Skip to content
View in the app

A better way to browse. Learn more.

Unraid

A full-screen app on your home screen with push notifications, badges and more.

To install this app on iOS and iPadOS
  1. Tap the Share icon in Safari
  2. Scroll the menu and tap Add to Home Screen.
  3. Tap Add in the top-right corner.
To install this app on Android
  1. Tap the 3-dot menu (⋮) in the top-right corner of the browser.
  2. Tap Add to Home screen or Install app.
  3. Confirm by tapping Install.

Intel 15th Gen 'Arrow Lake' Thoughts?

Featured Replies

H810 unfortunately has a very low amount of lanes. I did compile powertop from source for the arrow lake support but still only max of C2 was possible.

Not sure if the chiplet design really is the issue. I tried a meteor lake NUC some time ago and it could successfully reach C10 with an extremely low idle consumption of 2.4W. It seems that without support in intel_idle the ACPI tables are read from BIOS so maybe ACPI tables are wrong?

  • Replies 170
  • Views 66.4k
  • Created
  • Last Reply

Top Posters In This Topic

Most Popular Posts

  • I have my Ultra 7 265K iGPU working with the Unraid 7 release with a custom kernel 6.12.9.     It transcodes with my Jellyfin docker on stable, but need the beta version of Emby.

  • To whom it may concern: With the just released beta 7.1 (which runs on Kernel 6.12.19) the iGPU now works flawlessly for Arrow Lake CPUs (in my case a 245K) HW Transcoding via Plex as i write this

  • Just received my Motherboard (ASUS Z890M Plus WiFi) and been doing some quick and dirty testing to gauge the best settings with regard to performance/watt for the Intel 245K.    Based on the

Posted Images

powertop will not report correct c-states until intel_idle is fixed. File a bug with intel and see if you can get them to care about their own products. Intel explained why raptor lake didn't get support initially; they found no real improvement over ACPI. If you hit C3_ACPI with greater than 85%, you are getting C8 at least.

BGA vs LGA.

LGA Ultra 7 265T

https://www.energystar.gov/productfinder/product/certified-computers/details/3999018

3.3w idle (AIO screen off)

LGA Ultra 7 265T

https://www.energystar.gov/productfinder/product/certified-computers/details/3892093

2.1w idle

The issue is intel is chasing gaming benchmarks on PC desktops and dragging those that want low idle with them.

The ASRock board had similar results. Stuck in C2.

I'm still not convinced that intel_idle is at fault. As I explained there is no specific code for raptor lake in the kernel driver but raptor lake can still hit C10.

grep . /sys/devices/system/cpu/cpu0/cpuidle/state*/name

/sys/devices/system/cpu/cpu0/cpuidle/state0/name:POLL

/sys/devices/system/cpu/cpu0/cpuidle/state1/name:C1_ACPI (C1)

/sys/devices/system/cpu/cpu0/cpuidle/state2/name:C2_ACPI (C7)

/sys/devices/system/cpu/cpu0/cpuidle/state3/name:C3_ACPI (C10)

grep . /sys/devices/system/cpu/cpu0/cpuidle/state*/desc

/sys/devices/system/cpu/cpu0/cpuidle/state0/desc:CPUIDLE CORE POLL IDLE

/sys/devices/system/cpu/cpu0/cpuidle/state1/desc:ACPI FFH MWAIT 0x0

/sys/devices/system/cpu/cpu0/cpuidle/state2/desc:ACPI FFH MWAIT 0x21

/sys/devices/system/cpu/cpu0/cpuidle/state3/desc:ACPI FFH MWAIT 0x60

All of the c-states are just condensed into 3.


cat /sys/devices/system/cpu/cpuidle/current_driver

acpi_idle

I laughed out loud here. Switching to acpi_idle is a 2-3w drop over intel_idle. I'm seeing dips down to 16.5w.

  • 4 weeks later...

Just a short update on my X14SAE-F board - it finally arrived. Migration was straight forward.
The only issues I have, is with the Fan control and to get the IPMI support tool to control the Fans. The build in fan profiles from supermicro struggle all (except full speed) to keep the fans running, or at least from time to time the IPMI report 0 rpm.
I've played around with ipmitool to change the thresholds but that did not help so far.

