<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Ismael Alex Otero</title>
    <description>The latest articles on DEV Community by Ismael Alex Otero (@signalnotnoise).</description>
    <link>https://dev.to/signalnotnoise</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4250562%2F5aab5e39-248e-4fb7-985f-096af15da83d.jpg</url>
      <title>DEV Community: Ismael Alex Otero</title>
      <link>https://dev.to/signalnotnoise</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/signalnotnoise"/>
    <language>en</language>
    <item>
      <title>Two faults, one stutter: MacBookPro16,1 on Omarchy</title>
      <dc:creator>Ismael Alex Otero</dc:creator>
      <pubDate>Tue, 17 Nov 2026 20:42:58 +0000</pubDate>
      <link>https://dev.to/signalnotnoise/two-faults-one-stutter-macbookpro161-on-omarchy-40fo</link>
      <guid>https://dev.to/signalnotnoise/two-faults-one-stutter-macbookpro161-on-omarchy-40fo</guid>
      <description>&lt;p&gt;Field note from one machine, 17 November 2026. The desktop felt like it was&lt;br&gt;
dropping frames. Two separate faults were doing it at once. The CPU was pinned&lt;br&gt;
at 800 MHz by a stuck external PROCHOT pin, and Hyprland was drawing every&lt;br&gt;
frame on the Intel GPU and copying it onto the Radeon, which is the chip the&lt;br&gt;
panels are actually wired to.&lt;/p&gt;

&lt;p&gt;This is one afternoon on one laptop. It is not a general recommendation to&lt;br&gt;
disable PROCHOT, and it is not a recommendation to keep the Radeon awake on&lt;br&gt;
every T2 Mac.&lt;/p&gt;
&lt;h2&gt;
  
  
  Machine
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Model&lt;/td&gt;
&lt;td&gt;Apple MacBookPro16,1 (16-inch, 2019, T2)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CPU&lt;/td&gt;
&lt;td&gt;Intel Core i7-9750H, 6 cores / 12 threads, 800 MHz–4.5 GHz&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;iGPU&lt;/td&gt;
&lt;td&gt;UHD Graphics 630, PCI &lt;code&gt;00:02.0&lt;/code&gt;, &lt;code&gt;8086:3e9b&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;dGPU&lt;/td&gt;
&lt;td&gt;Navi 14, PCI &lt;code&gt;03:00.0&lt;/code&gt;, &lt;code&gt;1002:7340&lt;/code&gt; (the 5300M in this model)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Internal panel&lt;/td&gt;
&lt;td&gt;Apple Color LCD, 3072×1920 @ 60 Hz, on the Radeon (&lt;code&gt;eDP-1&lt;/code&gt;)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;External panel&lt;/td&gt;
&lt;td&gt;Acer XV271, 1920×1080 @ 240 Hz, on the Radeon (&lt;code&gt;DP-5&lt;/code&gt;)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;OS&lt;/td&gt;
&lt;td&gt;Omarchy, kernel &lt;code&gt;7.2.15-arch1-1-t2&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Desktop&lt;/td&gt;
&lt;td&gt;Hyprland 0.56.2, started by uwsm, SDDM autologin&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Both connectors that were lit, the internal panel and the Acer, belonged to&lt;br&gt;
the AMD card. The Intel card had no connected display.&lt;/p&gt;
&lt;h2&gt;
  
  
  Fault 1: the CPU would not leave 800 MHz
&lt;/h2&gt;

&lt;p&gt;Every thread sat at 800 MHz. A 12-thread busy loop did not move them. The&lt;br&gt;
package was 66°C. The performance power profile was selected, turbo was&lt;br&gt;
allowed, the frequency cap was 4.5 GHz, and the AC adapter was online. Chassis&lt;br&gt;
fans were already around 4400 RPM, so it looked like a thermal problem.&lt;/p&gt;

&lt;p&gt;It was not the CPU's own throttle. &lt;code&gt;MSR_POWER_CTL&lt;/code&gt; (&lt;code&gt;0x1FC&lt;/code&gt;) read &lt;code&gt;0x6c005d&lt;/code&gt;.&lt;br&gt;
Bit 0, the bidirectional PROCHOT enable, was set. On all 12 threads,&lt;br&gt;
&lt;code&gt;IA32_THERM_STATUS&lt;/code&gt; (&lt;code&gt;0x19C&lt;/code&gt;) and &lt;code&gt;IA32_PACKAGE_THERM_STATUS&lt;/code&gt; (&lt;code&gt;0x1B1&lt;/code&gt;) had&lt;br&gt;
bit 2 set (PROCHOT is asserted) and bits 0 and 4 clear (the CPU is not at its&lt;br&gt;
own thermal or critical limit). Package power limits were 100 W and 125 W.&lt;br&gt;
Clock modulation was off. Hardware P-state request already allowed the top&lt;br&gt;
ratio.&lt;/p&gt;

