveals how the
offered 1TB SSD card performs (admirably) in comparison with current
gadgets. The opposite new result’s the MicroSD card with a card which,
apparently, makes use of much less energy than the 1TB SSD drive.

Standby battery utilization

I wrote some quick hack to evaluate how much power is used during
sleep. Apparently, that is one
of the areas that should have improved since the first Framework
model, let’s discover out.

My baseline for comparability is the Purism laptop computer, which, in 10 minutes,
went from this:

sep 28 11:19:45 angela systemd-sleep[209379]: /sys/class/power_supply/BAT/charge_now                      =   6045 [mAh]

… to this:

sep 28 11:29:47 angela systemd-sleep[209725]: /sys/class/power_supply/BAT/charge_now                      =   6037 [mAh]

That is 8mAh per 10 minutes (and a couple of seconds), or 48mA, or, with this
battery, about 127 hours or roughly 5 days of standby. Not dangerous!

Compared, right here is my actually previous x220, earlier than:

sep 29 22:13:54 emma systemd-sleep[176315]: /sys/class/power_supply/BAT0/energy_now                     =   5070 [mWh]

… after:

sep 29 22:23:54 emma systemd-sleep[176486]: /sys/class/power_supply/BAT0/energy_now                     =   4980 [mWh]

… which is 90 mwH in 10 minutes, or a whopping 540mA, which was
presumably okay when this battery was new (62000 mAh, so about 100
hours, or about 5 days), however this battery is sort of lifeless and has
solely 5210 mAh when full, so solely 10 hours standby.

See Also

Chemists develop new methodology for water splitting

And right here is the Framework performing an analogous check, earlier than:

sep 29 22:27:04 angela systemd-sleep[4515]: /sys/class/power_supply/BAT1/charge_full                    =   3518 [mAh]
sep 29 22:27:04 angela systemd-sleep[4515]: /sys/class/power_supply/BAT1/charge_now                     =   2861 [mAh]

… after:

sep 29 22:37:08 angela systemd-sleep[4743]: /sys/class/power_supply/BAT1/charge_now                     =   2812 [mAh]

… which is 49mAh in just a little over 10 minutes (and 4 seconds), or
292mA, rather more than the Purism, however half of the X220. At this fee,
the battery would final on standby solely 12 hours!! That’s fairly
dangerous.

Be aware that this was accomplished with the next growth playing cards:

• 2 USB-C
• 1 1TB SSD drive
• 1 USB-A with a hub linked to it, with keyboard and LAN

Preliminary checks with out the hub (over one minute) present that it
would not considerably have an effect on this energy consumption (300mA).

This guide additionally suggests booting with nvme.noacpi=1 however this
nonetheless offers me about 5mAh/min (or 300mA).

Including mem_sleep_default=deep to the kernel command line does make a
distinction. Earlier than:

sep 29 23:03:11 angela systemd-sleep[3699]: /sys/class/power_supply/BAT1/charge_now                     =   2544 [mAh]

… after:

sep 29 23:04:25 angela systemd-sleep[4039]: /sys/class/power_supply/BAT1/charge_now                     =   2542 [mAh]

… which is 2mAh in 74 seconds, which is 97mA, brings us to a extra
cheap 36 hours, or a day and a half. It is nonetheless above the x220
energy utilization, and greater than an order of magnitude greater than the Purism
laptop computer. It is also removed from the 0.4% promised by upstream, which
could be 14mA for the 3500mAh battery.

It also needs to be famous that this “deep” sleep mode is a bit more
disruptive than common sleep. As you possibly can see by the timing, it took
greater than 10 seconds for the laptop computer to renew, which feels just a little
alarming as your banging the keyboard to carry it again to life.

You’ll be able to verify the present sleep mode with:

# cat /sys/energy/mem_sleep
s2idle [deep]

Within the above, deep is chosen. You’ll be able to change it on the fly with:

printf s2idle > /sys/energy/mem_sleep

This is one other check:

sep 30 22:25:50 angela systemd-sleep[32207]: /sys/class/power_supply/BAT1/charge_now                     =   1619 [mAh]
sep 30 22:31:30 angela systemd-sleep[32516]: /sys/class/power_supply/BAT1/charge_now                     =   1613 [mAh]

… higher! 6 mAh in about 6 minutes, works out to 63.5mA, so extra
than two days standby.

