Hi, Everyone! After just updating my DGX Spark (Gigabyte) an hour ago, it repeatedly reboots after a minute or so. However, this seems to stop when I unplug my second monitor (connected via USB-C). When connected to the one monitor (HDMI), it seems fine. John

Hi all, thanks for the logs and testing. @dengfa.cheng, @Kirk_SparkX4, @Torkwald, @pr0n, @georgejtcaovii and everyone else who reported back helped us narrow this down quickly. A fix for the reboot loop on kernel 7.0.0-1024-nvidia is now available. It ships as the package nvidia-spark-initcall-bl version 1.0-2. It applies the same kernel boot option that was tested in this thread (initcall_blacklist=power_wrap_init), so you don’t need to add it by hand. If your Spark stays up long enough to run commands Unplugging all monitors (HDMI and USB-C) and connecting over SSH usually gives you enough time. Then run: sudo apt update sudo apt dist-upgrade sudo reboot Please use dist-upgrade, not upgrade. To confirm the fix is installed: dpkg -l nvidia-spark-initcall-bl The version should show 1.0-2. If it reboots before you can do anything, boot the previous kernel Connect a wired USB keyboard. Bluetooth keyboards aren’t active this early in boot. Power the unit off, then power it on. As soon as the NVIDIA logo appears, press and hold Left Shift + Insert until the GRUB menu opens. Select Advanced options for DGX OS GNU/Linux, then DGX OS GNU/Linux, with Linux 7.0.0-1019-nvidia. Once it’s up, run the update commands above and reboot. After the update, the system boots 1024 with the fix applied. If you pinned 1019 as a workaround If you used apt-mark hold or a GRUB_DEFAULT override in /etc/default/grub.d/zz-timeout.cfg, undo it once you’ve applied the update, so you keep getting future kernel and security updates: sudo apt-mark showhold sudo apt-mark unhold sudo rm /etc/default/grub.d/zz-timeout.cfg sudo update-grub sudo reboot If you added initcall_blacklist=...power_wrap_init to your kernel command line by hand, it’s no longer needed after the update. If you can’t reach GRUB at all A recovery section is being added to the DGX Spark User Guide. It covers repairing an install from a live USB without wiping your data, and I’ll link it here once it’s published. Please don’t use the factory System Recovery unless you’re prepared to lose everything on the SSD. If you still see reboots after you’re on 1.0-2, please reply with your system model, the output of uname -r and cat /proc/cmdline, and attach an nvidia-bug-report.

[image] ashreaperone: I am having the same issue. Any update? Are you having the issue of the DGX Spark reboot very 30 seconds when connected to an USBC/DP display? When did you do the upgrade? Dashboard or cli?

I am also having the same issue. It started after I updated to 1024 using the dashboard. When a monitor is attached (HDMI or USB-C) the GB10 will crash and reboot itself in a loop. With the monitor disconnected, I can remotely connect. Booting 1019 resolves the issue; my GB10 is running for over 24h without a problem.

Same Issue after update to 1024 kernel on MSI EdgeXpert. I got a stable SSH Connection after disconnecting all monitors from the spark. Via SSH I was able to fix it as follows (Solution more or less the same as above: Go back to 1019. ) Via an ssh connection: Step 1: Make the GRUB menu visible and pin 1019 as default: — Make the GRUB menu visible and pin 1019 as default The NVIDIA default config in /etc/default/grub.d/*.cfg sets GRUB_TIMEOUT=0 and hides the menu, and editing /etc/default/grub alone has no effect. Create a file that sorts last alphabetically and overrides everything: bash cat DGX OS GNU/Linux, with Linux 7.0.0-1019-nvidia" EOF sudo update-grub Step 2 — Hold the kernel meta-packages so updates don’t reinstall 1024 bash sudo apt-mark hold linux-nvidia-hwe-24.04 \ linux-image-nvidia-hwe-24.04 \ linux-headers-nvidia-hwe-24.04 \ linux-tools-nvidia-hwe-24.04 \ linux-modules-nvidia-580-open-nvidia-hwe-24.04 \ linux-modules-nvidia-fs-nvidia-hwe-24.04 (Once a fixed kernel is released: sudo apt-mark unhold and update normally.) Step 3 — Disable auto-suspend (headless operation) While running headless, the Spark suspended itself after inactivity (power LED off, chassis warm, SSH dead). Mask the sleep targets: (My spark went into suspended mode over night and got warm - in fact this is another issue not directly connected to this update issue) bash sudo systemctl mask sleep.target suspend.target hibernate.target hybrid-sleep.target Step 4 — Reboot and verify bash sudo reboot After reconnecting: bash uname -r # → 7.0.0-1019-nvidia 3. Result After two restarts with the 1019 kernel, the system appears stable again Remark: I got very good help from Perplexity