&lt;p&gt;Clearing bit 0 of &lt;code&gt;0x1FC&lt;/code&gt; on every logical CPU, and leaving every other bit&lt;br&gt;
alone, let the same busy loop reach about 3.5 GHz within a second. After that&lt;br&gt;
the cores idled and boosted in the 3.6–4.3 GHz range at 74–79°C. One package&lt;br&gt;
sample read 100°C during the jump and then fell back with the rest. The CPU's&lt;br&gt;
own thermal-status bit stayed clear.&lt;/p&gt;

&lt;p&gt;Firmware puts bit 0 back on boot and again on resume, so the clear has to be&lt;br&gt;
reapplied. The hook used here refuses to write if bit 0 or bit 4 of either&lt;br&gt;
thermal MSR is set. That leaves the CPU's own overheat brake in place. It does&lt;br&gt;
not leave the external pin in place: a real PROCHOT from the SMC, a VRM, or&lt;br&gt;
the GPU will also be ignored until the next boot if those thermal bits are&lt;br&gt;
clear. That is the tradeoff, and it is why this should stay a per-machine&lt;br&gt;
workaround rather than a default.&lt;/p&gt;

&lt;p&gt;The &lt;a href="https://wiki.t2linux.org/guides/hybrid-graphics/" rel="noopener noreferrer"&gt;T2 Linux hybrid-graphics guide&lt;/a&gt;&lt;br&gt;
describes freezes, a fan ramp, and sudden shutdown (CPU CATERR) on this&lt;br&gt;
generation when the AMD GPU is unhappy. A stuck PROCHOT pin is one of the ways&lt;br&gt;
a hot or confused SMC gets the CPU's attention. Clearing it removes that&lt;br&gt;
signal. Watch the temperatures.&lt;/p&gt;
&lt;h2&gt;
  
  
  Fault 2: Hyprland drew on the Intel GPU
&lt;/h2&gt;

&lt;p&gt;At startup Aquamarine enumerated one GPU, &lt;code&gt;/dev/dri/card1&lt;/code&gt;, driver &lt;code&gt;i915&lt;/code&gt;, and&lt;br&gt;
made it primary. A moment later udev added &lt;code&gt;/dev/dri/card2&lt;/code&gt;, driver &lt;code&gt;amdgpu&lt;/code&gt;.&lt;br&gt;
The log then said the Radeon backend was starting with primary &lt;code&gt;/dev/dri/card1&lt;/code&gt;.&lt;br&gt;
The late card never became the renderer. It only received the finished frames,&lt;br&gt;
because that is where the connectors are.&lt;/p&gt;

&lt;p&gt;Over one second, Hyprland's Intel render engine accumulated more time than the&lt;br&gt;
wall clock, while the Radeon's graphics engine was around 11% and its shader&lt;br&gt;
clock sat on the 200 MHz step. The desktop was being painted on the small GPU&lt;br&gt;
and copied across the PCI bus onto the chip that scans out, including onto a&lt;br&gt;
240 Hz panel. The Radeon looked idle because it was.&lt;/p&gt;

&lt;p&gt;Hyprland chooses that primary from the DRM devices that exist when it starts.&lt;br&gt;
Listing the Radeon first in &lt;code&gt;AQ_DRM_DEVICES&lt;/code&gt; does nothing if the node is not&lt;br&gt;
there yet. The fallback becomes primary, and the hotplug stays a secondary&lt;br&gt;
display device.&lt;/p&gt;
&lt;h2&gt;
  
  
  Telling them apart
&lt;/h2&gt;

&lt;p&gt;These two show up together and feel like one stutter.&lt;/p&gt;

&lt;p&gt;The CPU fault is a clock that will not rise. Read &lt;code&gt;/proc/cpuinfo&lt;/code&gt; during a&lt;br&gt;
short all-core burn. If every thread is still at the minimum (800 MHz on this&lt;br&gt;
chip) and the package is far from 100°C, read &lt;code&gt;0x19C&lt;/code&gt; and &lt;code&gt;0x1B1&lt;/code&gt;. Bit 2 set,&lt;br&gt;
with bits 0 and 4 clear, is an external PROCHOT pin. Bit 0 or bit 4 set means&lt;br&gt;
the CPU itself is hot, and the pin should be left alone.&lt;/p&gt;