An extended check:

oct 01 09:22:56 angela systemd-sleep[62978]: /sys/class/power_supply/BAT1/charge_now                     =   3327 [mAh]
oct 01 12:47:35 angela systemd-sleep[63219]: /sys/class/power_supply/BAT1/charge_now                     =   3147 [mAh]

That is 180mAh in about 3.5h, 52mA! Now at 66h, or virtually 3 days.

I wasn’t certain why I used to be seeing such fluctuations in these checks, however
because it seems, expansion card power tests present that they do
considerably have an effect on energy utilization, particularly the SSD drive, which might
take as much as two full watts of energy even when idle. I did not management
for growth playing cards within the above checks — operating them with no matter
card I had plugged in with out paying consideration — so it is probably the
explanation for the excessive energy utilization and fluctuations.

It could be attainable to work round this downside by disabling USB
devices earlier than droop. TODO. See additionally this post.

Within the meantime, I’ve been in a position to get a lot higher droop
efficiency by unplugging all modules. Then I get this consequence:

oct 04 11:15:38 angela systemd-sleep[257571]: /sys/class/power_supply/BAT1/charge_now                     =   3203 [mAh]
oct 04 15:09:32 angela systemd-sleep[257866]: /sys/class/power_supply/BAT1/charge_now                     =   3145 [mAh]

Which is 14.8mA! Virtually precisely the quantity promised by Framework! With
a full battery, meaning a ten days droop time. That is really
fairly good, and much past what I used to be anticipating when beginning down
this journey.

So, as soon as the growth playing cards are unplugged, droop energy utilization is
really fairly cheap. Extra detailed standby checks can be found
within the standby-tests web page, with a abstract under.

There’s additionally some hope that the Chromebook edition —
particularly designed with a specification of 14 days standby
time — may carry some firmware enhancements again all the way down to the
regular line. A few of these points had been reported upstream in April
2022, however there would not appear to have been any progress there
since.

TODO: one remaining resolution right here is suspend-then-hibernate, which
Home windows makes use of for this

TODO: take into account implementing the S0ix sleep states , see additionally troubleshooting

TODO: take into account https://github.com/intel/pm-graph

Standby growth playing cards check outcomes

This desk is a abstract of the extra intensive standby-tests I’ve carried out:

System	Wattage	Amperage	Days	Be aware
baseline	0.25W	16mA	9	sleep=deep nvme.noacpi=1
s2idle	0.29W	18.9mA	~7	sleep=s2idle nvme.noacpi=1
regular nvme	0.31W	20mA	~7	sleep=s2idle with out nvme.noacpi=1
1 USB-C	0.23W	15mA	~10
2 USB-C	0.23W	14.9mA		similar as above
1 USB-A	0.75W	48.7mA	3	+500mW (!!) for the primary USB-A card!
2 USB-A	1.11W	72mA	2	+360mW
3 USB-A	1.48W	96mA	<2	+370mW
1TB SSD	0.49W	32mA	<5	+260mW
MicroSD	0.52W	34mA	~4	+290mW
DisplayPort	0.85W	55mA	<3	+620mW (!!)
1 HDMI	0.58W	38mA	~4	+250mW
2 HDMI	0.65W	42mA	<4	+70mW

Conclusions:

• USB-C playing cards take no additional energy on droop, presumably much less
  than empty slots, extra testing required

• USB-A playing cards take a lot extra energy on droop
  (300-500mW) than on common idle (~10mW, virtually negligible)

• 1TB SSD and MicroSD playing cards appear to take a cheap
  quantity of energy (260-290mW), in comparison with their runtime
  equivalents (1-6W!)

• DisplayPort takes a shocking lot of energy (620mW), virtually
  double its common runtime utilization (390mW)

• HDMI playing cards take, surprisingly, much less energy (250mW) in
  standby than the DP card (620mW)

• and oddly, a second card provides much less energy utilization (70mW?!) than the
  first, possibly a circuit is utilized by each?

