Ubuntu on the HP OmniBook Ultra 16 (RTX Spark N1X)

Tested in November 2026 on the HP OmniBook Ultra Laptop 16-bk0xxx (board 8E6A), BIOS F.06

Ubuntu 26.04 runs well on the HP OmniBook Ultra 16, but out of the box the GPU, speakers and battery don't work. You can fix all three in an afternoon with packages and two community workarounds. This post covers what I hit, how I fixed each problem, and what still doesn't work.

The laptop uses NVIDIA's RTX Spark N1X: an Arm chip (10 Cortex-X925 and 10 Cortex-A725 cores) with a Blackwell GPU, plus MediaTek Wi-Fi and audio. Support for this hardware is very new, and most of the fixes below aren't in Ubuntu yet.

At a glance

AreaOut of the boxAfter the fixes
Wi-Fi, Bluetooth, SSD, card readerWorksWorks
Webcam (5MP + IR), touchpad, touchscreen, keyboardWorksWorks
NVIDIA GPU, brightness, external displaysBasic framebuffer onlyWorks (driver 595)
Speakers, headphone jack, internal mic"Dummy Output"Works (speakers sound thinner than on Windows)
Battery, charger, lid, thermals, USB-C portsNot detectedWorks
Fan speed readingStuck at 0 RPMLive reading
Keyboard backlight controlNoneNone (the Fn key may still work)
Thunderbolt / USB4Not workingNot working
HibernateNot supportedNot supported

Which Ubuntu and kernel to use

Use Ubuntu 26.04 LTS (arm64) with the linux-nvidia-bos kernel. I ran 7.0.0-2024-nvidia-bos. This NVIDIA-specific kernel includes the N1X platform support: the MediaTek audio and SoundWire drivers, the bridge to the embedded controller, and USB.

What works out of the box

HardwareDriverNotes
MediaTek MT7925 Wi-Fi 7mt7925eThe log says sar cnt = 0 (no HP power-limit table). Harmless
MediaTek Bluetoothbtusb
KIOXIA 1 TB NVMe SSDnvme
HP 5MP camera + IR camerauvcvideoFace unlock with the IR camera needs extra software (e.g. Howdy)
Realtek RTS525A card readerrtsx_pci
Synaptics touchpad, ELAN touchscreen, keyboardHID
HDMI/DisplayPort audiosnd_hda_codec_tegrahdmi
TPM 2.0

What's broken out of the box

  1. GPU (10de:2e03): no driver. The desktop runs on simpledrm, a basic framebuffer. That means no acceleration, no brightness control, and no external monitors.

  2. Sound: "Dummy Output" only. The hardware is four Cirrus CS35L56 speaker amps and a CS42L43 headset/mic chip, on MediaTek's mt8901 SoundWire controller. The kernel drivers load fine. What's missing is the ALSA audio profile (UCM) that tells PipeWire how to route sound, so PipeWire falls back to a dummy device.

  3. Battery, charger, lid, thermals, fan and USB-C: not detected. /sys/class/power_supply/ is empty and the lid state reads unsupported. The kernel log repeats this firmware error dozens of times per boot:

    ACPI BIOS Error (bug): AE_AML_BUFFER_LIMIT, Field [DATA] at bit offset/length 208/400 exceeds size of target Buffer (512 bits)
    ACPI Error: Aborting method \_SB.FFA0.RBUF due to previous error (AE_AML_BUFFER_LIMIT)
    

    HP's firmware defines a 50-byte field inside a 64-byte buffer, 26 bytes in, so it runs 12 bytes past the end. That method (\_SB.FFA0.RBUF) handles every read from the embedded controller. So everything that talks to the controller fails: battery, AC adapter, thermal zone, fan, USB-C, lid, and one of the GPU's firmware calls.

    This also breaks power-based settings. With no AC adapter detected, GNOME and systemd assume you're always plugged in, so the "on battery" sleep timeout never applies.

