There was an issue with an update a while ago. I haven’t updated in a while, and like everyone, don’t want to destabilize the system. I usually use the series of commands reccomended on nvidia’s site, since mine is headless and I try to keep it lean:
sudo apt update
sudo apt dist-upgrade
sudo fwupdmgr refresh
sudo fwupdmgr upgrade
sudo reboot
Do you know if the issues have been addressed or should I wait?
Thanks!
I got burned by the original accidental push of kernel 7, well the version of kernel 7 they distribute now is working perfectly fine.
All update issues have been addressed, you should be able to update your system to the latest version. If you run into any other issues, please let us know.
Only issue I’m running into is firmware updates aren’t coming down for the versions needed to qualify for OTA2610 for me on an ASUS GX10.
Installed vs OTA2610-required
| Component | Installed | OTA2610 Requires |
|---|---|---|
| SOCFW | 2.155.14 |
2.158.15 |
| EC | 3.5.11 |
3.13.7 |
| EC Unfused | 3.5.11 |
3.13.7 |
| USBPD | 5.22 |
0.5.25 |
gx10-77e0:~$ fwupdmgr upgrade
Devices with the latest available firmware version:
• Embedded Controller
• UEFI CA
• UEFI Device Firmware
• UEFI Device Firmware
Devices with no available firmware updates:
• ASUS KEK Certificate
• ASUS UEFI SW Key Certificate
• ESL01TBTLCZ-27J2-TYN
• KEK CA
• SBAT
• UEFI dbx
• Windows UEFI CA
I looked at asus bios update download file, and the versions that offers doesn’t seem to fit OTA2610 either. I’m confused as to what I can do to meet OTA2610. When I check what I’ve got it always says I’m on OTA2607.
My dashboard says no available updates. I ran sudo apt update && sudo apt dist-upgrade -y && sudo fwupdmgr refresh && sudo fwupdmgr upgrade -y && sudo reboot to update. Is it a matter of waiting more? I feel like it’s been months and still can’t get the latest version from the firmware updates.
I asked a similar question a while ago: How safe is it to actually let my Sparks restart at this point? - #4 by 2893f57acbe7739fc460e9bd7
Thankfully these at least don’t auto just update.
Honestly, my machines are fairly locked down ufw-wise, and they’re running headless anyway. So I’m not particularly worried about security updates and whatnot. They’re just appliances, and specifically locked down to only be able to serve my LAN. I’ll wait until a stack actually needs me to update, probably.
Ok great thanks for the quick reply!
Same picture on our ASUS GX10, checked today (October 24). On LVFS the newest GX10 firmware is still the release of September 1 (its EC and SoC capsules), while the Founders Edition has had OTA2610’s there since September 28 (EC 3.13.7, SoC 2.158.15, USB PD 0.5.25). NVIDIA’s release notes say that GB10-based partner systems may not receive updates at the same time, so fwupdmgr is right that you have the latest it can offer: it is a matter of waiting for ASUS.
I updated mine very recently and it’s now failing to reboot properly when no monitor is attached:
I don’t know if this is widespread though or just me. Of course, if you have a monitor attached then it’s probably not a concern anyway (I’m not 100% certain it’s the monitor, but it seems to reliably happen that way in the testing I have done).
Earlier today I updated (to kernel 7 and the fw update that allows limiting reserved display memory) and it doesn’t appear sudo poweroff actually turns the system off. Still troubleshooting…
Not sure if you saw my link above yours - it sounds very much like the issue I’m also having since updating (I just noticed it trying to reboot and not power off).
I successfully updated both of my DGX Sparks last night for the first and this morning for the second.
The adventure, drama and excitement I see with every DGX Spark update belongs in a theme park. I haven’t updated in months and still can’t find the nerve to queue for the ride. Replacing these machines looks like a whole other park.
Yes, at least where firmware or kernel updates are involved, the procedure doesn’t seem to be very well tested, which consequently leads to frequent problems and reduces trust.
The procedures may be complex, and it’s worth noting that Nvidia responds quickly to problem reports and usually provides a fix within 1-2 business days. Of course, it would be better if this weren’t necessary in the first place.
Both Sparks completed the update and then crashed twice. Hopefully this last update holds. This last firmware push is not smooth.
Serious question, but what is the reason to update to begin with? I guess security reasons aside. It seems like updates are just poorly tested and don’t really improve speeds?
The eMudhra issue last week seemed like quite a serious problem. I updated with the hope/assumption that it would remove their certs from the OS store (if they were there), reducing the risk of MITM attacks for HTTPS.
Updates are supposed to fix things, not break them. It unfortuantely says a lot about the state of software if this isn’t often the case these days and it’s better to avoid updates than install them.
I’m conflicted about whether AI is good in this regard. On one hand, it seems like it should be easier/faster than ever to find and fix bugs. On the other hand, it’s also easier/faster to produce them 😞
(that’s not to say I think AI is responsible for any of these issues, it’s more of a comment on the software industry 🙃)
The very latest firmware released 3 days ago allows you to allocate 1.5GB more to CUDA by reducing the reserve on headless Sparks. 3GB extra across both my Sparks would be welcome, but the ASUS GX10 gets firmware updates ~3 months after they are released on the DGX Spark FE.
3GB Extra sounds awesome..
still after reading some posts
Stuck at grub after last update
DGX Spark repeatedly rebooting after most recent update
`sudo reboot now` and Spark doesn’t come back.. started after installing updates today
GIGABYTE AI TOP ATOM: stopped fan, severe heat and loss of access after updates
I am quite afraid of updating
what is the savest method to update?
the click of that button or a terminal command?
That depends so much on how you use your spark(s).
I am not experimenting / playing around with them. I simply need them running 24/7 as LLM providers. So for me reliability is key. I switched to the new Deepseek V4.1 release a few days ago and therefore took the opportunity to update firmware etc. That was not a good idea. Took me 2 days to straighten the system out. For me: If it’s not broken, don’t fix it.
If your main purpose is to learn and do things, then go for it. Be experimental, with all the benefits and all the consequences.
ok I wait until there is a “stable” update
cause my system runs fine and stable so far.