A dialogue of these outcomes is in this forum post.

Standby growth playing cards check outcomes, 3.06 beta BIOS

Framework just lately (2022-11-07) announced that they may publish
a firmware improve to deal with a number of the USB-C points, together with
energy administration. This might positively have an effect on the above consequence,
bettering each standby and runtime energy utilization.

The replace came out in December 2022 and I redid my evaluation with
the next outcomes:

Full ends in standby-tests-306. The massive takeaway for me is that
the replace did not enhance energy utilization on the USB-A ports which is a
huge downside for my use case. There’s a notable enchancment on the
DisplayPort energy consumption which brings it extra in step with the
HDMI connector, nevertheless it nonetheless would not correctly flip off on droop
both.

Even worse, the USB-A ports now typically fails to resume after
suspend, which is fairly annoying. It is a known problem
that may hopefully get fastened within the remaining launch.

Battery put on safety

The BIOS has an choice to restrict cost to 80% to mitigate battery
wear. There is a strategy to management the embedded controller from
runtime with fw-ectool, partly documented here. The command
would be:

sudo ectool fwchargelimit 80

I checked out constructing this myself however failed to run it. I opened a
RFP in Debian in order that we will ship this in Debian, and also documented
my work there.

Be aware that there’s now a counter that tracks cost/discharge
cycles. It is seen in tlp-stat -b, which is a pleasant
enchancment:

root@angela:/residence/anarcat# tlp-stat -b
--- TLP 1.5.0 --------------------------------------------

+++ Battery Care
Plugin: generic
Supported options: none out there

+++ Battery Standing: BAT1
/sys/class/power_supply/BAT1/producer                   = NVT
/sys/class/power_supply/BAT1/model_name                     = Framewo
/sys/class/power_supply/BAT1/cycle_count                    =      3
/sys/class/power_supply/BAT1/charge_full_design             =   3572 [mAh]
/sys/class/power_supply/BAT1/charge_full                    =   3541 [mAh]
/sys/class/power_supply/BAT1/charge_now                     =   1625 [mAh]
/sys/class/power_supply/BAT1/current_now                    =    178 [mA]
/sys/class/power_supply/BAT1/standing                         = Discharging

/sys/class/power_supply/BAT1/charge_control_start_threshold = (not out there)
/sys/class/power_supply/BAT1/charge_control_end_threshold   = (not out there)

Cost                                                      =   45.9 [%]
Capability                                                    =   99.1 [%]

One factor that’s nonetheless lacking is the charge threshold data (the
(not out there) above). There’s been some work to make that
accessible in August, keep tuned? This could additionally make it attainable
implement hysteresis support.

Ethernet growth card

The Framework ethernet expansion card is a flowery little doodle:
“2.5Gbit/s and 10/100/1000Mbit/s Ethernet”, the “clear housing lets
you peek on the RTL8156 controller that powers it”. Which is one other
strategy to say “we did not utterly end prod on this one, so it form
of seems to be like we 3D-printed this within the store”….

The cardboard is just a little cumbersome, however I suppose that is inevitable contemplating
the RJ-45 kind issue when in comparison with the skinny Framework laptop computer.

I’ve had a serious issue when making an attempt it at first: the hyperlink LEDs
simply would not come up. I made a full bug report in the forum and
with upstream help, however ultimately figured it out alone. It is
(in fact) an influence saving subject: if you happen to reboot the machine, the hyperlinks
come up when the laptop computer is operating the BIOS POST examine and even when
the Linux kernel boots.

I first thought that the issue is probably going associated to the
powertop service which I run at boot time to tweak some energy saving
settings.

It looks like this:

echo 'on' > '/sys/bus/usb/gadgets/4-2/energy/management'

… is an effective workaround to carry the cardboard again on-line. You’ll be able to even
return to energy saving mode and the cardboard will nonetheless work:

echo 'auto' > '/sys/bus/usb/gadgets/4-2/energy/management'

Further research by Matt_Hartley from the Framework
Team discovered this issue in the tlp tracker that reveals how the
USB_AUTOSUSPEND setting allows the facility saving even when the
driver would not help it, which, looking back, simply appears like a
dangerous thought. To cite that subject:

By default, USB energy saving is lively within the kernel, however not
force-enabled for incompatible drivers. That’s, gadgets that
help suspension will droop, drivers that don’t, is not going to.