Fix 1: NVIDIA GPU driver

Install the open-kernel driver together with the prebuilt modules for the nvidia-bos kernel:

sudo apt install --no-install-recommends nvidia-driver-595-open linux-modules-nvidia-595-open-nvidia-bos
sudo reboot

If you get a black screen: pick Advanced options → recovery mode in the boot menu (GRUB), then run apt purge 'nvidia-*'.

Fix 2: Speakers and microphone

NVIDIA has submitted an audio profile for this sound chip, but it isn't merged yet: alsa-ucm-conf PR #862. It covers six N1X laptop layouts. The OmniBook Ultra 16 is the one with four CS35L56 amps on link 0 and a CS42L43 on link 1. Ubuntu's own alsa-ucm-conf (1.2.15.3-1ubuntu1.7 at the time of writing) doesn't include it yet.

You can install it for your user only, without touching /usr. Build a private profile folder that links to the system's files, add the new profile, and point PipeWire at it with ALSA_CONFIG_UCM2:

#!/usr/bin/env bash
set -euo pipefail
SHA=c22cf4c960c7f68e13420140e6ec38f1e0ce70b5   # PR #862 head as of 2026-11-21
RAW=https://raw.githubusercontent.com/dcemin-nv/alsa-ucm-conf/$SHA/ucm2
U=$HOME/.config/alsa/ucm2; SYS=/usr/share/alsa/ucm2

mkdir -p "$U/conf.d" "$U/MediaTek/mt8901-soundwir" "$U/conf.d/mt8901-soundwir"
for e in "$SYS"/*; do b=$(basename "$e"); case $b in conf.d|MediaTek) ;; *) ln -sfn "$e" "$U/$b";; esac; done
for e in "$SYS"/conf.d/*;   do ln -sfn "$e" "$U/conf.d/"; done
for e in "$SYS"/MediaTek/*; do ln -sfn "$e" "$U/MediaTek/"; done

for f in mt8901-soundwir.conf HiFi-cs35l56x4-cs42l43.conf HiFi-cs42l43-cs35l56x4.conf \
         HiFi-cs42l43-cs35l56x2.conf HiFi-rt1320x2-rt712.conf HiFi-rt1320x2-rt712-refboard.conf \
         HiFi-rt1321x2-rt713.conf; do
  curl -sSfo "$U/MediaTek/mt8901-soundwir/$f" "$RAW/MediaTek/mt8901-soundwir/$f"
done
ln -sfn ../../MediaTek/mt8901-soundwir/mt8901-soundwir.conf "$U/conf.d/mt8901-soundwir/mt8901-soundwir.conf"

for s in pipewire wireplumber; do
  mkdir -p ~/.config/systemd/user/$s.service.d
  printf '[Service]\nEnvironment=ALSA_CONFIG_UCM2=%%h/.config/alsa/ucm2\n' > ~/.config/systemd/user/$s.service.d/10-alsa-ucm2.conf
done
mkdir -p ~/.config/environment.d
echo 'ALSA_CONFIG_UCM2=${HOME}/.config/alsa/ucm2' > ~/.config/environment.d/10-alsa-ucm2.conf

systemctl --user daemon-reload
systemctl --user restart wireplumber pipewire pipewire-pulse

Check that it loaded:

ALSA_CONFIG_UCM2=~/.config/alsa/ucm2 alsaucm -c hw:mt8901soundwire list _verbs   # expect: HiFi
wpctl status   # expect "Internal speakers", "Headphone jack", "Internal microphone array"

The profile chooses a layout based on which SoundWire devices are present. On this laptop sdw:0:0:01fa:3556:01:0 exists, so it picks HiFi-cs35l56x4-cs42l43.conf.

