Include intel_idle Kernel Driver in TrueNAS SCALE to Improve Idle Power Efficiency on Intel CPUs

Problem/Justification
TrueNAS SCALE currently disables the intel_idle driver by default. This prevents Intel CPUs from entering deep C-states like C6–C10, which is critical for achieving low idle power consumption. On modern platforms (12th Gen and newer), intel_idle provides significantly better power management than the fallback acpi_idle driver.

As a result, users running TrueNAS SCALE on energy-efficient hardware—especially home lab users and those deploying 12th Gen, 13th Gen, or newer Intel CPUs—experience unnecessarily high idle power consumption. Community members have attempted workarounds (e.g., kernel boot parameters like intel_idle=force), but these fail because the driver is not included in the TrueNAS kernel (modprobe intel_idle fails, and PowerTOP confirms no deep C-states).


Impact
Re-enabling or supporting the intel_idle driver would benefit TrueNAS users on Intel platforms by significantly reducing idle power draw, thermal output, and long-term energy costs.

Benefits:

  • Lower idle power draw (~10–15W savings possible)
  • Cooler and quieter systems
  • Better efficiency for 24/7 deployments
  • No behavior change unless deliberately enabled

Disadvantages:

  • Slightly increased kernel footprint
  • Potential misuse by inexperienced users (can be mitigated with documentation or a toggle)

User Story
I am running TrueNAS SCALE on an Intel Core i7-13700K system. On Debian- or Ubuntu-based distributions with intel_idle support, my system idles below 10W with deep C-state residency. On TrueNAS SCALE, idle power is consistently 10–20W higher due to lack of intel_idle.

Even with proper kernel boot arguments (intel_idle=force, intel_idle.max_cstate=9, processor.max_cstate=9), the driver fails to load.

  • modprobe intel_idle fails with:
    FATAL: Module intel_idle not found in directory /lib/modules/...
  • dmesg | grep intel_idle returns nothing
  • PowerTOP shows only: C1_ACPI, C2_ACPI, C3_ACPI, confirming fallback to acpi_idle

This issue directly impacts power-conscious deployments and home lab users running 24/7 systems. If TrueNAS SCALE supported intel_idle, I could dramatically improve idle efficiency without losing platform functionality.

Request
Please consider enabling intel_idle in the TrueNAS SCALE kernel (either built-in or as a module), or providing a supported method to activate it through system settings. This would bring SCALE in line with other modern Linux distros and deliver meaningful power savings to a wide segment of the user base.

7 Likes

Update: This feature request is gaining traction — we’re now at 22 votes! Thanks to everyone who’s supported this so far.

If you’re running TrueNAS SCALE on a modern Intel CPU (especially 12th Gen and up) and noticing higher-than-expected idle power draw, please consider adding your vote or sharing your use case. The more visibility this gets, the more likely it is to catch the attention of the dev team.

Feature Request Link

Quite the vote count. I imagine the number of users with TrueNAS on Intel N-series is driving this a little bit.

One consideration I have immediately is that there were some (older, yes) reports about potential instability at idle with AMD CPUs if this driver is enabled.

this is not an official yes/no, just my own observations

1 Like

Assuming that these instabilities are confirmed and still exist, this looks like a problem which could be solved by a toggle with a fat red warning about enabling intel_idle only with intel CPUs.
(Insert “Big Red Button” meme here…)

2 Likes

Thoughts/Opinions:
Perhaps this can be a CE feature. I like the idea of it being user configurable in a safeish way personally, and I might actually use this feature for benchmarks/burn-ins on my systems. If it is accepted, of course.

However, there is a general downside.

It’s usually a hardware problem, or like, a hardware problem that Intel/AMD caught too late in their dev cycles and had to fix after in software a.la. Spectre/Meltdown. Over the years both Intel and AMD, older platforms and even very recent ones, have had firmware bugs with logic operations and this is all very well documented.
But there has also been alot of chatter about issues with very low power states from both vendors, and these types of things have resulted in system crashing and instability (including the some of the very platforms we are discussing in this thread, at varying times/firmware revisions over the years, most recently 14th gen).

Very low power C states should work fine on most systems but not all, especially those with older “BIOS” versions. I think the AliExpress stuff might work great here but there’s probably alot of duds from poor QC on there too. It’s a roll of the dice but you can sometimes find good ones. I liked Topping stuff, Netgate used them as an OEM for something and I started buying from them instead for a couple of projects.

I’ve been a big part of the overclocking community for years and years. Generally, the goal is moar power moar fast, but this cycle I spent a long time trying to understand the efficiency curves of what I was doing and it was fun but a waste of money.

I say all this to say, I have personally and professionally (pre-iX) managed Enterprise Grade equipment and also experimented with homelab shenanigans.

I would not want TrueNAS Enterprise users to modify this default very conservative behavior. Stability > Power consumption would be the mindset there.

Community Users remember…Its Open Source. You can do this today in various non-supported ways if you are that type of user, you’ll just be mostly on your own.

Opinion:
A downstream of TrueNAS does exist.

There are other places, like the Forum I linked above where thare are actually other people who might also be interested in “rolling their own solutions” and “coloring in the margains” for “TrueNAS”. In at least 1 example, there was a guy doing Docker on TrueNAS since Angelfish and he actually was actively supporting them for a surprisingly long time, despite doing wildly unsupported things from the iX perspective. Wendell’s community is filled with mad scientists like him and I love him for it.

There’s also HexOS where I’m fairly confident this would not be the default for them, but you would have to check with their forums.

This also can probably be solved in a simple but “unsupported/use at your own risk” way if other users are interested in documenting it in a forum thread. I would recommend “intel_idle.max_cstate=5” for stability if nothing else. This is a good “Use at your own risk” thing, I think.