&lt;p&gt;The GPU fault is in the Hyprland log. "Found 1 GPUs" followed by an Intel&lt;br&gt;
backend, then an amdgpu backend "with primary" pointing at the Intel card,&lt;br&gt;
while &lt;code&gt;hyprctl monitors&lt;/code&gt; shows the live panels on the AMD connectors. The&lt;br&gt;
compositor is copying.&lt;/p&gt;
&lt;h2&gt;
  
  
  What was changed
&lt;/h2&gt;

&lt;p&gt;PROCHOT, reapplied at boot and after wake:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;/usr/local/sbin/disable-bd-prochot&lt;/code&gt; reads the two thermal MSRs and, when
bits 0 and 4 are clear, clears only bit 0 of &lt;code&gt;0x1FC&lt;/code&gt; on each logical CPU.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;disable-bd-prochot.service&lt;/code&gt; runs that at &lt;code&gt;multi-user.target&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;/usr/lib/systemd/system-sleep/disable-bd-prochot&lt;/code&gt; runs it on the &lt;code&gt;post&lt;/code&gt;
sleep action, because resume restores the firmware value.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;/etc/modules-load.d/msr.conf&lt;/code&gt; loads the &lt;code&gt;msr&lt;/code&gt; module so &lt;code&gt;/dev/cpu/*/msr&lt;/code&gt;
exists early enough.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;GPU order, applied on the next Hyprland start:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight conf"&gt;&lt;code&gt;&lt;span class="c1"&gt;# /etc/udev/rules.d/69-drm-gpu-names.rules
&lt;/span&gt;&lt;span class="na"&gt;KERNEL&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"card*"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;KERNELS&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"0000:03:00.0"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;SUBSYSTEM&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"drm"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;DRIVERS&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"amdgpu"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;SYMLINK&lt;/span&gt;&lt;span class="o"&gt;+=&lt;/span&gt;&lt;span class="s2"&gt;"dri/radeon"&lt;/span&gt;
&lt;span class="na"&gt;KERNEL&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"card*"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;KERNELS&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"0000:00:02.0"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;SUBSYSTEM&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"drm"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;DRIVERS&lt;/span&gt;&lt;span class="o"&gt;==&lt;/span&gt;&lt;span class="s2"&gt;"i915"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="no"&gt;SYMLINK&lt;/span&gt;&lt;span class="o"&gt;+=&lt;/span&gt;&lt;span class="s2"&gt;"dri/intel"&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;/dev/dri/cardN&lt;/code&gt; numbers move between boots. &lt;code&gt;/dev/dri/by-path/pci-0000:03:00.0-card&lt;/code&gt;&lt;br&gt;
is stable and still unusable here: Aquamarine splits &lt;code&gt;AQ_DRM_DEVICES&lt;/code&gt; on every&lt;br&gt;
colon, and the PCI path contains two. The symlinks avoid both problems. On&lt;br&gt;
this Hyprland 0.56.2 the symlink was resolved; the log said the explicit device&lt;br&gt;
&lt;code&gt;/dev/dri/card2&lt;/code&gt; was found. An older report&lt;br&gt;
(&lt;a href="https://github.com/hyprwm/Hyprland/issues/6997" rel="noopener noreferrer"&gt;hyprwm/Hyprland#6997&lt;/a&gt;)&lt;br&gt;
described symlinks being ignored. Check the log for "Explicit device … found"&lt;br&gt;
after changing this.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;AQ_DRM_DEVICES&lt;/code&gt; has to be in the environment before Hyprland starts.&lt;br&gt;
&lt;code&gt;hl.env()&lt;/code&gt; in the Lua config runs too late. Under uwsm the file is&lt;br&gt;
&lt;code&gt;~/.config/uwsm/env-hyprland&lt;/code&gt;, because &lt;code&gt;XDG_CURRENT_DESKTOP&lt;/code&gt; is &lt;code&gt;Hyprland&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="kd"&gt;export&lt;/span&gt; &lt;span class="nv"&gt;AQ_DRM_DEVICES&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"/dev/dri/radeon:/dev/dri/intel"&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The first entry is the renderer. The Intel node stays in the list so a missing&lt;br&gt;
Radeon can still bring the session up.&lt;/p&gt;