Fix the hissing microphone

Once the mic works it records loud hiss, because PipeWire's 100% mic volume equals the CS42L43's maximum boost of +31.5 dB (Decimator 3/4 Volume = 191 of 191). At that level the mic's own noise is boosted until it clips, even in a silent room. I measured silence by mic volume:

PipeWire mic volumeHardware gainSilent-room level
100%+31.5 dBclipping, about 9% of samples at full scale
40%+7.5 dBsteady, about −54 dBFS
30%0 dBabout −62 dBFS

Set it to about 40% and leave it there. WirePlumber remembers it:

wpctl set-volume @DEFAULT_AUDIO_SOURCE@ 0.4

Speaker caveat

The speaker amps report Cirrus Logic CS35L56 Rev B0 OTP3 fw:3.4.4 (patched=0). That's because HP's tuning files for this laptop (subsystem 103c8e6a) aren't in linux-firmware. Without them the amps run in a basic fallback mode, so expect the speakers to sound thinner and quieter than on Windows. There's no fix until HP or Cirrus submits the files.

When Ubuntu ships the profile, remove the override so you get updates:

rm -r ~/.config/alsa/ucm2 ~/.config/systemd/user/*/10-alsa-ucm2.conf ~/.config/environment.d/10-alsa-ucm2.conf

Fix 3: Battery, charger, lid, thermals and USB-C

There's no BIOS fix yet. You also can't load corrected firmware tables (a DSDT override) at boot, because Secure Boot puts the kernel in lockdown, which blocks it. The workaround is omnibook_ffa_fix, a small community kernel module from keejkrej/omarchy-omnibook-ultra-16 (root cause in Omarchy issue #18112). It makes two in-memory fixes when it loads:

After it loads, it re-probes the battery, AC adapter, thermal zone, USB-C and fan devices.

Read the source before you load it. It works by changing the kernel's internal ACPI structures (ACPICA) directly, at hard-coded positions: aml_start at +32 and aml_length at +48 in struct acpi_object_method. I checked these against Linux 7.0 (ACPICA 20251212) and they match. The module also checks the object's type before writing anything. Still, if a future kernel changes that structure, re-check it.

Install it with DKMS, so it's rebuilt for every kernel update:

sudo apt install build-essential dkms
git clone https://github.com/keejkrej/omarchy-omnibook-ultra-16
sudo mkdir -p /usr/src/omnibook-ffa-fix-1.0.0
sudo cp omarchy-omnibook-ultra-16/kmod/{omnibook_ffa_fix.c,Makefile,dkms.conf} /usr/src/omnibook-ffa-fix-1.0.0/
sudo dkms add -m omnibook-ffa-fix -v 1.0.0
sudo dkms install -m omnibook-ffa-fix -v 1.0.0
echo omnibook_ffa_fix | sudo tee /etc/modules-load.d/omnibook-ffa-fix.conf

Secure Boot: enroll the signing key

DKMS signs the module with a local key, /var/lib/shim-signed/mok/MOK.der. The kernel won't load it until that key is enrolled. Until then you'll see Key was rejected by service and a failed systemd-modules-load.service.

sudo mokutil --import /var/lib/shim-signed/mok/MOK.der   # choose a one-time password
sudo reboot

Watch the screen during this reboot. A blue "MOK management" screen appears and continues booting after about 10 seconds, which drops the request. I missed it the first time. Press a key straight away, then choose Enroll MOK → Continue → Yes, type the password, and reboot. Check it worked with mokutil --test-key /var/lib/shim-signed/mok/MOK.der, which should say "already enrolled".