So the repair is definitely to uninstall tlp or disable that setting by
including this to /and many others/tlp.conf:

USB_AUTOSUSPEND=0

… however that disables auto-suspend on all USB gadgets, which can
damage different energy utilization efficiency. I have found {that a} a
mixture of:

USB_AUTOSUSPEND=1
USB_DENYLIST="0bda:8156"

and this on the kernel commandline:

usbcore.quirks=0bda:8156:ok

… really does work accurately. I now have this in my
/and many others/default/grub.d/framework-tweaks.cfg file:

# internet.ifnames=0: regular interface names ffs (e.g. eth0, wlan0, not wlp166
s0)
# nvme.noacpi=1: scale back SSD disk energy utilization (not working)
# mem_sleep_default=deep: scale back energy utilization throughout sleep (not working)
# usbcore.quirk is a workaround for the ethernet card droop bug: https:
//guides.body.work/Information/Fedora+37+Set up+on+the+Framework+Laptop computer/
108?lang=en
GRUB_CMDLINE_LINUX="internet.ifnames=0 nvme.noacpi=1 mem_sleep_default=deep usbcore.quirks=0bda:8156:ok"

# repair the decision in grub for fonts to not be tiny
GRUB_GFXMODE=1024x768

Apart from that, I have not been in a position to max out the cardboard as a result of I
do not produce other 2.5Gbit/s tools at residence, which is unusually
satisfying. However operating in opposition to my Turris Omnia
router, I may just about max a gigabit pretty simply:

[ ID] Interval           Switch     Bitrate         Retr
[  5]   0.00-10.00  sec  1.09 GBytes   937 Mbits/sec  238             sender
[  5]   0.00-10.00  sec  1.09 GBytes   934 Mbits/sec                  receiver

The cardboard would not require any proprietary firmware blobs which is
shocking. Apart from the facility saving points, it simply works.

In my energy checks (see powerstat-wayland), the Ethernet card appears
to make use of about 1.6W of energy idle, with out hyperlink, within the above “quirky”
configuration the place the cardboard is practical however with out autosuspend.

Proprietary firmware blobs

• the WiFi community card shipped with the DIY equipment is a AX210 card that
  requires a 5.19 kernel or later, and the firmware-iwlwifi non-free firmware bundle
• the Bluetooth adapter additionally hundreds the firmware-iwlwifi
  bundle (untested)
• the graphics work out of the field with out firmware, however sure
  energy administration options come solely with particular proprietary
  firmware, usually shipped within the firmware-misc-nonfree
  however presently lacking from the bundle

cd /lib/firmware && rm adlp_guc_70.1.1.bin adlp_guc_69.0.3.bin
update-initramfs -u

W: Potential lacking firmware /lib/firmware/i915/adlp_guc_69.0.3.bin for module i915
W: Potential lacking firmware /lib/firmware/i915/adlp_guc_70.1.1.bin for module i915

[Unit]
Description=begin compositing supervisor
PartOf=graphical-session.goal
ConditionHost=angela

[Service]
Sort=exec
ExecStart=compton --show-all-xerrors --backend glx --vsync opengl-swc
Restart=on-failure

[Install]
RequiredBy=graphical-session.goal

anarcat@angela:~$ compiz --replace
compiz (core) - Warn: No XI2 extension
compiz (core) - Error: One other composite supervisor is already operating on display: 0
compiz (core) - Deadly: No manageable screens discovered on show :0

compiz (core) - Warn: No XI2 extension
compiz (core) - Error: Could not load plugin 'ccp'
compiz (core) - Error: Could not load plugin 'ccp'