I will also say this is kinda a new-ish problem for TrueNAS in general in a more TrueNAS Community/Enterprise world. In CORE, the FreeBSD driver stack artificially held us back in terms of hardware support anyway.

The name of this driver is “intel_idle”. Does it even work on AMD CPUs?

Hey everyone,

First off, big thanks to the community and iX staff like Honey Badger and Nick for weighing in on the intel_idlediscussion. I really appreciate the focus on stability and understand why TrueNAS SCALE defaults to conservative CPU power management settings.

From what I gather, deep C-state drivers like intel_idle for Intel and amd_pstate for AMD can cause instability or crashes on some systems, especially depending on BIOS versions or firmware quirks. That’s definitely something enterprise users can’t afford to risk.

That said, for home labs or always-on NAS setups where power efficiency matters, having the option to enable these drivers—or at least have user-tunable settings like max_cstate=5—would be a huge win. It’d let power-conscious users safely balance performance, stability, and energy savings.

Maybe this could be an advanced system tunable or UI toggle with clear warnings, so those who want to experiment or optimize their power use can do so without impacting everyone else.

Also wanted to flag that AMD users face a similar situation: TrueNAS SCALE doesn’t enable the newer amd_pstatedriver by default, which means they’re missing out on better frequency scaling and idle power savings too. Including this alongside intel_idle would help cover a broader set of users.

Thanks again to everyone for the great discussion!

4 Likes

Update: Huge thanks for 31 votes — Let’s keep pushing for better power management!

Hi all,

Wow, I’m really encouraged to see this post already at 31 votes! Thanks so much to everyone who’s taken the time to read, comment, and support the idea of enabling intel_idle in TrueNAS SCALE.

The community clearly cares about making SCALE more power-efficient, especially for home labs and always-on setups where idle power consumption really adds up.

I want to remind folks that while intel_idle unlocks deeper CPU C-states and can dramatically reduce idle wattage, stability and hardware compatibility are understandably top priorities – so I’m fully on board with cautious implementation and user-configurable options.

For AMD users, the similar situation with amd_pstate not being enabled by default is worth considering too – broadening the scope could benefit even more of us.

Please keep sharing your feedback, upvoting if you haven’t already, and posting any testing data or experiences you have. The more the devs see this community interest and get practical info, the better chance we have to get it included in an upcoming release.

Thanks again for the great discussion!

This request got a mention on T3 today so hopefully in the future…

2 Likes

Since this request was mentioned on T3, I wanted to provide a concrete comparison from my own lab to highlight just how much of an impact the missing intel_idle driver has on idle power consumption in TrueNAS SCALE.

I run two systems with similar workloads and ZFS configurations. The key difference? One uses intel_idle (Proxmox), and the other is running SCALE with only acpi_idle.

The results speak for themselves:

Feature TrueNAS SCALE Proxmox (with intel_idle)
CPU Intel i7-13700K Intel i5-12600K
Storage 9 × HDDs 5 × HDDs
Idle Power Draw ~145W ~66W
Idle Driver acpi_idle intel_idle
Deepest C-State C3_ACPI only C10 (pc10)
C10 Residency :x: Not available :white_check_mark: Up to 85% per core
Package Sleep State :x: Pkg(OS) 0.0% :white_check_mark: Pkg(HW) active in C10

Power Consumption Breakdown

While the HDDs and fans contribute to the total power draw, the real difference is primarily due to CPU idle management.

HDD Power Consumption:

  • TrueNAS SCALE (9 HDDs): ~36W (4W per HDD at idle)
  • Proxmox (5 HDDs): ~20W (4W per HDD at idle)

Difference due to HDDs:

  • 36W (TrueNAS SCALE) - 20W (Proxmox) = 16W.
    This means the extra 4 HDDs in TrueNAS SCALE add an additional 16W to the total power draw compared to the Proxmox setup.

Fan Power Consumption:

  • TrueNAS SCALE has 9 fans compared to Proxmox’s 3 fans.
  • Each fan can draw between 2W–5W depending on the speed and model. Assuming 3W per fan for a rough estimate:
    • TrueNAS SCALE: 9 fans × 3W = 27W
    • Proxmox: 3 fans × 3W = 9W

Difference due to fans:

  • 27W (TrueNAS SCALE) - 9W (Proxmox) = 18W
    This means that the additional 6 fans in TrueNAS SCALE contribute about 18W more to the power draw compared to the Proxmox setup.

CPU & Driver Impact:

  • The remaining 45W difference in idle power consumption comes from the missing intel_idle driver in TrueNAS SCALE, which results in less efficient CPU idle states.
  • Proxmox with intel_idle can achieve C10 (PC10) state with up to 85% residency per core, significantly reducing power consumption at idle. Meanwhile, TrueNAS SCALE can only achieve C3_ACPI state with acpi_idle, which is much less efficient.

Community Support

The community has shown strong support for this enhancement. The feature request to include the intel_idle driver in TrueNAS SCALE has garnered 41 votes as of now. This growing interest underscores the demand for enhanced idle power efficiency, particularly among users with modern Intel CPUs.

Conclusion:

While the additional 4 HDDs and 6 fans in TrueNAS SCALE account for 16W and 18W of the power difference respectively, the remaining 45W is attributed to the CPU and idle driver differences.
The lack of intel_idle in TrueNAS SCALE prevents the CPU from entering deeper idle states like C10, which is available in Proxmox and helps save power, especially when multiple cores are involved.

It would be great to see support for intel_idle in TrueNAS SCALE to close this gap, especially for users with similar setups looking to optimize power consumption.

2 Likes