Result

omnibook_ffa_fix: Patched \_SB.FFA0.RBUF CreateField DATA bit length (0x0190 -> 0x0130 at +0x64)
omnibook_ffa_fix: Redirected \_SB.BAT0._BST to evaluate \_SB.BAT0.UPBS()
ACPI: battery: Slot [BAT0] (battery present)
ACPI: AC: AC Adapter [ADP1] (on-line)
ACPI: thermal: Thermal Zone [TZ01] (47 C)
$ ls /sys/class/power_supply/
ADP1  BAT0  ucsi-source-psy-USBC000:001  ucsi-source-psy-USBC000:002
$ upower -i /org/freedesktop/UPower/devices/battery_BAT0 | grep -E 'state|percentage|energy-full'
    state:               charging
    energy-full:         98.463 Wh
    energy-full-design:  99.33 Wh
    percentage:          35%

The battery percentage, time remaining, charger detection, lid switch, the second thermal zone and both USB-C ports all work now. GNOME's AC and battery power settings behave properly again. You'll still see about 70 AE_AML_BUFFER_LIMIT errors in the first 5 seconds of each boot, before the module loads. That's expected, and none appear after it.

Fix 4: Fan speed reading

With Fix 3 in place, the fan still reads 0 RPM, even under load. HP's \_SB.FAN0._FST method only returns a stored value (F1CS). That value is refreshed by F1LS() only when the embedded controller sends a fan event, and all of those failed before the module loaded. I added a few lines to the module so _FST calls F1LS() first, which makes every read live. It applies on top of the module from Fix 3; bump PACKAGE_VERSION in dkms.conf to 1.1.0 when you reinstall.

fan-speed.patch (applies to kmod/omnibook_ffa_fix.c)
--- a/omnibook_ffa_fix.c
+++ b/omnibook_ffa_fix.c
@@ -67,6 +67,18 @@
 static u8 *orig_bst_aml;
 static u32 orig_bst_len;
 
+/*
+ * \_SB.FAN0._FST only returns F1CS, which F1LS() refreshes from the EC when
+ * an EC fan event arrives. Prepend a call to F1LS so every _FST read is live:
+ *   46 31 4c 53       F1LS ()   (resolved in the enclosing FAN0 scope)
+ */
+static const u8 fst_prefix_aml[] = { 0x46, 0x31, 0x4c, 0x53 };
+
+static void *fst_method_obj;
+static u8 *orig_fst_aml;
+static u32 orig_fst_len;
+static u8 *patched_fst_aml;
+
 static struct delayed_work rescan_work;
 static int rescan_attempts;
 
@@ -159,6 +171,45 @@
 	return 0;
 }
 
+static int patch_fst_aml(void)
+{
+	acpi_handle h;
+
+	if (ACPI_FAILURE(acpi_get_handle(NULL, "\\_SB.FAN0.F1LS", &h))) {
+		pr_warn("\\_SB.FAN0.F1LS not found, skipping _FST patch\n");
+		return -ENODEV;
+	}
+
+	fst_method_obj = get_acpica_method_obj("\\_SB.FAN0._FST");
+	if (!fst_method_obj) {
+		pr_warn("\\_SB.FAN0._FST not found, skipping _FST patch\n");
+		return -ENODEV;
+	}
+
+	orig_fst_aml = *(u8 **)((u8 *)fst_method_obj + ACPICA_METHOD_AML_START_OFFSET);
+	orig_fst_len = *(u32 *)((u8 *)fst_method_obj + ACPICA_METHOD_AML_LEN_OFFSET);
+	if (!orig_fst_aml || !orig_fst_len || orig_fst_len > 4096) {
+		pr_warn("\\_SB.FAN0._FST invalid AML pointer/length (%u)\n", orig_fst_len);
+		orig_fst_aml = NULL;
+		return -EINVAL;
+	}
+
+	patched_fst_aml = kmalloc(sizeof(fst_prefix_aml) + orig_fst_len, GFP_KERNEL);
+	if (!patched_fst_aml) {
+		orig_fst_aml = NULL;
+		return -ENOMEM;
+	}
+	memcpy(patched_fst_aml, fst_prefix_aml, sizeof(fst_prefix_aml));
+	memcpy(patched_fst_aml + sizeof(fst_prefix_aml), orig_fst_aml, orig_fst_len);
+
+	*(u32 *)((u8 *)fst_method_obj + ACPICA_METHOD_AML_LEN_OFFSET) =
+		sizeof(fst_prefix_aml) + orig_fst_len;
+	*(u8 **)((u8 *)fst_method_obj + ACPICA_METHOD_AML_START_OFFSET) = patched_fst_aml;
+
+	pr_info("Patched \\_SB.FAN0._FST to refresh fan speed via F1LS()\n");
+	return 0;
+}
+
 static void attach_platform_dev_by_name(const char *name)
 {
 	struct device *dev;
@@ -229,6 +280,7 @@
 		return ret;
 