[   19.534429] Intel(R) Wi-fi WiFi driver for Linux
[   19.534691] iwlwifi 0000:a6:00.0: enabling system (0000 -> 0002)
[   19.541867] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-72.ucode (-2)
[   19.541881] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-72.ucode (-2)
[   19.541882] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-72.ucode failed with error -2
[   19.541890] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-71.ucode (-2)
[   19.541895] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-71.ucode (-2)
[   19.541896] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-71.ucode failed with error -2
[   19.541903] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-70.ucode (-2)
[   19.541907] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-70.ucode (-2)
[   19.541908] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-70.ucode failed with error -2
[   19.541913] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-69.ucode (-2)
[   19.541916] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-69.ucode (-2)
[   19.541917] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-69.ucode failed with error -2
[   19.541922] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-68.ucode (-2)
[   19.541926] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-68.ucode (-2)
[   19.541927] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-68.ucode failed with error -2
[   19.541933] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-67.ucode (-2)
[   19.541937] iwlwifi 0000:a6:00.0: firmware: did not load iwlwifi-ty-a0-gf-a0-67.ucode (-2)
[   19.541937] iwlwifi 0000:a6:00.0: Direct firmware load for iwlwifi-ty-a0-gf-a0-67.ucode failed with error -2
[   19.544244] iwlwifi 0000:a6:00.0: firmware: direct-loading firmware iwlwifi-ty-a0-gf-a0-66.ucode
[   19.544257] iwlwifi 0000:a6:00.0: api flags index 2 bigger than supported by driver
[   19.544270] iwlwifi 0000:a6:00.0: TLV_FW_FSEQ_VERSION: FSEQ Model: 0.63.2.1
[   19.544523] iwlwifi 0000:a6:00.0: firmware: did not load iwl-debug-yoyo.bin (-2)
[   19.544528] iwlwifi 0000:a6:00.0: firmware: did not load iwl-debug-yoyo.bin (-2)
[   19.544530] iwlwifi 0000:a6:00.0: loaded firmware model 66.55c64978.0 ty-a0-gf-a0-66.ucode op_mode iwlmvm

Graphics acceleration

Working synchronized to the vertical refresh.  The framerate ought to be
roughly the identical because the monitor refresh fee.
137 frames in 5.1 seconds = 26.984 FPS
27 frames in 5.4 seconds =  5.022 FPS

300 frames in 5.0 seconds = 60.000 FPS

I put in intel-gpu-tools for the intel_gpu_top
command to verify the GPU was engaged when doing these simulations. A
good discover. Different helpful diagnostic instruments embrace glxgears and
glxinfo (in mesa-utils) and (vainfo in vainfo).

Following to this post, I additionally made certain to have these settings
in my about:config in Firefox, or, in person.js:

user_pref("media.ffmpeg.vaapi.enabled", true);

Be aware that the information suggests many different settings to tweak, however these
may really be overkill, see this comment and its parents. I
did strive forcing {hardware} acceleration by setting gfx.webrender.all
to true, however every little thing grew to become uneven and peculiar.

The information additionally mentions putting in the intel-media-driver bundle,
however I couldn’t discover that in Debian.

The Arch wiki has, as normal, a superb reference on hardware
acceleration in Firefox.

Chromium / Sign desktop bugs

It seems to be like each Chromium and Sign Desktop misbehave with my
compositor setup (compton + i3). The fix is so as to add a persistent
flag to Chromium. In Arch, it is conveniently in
~/.config/chromium-flags.conf however that does not really work in
Debian. I needed to put the flag in
/and many others/chromium.d/disable-compositing, like this:

export CHROMIUM_FLAGS="$CHROMIUM_FLAGS --disable-gpu-compositing"

It is attainable one other one of many hundreds of flags may repair this
subject higher, however I do not actually have time to undergo this complete,
incomplete, and unofficial checklist (!?!).

Sign Desktop is an analogous downside, and would not reuse these flags
(due to course it would not). As a substitute I needed to rewrite the wrapper
script in /usr/native/bin/signal-desktop to make use of this as an alternative:

exec /usr/bin/flatpak run --branch=secure --arch=x86_64 org.sign.Sign --disable-gpu-compositing "$@"

This was largely accomplished in this Puppet commit.