Acer Veriton GN100 — repeated unexpected reboots on 7.0.0-1024-nvidia; successful temporary boot into 1019 I’m reporting the same issue on an Acer Veriton GN100 running NVIDIA DGX OS 7.6.0. System configuration Hardware: Acer Veriton GN100, NVIDIA GB10, 128 GB unified memory OS: NVIDIA DGX OS 7.6.0, aarch64 Affected kernel: 7.0.0-1024-nvidia Previous kernel: 7.0.0-1019-nvidia NVIDIA driver: 580.178.04 CUDA version: 13.0 Symptoms On November 1, 2026, the GN100 repeatedly rebooted unexpectedly following recent system updates. The connected monitor entered power-saving mode and lost its video signal. The computer restarted unexpectedly, sometimes within minutes of the previous boot. The failures occurred during light desktop activity, without heavy GPU inference workloads. Changing power strips and checking physical power connections did not eliminate the failures. SSH access was intermittently available between reboots. Diagnostic evidence last -x recorded multiple abnormal session terminations, including a session from approximately 06:40 to 06:44 PST classified as crash. Kernel logs also contained: A call trace referencing cpuidle_enter_state Correctable PCIe AER physical-layer receiver errors (RxErr) Journal recovery following unclean shutdowns These observations do not establish the precise root cause, but they may help NVIDIA correlate this report with the ongoing investigation. Recovery procedure I verified that kernel 7.0.0-1019-nvidia, its initrd, kernel modules, and GRUB entry remained installed. Using SSH, I scheduled a one-time boot into the previous kernel: sudo grub-reboot '1>2' sudo grub-editenv list This confirmed: next_entry=1>2 After a controlled reboot, verification showed: uname -r 7.0.0-1019-nvidia The NVIDIA GB10 initialized normally, and nvidia-smi reported driver 580.178.04, CUDA 13.0, and normal GPU temperatures. The one-time GRUB selection was subsequently consumed (next_entry=). Current status The GN100 successfully booted into 7.0.0-1019-nvidia. Longer-term stability testing is still underway. I have not permanently changed the default kernel or pinned kernel packages. Questions for NVIDIA Can NVIDIA confirm whether 7.0.0-1024-nvidia has a known regression affecting Acer GN100 systems, including CPU idle, display, or watchdog behavior? Is 7.0.0-1019-nvidia an officially recommended temporary recovery kernel for affected GN100 owners? What is the recommended security-safe procedure for maintaining the older kernel until a fixed release is available? Are there any known implications for ConnectX-7/RDMA and multi-node distributed inference on the older kernel? What additional diagnostic information would be most useful to NVIDIA? I can provide relevant redacted kernel logs and boot history if needed. Thank you for investigating this issue.

For what it’s worth, I have two founders editions that I factory reset last night. After resetting, and using Nvidia Sync to update 100% up to date, no issues at all.

ASUS Ascent GX10 — same 7.0.0-1024-nvidia reboot loop; 7.0.0-1019-nvidia stable Hardware: ASUS Ascent GX10 (partner GB10), DGX OS 7.6.0 (Dashboard OTA 2026-11-02), driver 580.178.04 open. After the OTA the unit ran 7.0.0-1024-nvidia and rebooted uncleanly about every 2 minutes. No Xid / panic / Oops in the boots I sampled. Journals ended mid-work. Near the cuts I saw correctable PCIe AER Physical Layer RxErr and cx7-pcie-hotplug … Cable removal. Temps at boot were ordinary. An HDMI monitor was connected. Unplugging that HDMI cable let 1024 stay up past ~5 minutes; it was only a mitigation. Recovery: installed linux-image-7.0.0-1019-nvidia, set GRUB default to the Advanced-options 1019 entry (and zz-timeout.cfg so no-grubmenu.cfg does not hide the menu), held the HWE metapackages plus the 1019 image/headers, rebooted. Now: uname -r → 7.0.0-1019-nvidia. nvidia-smi → driver 580.178.04, GPU NVIDIA GB10. Same evening, after the OTA, linux-modules-nvidia-580-open for the running kernel was also missing until installed (nvidia-smi could not talk to the driver). Separate from the reboot loop, but easy to hit on the same update path.

Can someone try this workaround? Add this in the kernel argument. initcall_blacklist=tegra234_cbb_init,power_wrap_init tegra234_cbb_init should be already there. Just need to add power_wrap_init

Feedback: The kernel arguments have kept my GX10 running for over 40 minutes without sign of freezing/restarting. I was also able to log in to the Founders Edition’s GUI, which previously restarted immediately after entered the password when login. It looks effective so far.

Not being able to control fans using /sys/bus/arm_ffa/devices/arm-ffa-17 on 1024 is a dealbreaker , the unit runs way too hot, and freeze / reboot without it. Any way to make the default fan profile more aggressive or give back user override on 1024 ?

DGX Spark repeatedly rebooting after most recent update

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10

ashreaperone November 2, 2026, 4:03am 33

Yep, two monitors (HDMI/USB-C). Checked the CLI yesterday and confirmed I’m caught up on all updates (last update was a dashboard and it was browser update a few days ago).

How safe is it to update right now?

Topic		Replies	Views
Random reboots and 00 screen DGX Spark / GB10 reboot	46	2485	April 20, 2026
DGX spark keeps rebooting every 20-30 minutes DGX Spark / GB10 boot	6	1921	March 19, 2026
`sudo reboot now` and Spark doesn't come back.. started after installing updates today DGX Spark / GB10	11	839	October 31, 2026
After the latest update of DGX SPark it's roasting like a hell in stale DGX Spark / GB10 dgx-spark-issue	7	1149	August 4, 2026
Any use of the HDMI (TMDS) output causes a hard reset on kernel 7.0.0-1024-nvidia DGX Spark / GB10 kernel , driver , linux-driver , reboot , dgx-spark-issue	13	601	November 3, 2026
DGX Spark reboots every ~20 minutes DGX Spark / GB10	11	872	September 3, 2026
DGX Spark resetting at idle after latest DGX Dashboard update (kernel 7.0.0-1024) DGX Spark / GB10 kernel , dgx-spark-issue	1	384	October 30, 2026
DGX Spark constantly shut down after first set up DGX Spark / GB10	8	793	January 11, 2026
Spark abruptly shuts down DGX Spark / GB10 dgx-spark-issue	47	4339	October 9, 2026
New DGX OS 7.4.0 DGX Spark / GB10	39	4481	March 28, 2026

DGX Spark repeatedly rebooting after most recent update

Related topics