 	patch_bst_aml();
+	patch_fst_aml();
 
 	INIT_DELAYED_WORK(&rescan_work, rescan_ffa_consumers);
 	rescan_ffa_consumers(&rescan_work.work);
@@ -240,6 +292,12 @@
 {
 	cancel_delayed_work_sync(&rescan_work);
 
+	if (fst_method_obj && orig_fst_aml) {
+		*(u8 **)((u8 *)fst_method_obj + ACPICA_METHOD_AML_START_OFFSET) = orig_fst_aml;
+		*(u32 *)((u8 *)fst_method_obj + ACPICA_METHOD_AML_LEN_OFFSET) = orig_fst_len;
+	}
+	kfree(patched_fst_aml);
+
 	if (bst_method_obj && orig_bst_aml) {
 		*(u8 **)((u8 *)bst_method_obj + ACPICA_METHOD_AML_START_OFFSET) = orig_bst_aml;
 		*(u32 *)((u8 *)bst_method_obj + ACPICA_METHOD_AML_LEN_OFFSET) = orig_bst_len;
@@ -259,4 +317,5 @@
 
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("keejkrej");
-MODULE_DESCRIPTION("HP OmniBook Ultra 16-bk0xxx ACPI FF-A EC RBUF & Battery Fix");
+MODULE_DESCRIPTION("HP OmniBook Ultra 16-bk0xxx ACPI FF-A EC RBUF, Battery & Fan Speed Fix");
+MODULE_VERSION("1.1.0");

I ran all 20 cores flat out for 75 seconds to test it:

TimeFan (fan_speed_rpm)SoC / system temp
Idle0 RPM (the fan really is off)51 °C / 52 °C
15 s2658 RPM98 °C / 97 °C
30–75 sabout 3990 RPM95–98 °C
20 s after the test3711 RPM57 °C / 76 °C

What still doesn't work

ProblemWhat's going onOutlook
Keyboard backlight controlLinux sees no keyboard backlight. I checked the firmware tables: HP gives the operating system no control, only a BIOS timeout setting. The embedded controller handles brightness, presumably via the Fn key.No OS control until HP adds one. Try the Fn key.
Thunderbolt / USB4The USB4 driver loads but no USB4 controller ever appears, and the log shows NVIDIA USB4 disable _DSM: returned 0x00 (failure). Docks and enclosures that need USB4/Thunderbolt probably won't run at full speed.Waiting on NVIDIA/HP. Ordinary USB-C devices and charging work.
USB-C alternate modesucsi_acpi: failed to register partner alt modes (-5), GET_CABLE_PROPERTY failed.Video over USB-C to some docks may not work. Untested.
Speaker tuningNo HP tuning files for the CS35L56 amps (patched=0).Waiting on linux-firmware.
HibernateThere's no /sys/power/disk on this platform.Normal for Arm laptops.
Slow ACPI warningsworkqueue: acpi_os_execute_deferred hogged CPU for >20000us: each embedded-controller call over the FF-A bridge takes over 20 ms.Cosmetic. It may cost a little battery life.

Not yet tested: suspend/resume (I haven't run a sleep cycle yet), an external monitor over HDMI or USB-C, the headphone jack and headset mic with something plugged in, and the HP-specific function keys.

Keeping it working

Sources