I have not found out the basis of this downside. I did strive utilizing
picom and xcompmgr; they each endure from the identical subject. One other
Debian testing person on Wayland advised me they have not seen this downside,
so hopefully this may be fastened by switching to
wayland.

Graphics card hangs

I imagine I may need this bug which ends up in a complete
graphical cling for 15-30 seconds. It is pretty uncommon so it isn’t too
disruptive, however when it does occur, it is fairly alarming.

The feedback on that bug report are encouraging although: it appears this
is a bug in both mesa or the Intel graphics driver, which suggests many
folks have this downside so it is prone to be fastened. There’s really
a merge request on mesa already (2022-12-29).

It may even be that bug as a result of the error message I get is
really:

Jan 20 12:49:10 angela kernel: Asynchronous wait on fence 0000:00:02.0:sway[104431]:cb0ae timed out (trace:intel_atomic_commit_ready [i915]) 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 12:0:00000000 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] Resetting chip for stopped heartbeat on rcs0 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] GuC firmware i915/adlp_guc_70.1.1.bin model 70.1 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] HuC firmware i915/tgl_huc_7.9.3.bin model 7.9 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] HuC authenticated 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] GuC submission enabled 
Jan 20 12:49:15 angela kernel: i915 0000:00:02.0: [drm] GuC SLPC enabled

It is a stable 30 seconds graphical cling. Perhaps the keyboard and
every little thing else retains working. The latter bug report is kind of lengthy,
with many feedback, however this one from January 2023 appears to say
that Sway 1.8 fastened the issue. There’s additionally an earlier patch to
add an additional kernel parameter that supposedly fixes that too. There’s
all types of different workarounds in there, for instance this:

echo "choices i915 enable_dc=1 enable_guc_loading=1 enable_guc_submission=1 edp_vswing=0 enable_guc=2 enable_fbc=1 enable_psr=1 disable_power_well=0" | sudo tee /and many others/modprobe.d/i915.conf

from this comment… In order that one is unsolved, so far as the
upstream drivers are involved, however possibly could possibly be fastened via Sway.

Bizarre USB hangs / graphical glitches

I’ve had bizarre connectivity glitches higher described in this
post, however principally: my USB keyboard and mice (linked over a
USB hub) drop keys, lag rather a lot or cling, and I get visible glitches.

The repair was to tighten the screws across the CPU on the motherboard
(!), which is, fortunately, a fairly easy restore.

I ordered the Framework in August 2022 and acquired it a couple of month
later, which is ahead of anticipated as a result of the August batch was
late.

Individuals (together with me) anticipated this to have an effect on the September
batch, nevertheless it appears Framework have been in a position to repair the supply
issues and sustain with the demand.

As of early 2023, their web site publicizes that laptops ship “inside 5
days”. I’ve myself ordered a couple of growth playing cards in November 2022,
and so they shipped on the identical day, arriving 3-4 days later.

The availability pipeline

This comes from the crowdsourced spreadsheet, which ought to be
up to date when the standing adjustments right here.

I used to be a part of the “third batch” of the twelfth technology laptop computer, which
was imagined to ship in September. It ended up arriving on my door
step on September twenty seventh, about 33 days after ordering.

It appears present orders aren’t processed in “batches”, however in actual
time, see this blog post for details on shipping.

Delivery trivia

• Taoyuan, Taiwan 2022-09-23
• Anchorage, Alaska, USA (?!) 2022-09-24
• Memphis, Tennessee, USA 2022-09-25
• Winnipeg, Manitoba, Canada (!! again the mistaken approach!) 2022-09-26
• Mirabel, Québec, Canada 2022-09-27
• Montréal, Québec, Canada

I can not fairly determine tips on how to calculate precisely how a lot mileage
that’s, nevertheless it’s enormous. The trip via Alaska is shocking sufficient
however the bounce again via Winnipeg is particularly bizarre. I suppose
the route occurs that approach due to Fedex delivery hubs.

There was a associated oddity once I had my Purism laptop computer shipped: it
left from the west coast and appeared to enter on an countless, two week
lengthy highway journey throughout the continental US.