&lt;p&gt;A user drop-in waits for the name to appear, up to five seconds, then starts&lt;br&gt;
Hyprland anyway:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight systemd"&gt;&lt;code&gt;&lt;span class="c1"&gt;# ~/.config/systemd/user/wayland-wm@hyprland.desktop.service.d/wait-radeon.conf
&lt;/span&gt;&lt;span class="k"&gt;[Service]&lt;/span&gt;
&lt;span class="na"&gt;ExecStartPre&lt;/span&gt;&lt;span class="o"&gt;=%&lt;/span&gt;&lt;span class="n"&gt;h&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;config&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="n"&gt;uwsm&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="n"&gt;wait&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="k"&gt;for&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="n"&gt;radeon&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;wait-for-radeon&lt;/code&gt; polls &lt;code&gt;/dev/dri/radeon&lt;/code&gt; every tenth of a second and exits 0&lt;br&gt;
whether or not it showed up. &lt;code&gt;systemctl --user daemon-reload&lt;/code&gt; picks the drop-in&lt;br&gt;
up. The value is not applied to a compositor that is already running. This&lt;br&gt;
session was restarted with &lt;code&gt;uwsm stop&lt;/code&gt;; SDDM's autologin brought it back.&lt;/p&gt;
&lt;h2&gt;
  
  
  Result
&lt;/h2&gt;

&lt;p&gt;The new log enumerated both cards from the explicit list, started the amdgpu&lt;br&gt;
backend first, and recorded "gpu /dev/dri/card2 becomes primary". The renderer&lt;br&gt;
line was:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;AMD Radeon Graphics (radeonsi, navi14, ACO, DRM 3.64, 7.2.15-arch1-1-t2)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The Intel backend then started with the Radeon as primary. The same two panels&lt;br&gt;
came back at the same modes. With the desktop idle, the Radeon sat near 7% on&lt;br&gt;
its 200 MHz shader step, which is a quiet compositor rather than a copy engine.&lt;br&gt;
The cores were boosting in the upper 3 GHz range. Subjectively the machine&lt;br&gt;
felt much faster. That is the CPU no longer being stuck at its floor, and the&lt;br&gt;
240 Hz panel no longer being fed through a copy from the other GPU.&lt;/p&gt;

&lt;h2&gt;
  
  
  What was not retested
&lt;/h2&gt;

&lt;p&gt;The PROCHOT service was started once, after the bit had already been cleared,&lt;br&gt;
and logged that BD PROCHOT was already disabled. A reboot, and a&lt;br&gt;
suspend/resume cycle, were not done after the hook was installed. The sleep&lt;br&gt;
hook is the standard &lt;code&gt;post&lt;/code&gt; path, and the register is documented to come back&lt;br&gt;
on those events, but this note does not claim those paths were observed.&lt;/p&gt;

&lt;p&gt;No heavy GPU workload was measured after the switch. The claim is that the&lt;br&gt;
Radeon became the renderer, which the log shows.&lt;/p&gt;

&lt;h2&gt;
  
  
  Prior art, and the opposite advice
&lt;/h2&gt;

&lt;p&gt;Stuck BD PROCHOT at the minimum ratio is an old laptop failure. ThrottleStop&lt;br&gt;
and several Linux MSR scripts clear the same bit. Nothing about that mechanism&lt;br&gt;
is new.&lt;/p&gt;

&lt;p&gt;The T2 Linux guide, and at least one Omarchy write-up for this exact model&lt;br&gt;
(&lt;a href="https://github.com/novuon/mbp2019-omarchy" rel="noopener noreferrer"&gt;mbp2019-omarchy&lt;/a&gt;), go the other&lt;br&gt;
way: force the Intel GPU primary with &lt;code&gt;apple_gmux.force_igd&lt;/code&gt; so the Radeon can&lt;br&gt;
sleep. That is the right goal when the dGPU is only burning power, and it is&lt;br&gt;
the path those pages document for the freezes and shutdowns. On this boot the&lt;br&gt;
panels were already on the Radeon, so it was powered either way. The waste was&lt;br&gt;
painting on the other chip and copying. Forcing Intel primary would be a&lt;br&gt;
different setup, with its own mux change, and was not tried.&lt;/p&gt;

&lt;p&gt;Omarchy's own note on hybrid GPUs covers the colon-splitting trap in&lt;br&gt;
&lt;code&gt;AQ_DRM_DEVICES&lt;/code&gt; and the udev names. The piece that note does not spell out,&lt;br&gt;
and the reason a correct &lt;code&gt;AQ_DRM_DEVICES&lt;/code&gt; can still leave you on the Intel&lt;br&gt;
renderer, is the probe order on this Mac: the Radeon appears after Hyprland&lt;br&gt;
has already chosen.&lt;/p&gt;

</description>
      <category>linux</category>
      <category>hyprland</category>
      <category>archlinux</category>
      <category>mac</category>
    </item>
  </channel>
</rss>