Besides that, I'm quite happy with the performance... I'll write an update after some more testing.
Just one remark, even that it took ages to get the board - the communication and support via Serverparts.pl where I got the board from was perfect, and they did everything possible to get the board from supermicro.

  • 3 weeks later...

https://github.com/amitxv/SCEWIN

Has anyone been able to get SCEWIN working on arrow lake? It complains about missing a driver or it just does a soft lock when trying to read values.

  • 1 month later...
On 7/9/2025 at 1:23 AM, knotgerman said:

grep . /sys/devices/system/cpu/cpu0/cpuidle/state*/name

/sys/devices/system/cpu/cpu0/cpuidle/state0/name:POLL

/sys/devices/system/cpu/cpu0/cpuidle/state1/name:C1_ACPI (C1)

/sys/devices/system/cpu/cpu0/cpuidle/state2/name:C2_ACPI (C7)

/sys/devices/system/cpu/cpu0/cpuidle/state3/name:C3_ACPI (C10)

grep . /sys/devices/system/cpu/cpu0/cpuidle/state*/desc

/sys/devices/system/cpu/cpu0/cpuidle/state0/desc:CPUIDLE CORE POLL IDLE

/sys/devices/system/cpu/cpu0/cpuidle/state1/desc:ACPI FFH MWAIT 0x0

/sys/devices/system/cpu/cpu0/cpuidle/state2/desc:ACPI FFH MWAIT 0x21

/sys/devices/system/cpu/cpu0/cpuidle/state3/desc:ACPI FFH MWAIT 0x60

All of the c-states are just condensed into 3.


cat /sys/devices/system/cpu/cpuidle/current_driver

acpi_idle

I laughed out loud here. Switching to acpi_idle is a 2-3w drop over intel_idle. I'm seeing dips down to 16.5w.

Hi, how did you accually managed to switch to acpi_idle from intel_idle ? I dont know if the question is kind of silly, but i did not find out how tto do it.

On 8/23/2025 at 8:30 PM, knotgerman said:

https://github.com/amitxv/SCEWIN

Has anyone been able to get SCEWIN working on arrow lake? It complains about missing a driver or it just does a soft lock when trying to read values.

I did go down the rabbit of trying to set the "Low Power S0 Idle Capability". I never got the Linux version von SCELNX to work, But i was able to boot into windwos to got SCEWIN to work an could export the NVRAM but there was no "Low Power S0 Idle Capability" listed. So i was unable to set that setting.

Edited by buisiham

8 hours ago, buisiham said:

Hi, how did you accually managed to switch to acpi_idle from intel_idle ? I dont know if the question is kind of silly, but i did not find out how tto do it.

Okay i found out how to do it. To enforce aspc_idle you have to add the kernel boot argument "intel_idle.max_cstate=0". This resulted for me in a 2w drop over intel_idle.

Im also having the C2 Package issue with a 245k on a ASRock Z890. Interesting to hear that other mainboard vendors are having the same issue.

Ive tested on Ubuntu and Windows, so it seems to be a generic issue. CPU c states work fine, it's just the package state wich is stuck at C2. Someone reported on ASRock forums, that he got package states larger than C2 on Windows but not on Linux with a beta FW. I was not able to reproduce.

Overall im at 22W on a barebone system with onboard LAN - not bad, not great either. I hope this gets fixed with some bios/microcode update at some point. Or someone can report at a working configuration.

On 10/15/2025 at 4:29 AM, buisiham said:

I did go down the rabbit of trying to set the "Low Power S0 Idle Capability". I never got the Linux version von SCELNX to work, But i was able to boot into windwos to got SCEWIN to work an could export the NVRAM but there was no "Low Power S0 Idle Capability" listed. So i was unable to set that setting.

$ apt install acpica-tools
$ acpidump -b -n FACP
$ iasl -d facp.dat

Use the above to pull the values for your board and convert to text. All desktop boards should have that setting but it is off for some reason.

On 10/18/2025 at 3:24 PM, JanTenner said:

Im also having the C2 Package issue with a 245k on a ASRock Z890. Interesting to hear that other mainboard vendors are having the same issue.

Ive tested on Ubuntu and Windows, so it seems to be a generic issue. CPU c states work fine, it's just the package state wich is stuck at C2. Someone reported on ASRock forums, that he got package states larger than C2 on Windows but not on Linux with a beta FW. I was not able to reproduce.

Overall im at 22W on a barebone system with onboard LAN - not bad, not great either. I hope this gets fixed with some bios/microcode update at some point. Or someone can report at a working configuration.

C-states work. intel_idle is busted which means powertop is busted. The best you can do is 90-100% C3_ACPI i.e. you hit C10. This is across all arrow lake boards.

Every single bios update has raised power consumption. Roll your bios back through each microcode and watch idle power drop. For me, anything beyond 0x116 is unacceptable. Sounds like you are on the latest.

I believe the absolute floor for (at the wall) idle power is 10w for desktop arrow lake. I gave up at 16w and moved on.

On 10/15/2025 at 12:32 PM, buisiham said:

Okay i found out how to do it. To enforce aspc_idle you have to add the kernel boot argument "intel_idle.max_cstate=0". This resulted for me in a 2w drop over intel_idle.

I reverted back to intel_idle as there does not seem to be a discernible difference over time. I think hybrid core handing is not optimized.

6.15 - SLPC low power GuC mode

6.16 - EAS support (didn't do much for me)

6.19 - ACPI_idle (and intel_idle) should be improved:

https://www.phoronix.com/news/Linux-ACPI-CPU-Idle-Hybrid

Hello!

I did also a lot of testing on Asus Pro WS W880-ACE SE.

So you mean powertop (built from source) and also turbostat is wrong about the package cstates?

I also tested on windows using throttlestop and also could only observe package c2 states, so it would match powertop and turbostat.

3 hours ago, lois said:

Hello!

I did also a lot of testing on Asus Pro WS W880-ACE SE.

So you mean powertop (built from source) and also turbostat is wrong about the package cstates?

I also tested on windows using throttlestop and also could only observe package c2 states, so it would match powertop and turbostat.

No. The correct c-state reporting won't work until intel fixes intel_idle. There might be fixes for kernel 6.19 or whatever number comes after that.

https://forums.unraid.net/topic/176866-intel-15th-gen-arrow-lake-thoughts/page/6/#findComment-1566181

So you mean that on intel 15th gen there is only one package c-state?

Because the info from powertop and turbostat matches my windows findings with throttlestop.
The idle power (package W from hwinfo) matches the reading from turbostat. And also power from the wall is identical.
I guess Linux pstate handling has nothing in common with Windows but shows the same results.
image.png

Im in the same boat as lois. Tested on windows with throttelstop and on linux with turbostat. Never reach higher than PKgC2.

My theory is that intel_pstate would expose hardware states via sysfs, but it is relying on ACPI for 15th gen CPUs so no extra package states exposed. I would consider it as as interface tool. Other tools like turbostat query hardware states too. They show package states.

knotgerman, you seem to be very much into the topic. Please correct me if Im understanding this wrong.

It wonder what mobile variants e.g. Core Ultra 235T would do. Or is it a differently configured motherboard/bios. Can't imagine that those have the same power draw.

Here is the output of decoding ACPI tables,

The line Low Power S0 Idle (V5) : 0 stands out.

/*
 * Intel ACPI Component Architecture
 * AML/ASL+ Disassembler version 20250404 (64-bit version)
 * Copyright (c) 2000 - 2025 Intel Corporation
 * 
 * Disassembly of facp.dat
 *
 * ACPI Data Table [FACP]
 *
 * Format: [HexOffset DecimalOffset ByteLength]  FieldName : FieldValue (in hex)
 */

[000h 0000 004h]                   Signature : "FACP"    [Fixed ACPI Description Table (FADT)]
[004h 0004 004h]                Table Length : 00000114
[008h 0008 001h]                    Revision : 06
[009h 0009 001h]                    Checksum : 64
[00Ah 0010 006h]                      Oem ID : "ALASKA"
[010h 0016 008h]                Oem Table ID : "A M I "
[018h 0024 004h]                Oem Revision : 01072009
[01Ch 0028 004h]             Asl Compiler ID : "AMI "
[020h 0032 004h]       Asl Compiler Revision : 01000013

[024h 0036 004h]                FACS Address : 76487000
[028h 0040 004h]                DSDT Address : 762E8000
[02Ch 0044 001h]                       Model : 00
[02Dh 0045 001h]                  PM Profile : 01 [Desktop]
[02Eh 0046 002h]               SCI Interrupt : 0009
[030h 0048 004h]            SMI Command Port : 000000B2
[034h 0052 001h]           ACPI Enable Value : 00
[035h 0053 001h]          ACPI Disable Value : 00
[036h 0054 001h]              S4BIOS Command : 00
[037h 0055 001h]             P-State Control : 00
[038h 0056 004h]    PM1A Event Block Address : 00001800
[03Ch 0060 004h]    PM1B Event Block Address : 00000000
[040h 0064 004h]  PM1A Control Block Address : 00001804
[044h 0068 004h]  PM1B Control Block Address : 00000000
[048h 0072 004h]   PM2 Control Block Address : 00001850
[04Ch 0076 004h]      PM Timer Block Address : 00001808
[050h 0080 004h]          GPE0 Block Address : 00001860
[054h 0084 004h]          GPE1 Block Address : 00001810
[058h 0088 001h]      PM1 Event Block Length : 04
[059h 0089 001h]    PM1 Control Block Length : 02
[05Ah 0090 001h]    PM2 Control Block Length : 01
[05Bh 0091 001h]       PM Timer Block Length : 04
[05Ch 0092 001h]           GPE0 Block Length : 20
[05Dh 0093 001h]           GPE1 Block Length : 18
[05Eh 0094 001h]            GPE1 Base Offset : 80
[05Fh 0095 001h]                _CST Support : 00
[060h 0096 002h]                  C2 Latency : 0065
[062h 0098 002h]                  C3 Latency : 03E9
[064h 0100 002h]              CPU Cache Size : 0400
[066h 0102 002h]          Cache Flush Stride : 0010
[068h 0104 001h]           Duty Cycle Offset : 00
[069h 0105 001h]            Duty Cycle Width : 00
[06Ah 0106 001h]         RTC Day Alarm Index : 0D
[06Bh 0107 001h]       RTC Month Alarm Index : 00
[06Ch 0108 001h]           RTC Century Index : 32
[06Dh 0109 002h]  Boot Flags (decoded below) : 0000
               Legacy Devices Supported (V2) : 0
            8042 Present on ports 60/64 (V2) : 0
                        VGA Not Present (V4) : 0
                      MSI Not Supported (V4) : 0
                PCIe ASPM Not Supported (V4) : 0
                   CMOS RTC Not Present (V5) : 0
[06Fh 0111 001h]                    Reserved : 00
[070h 0112 004h]       Flags (decoded below) : 0003C6E5
      WBINVD instruction is operational (V1) : 1
              WBINVD flushes all caches (V1) : 0
                    All CPUs support C1 (V1) : 1
                  C2 works on MP system (V1) : 0
            Control Method Power Button (V1) : 0
            Control Method Sleep Button (V1) : 1
        RTC wake not in fixed reg space (V1) : 1
            RTC can wake system from S4 (V1) : 1
                        32-bit PM Timer (V1) : 0
                      Docking Supported (V1) : 1
               Reset Register Supported (V2) : 1
                            Sealed Case (V3) : 0
                    Headless - No Video (V3) : 0
        Use native instr after SLP_TYPx (V3) : 0
              PCIEXP_WAK Bits Supported (V4) : 1
                     Use Platform Timer (V4) : 1
               RTC_STS valid on S4 wake (V4) : 1
                Remote Power-on capable (V4) : 1
                 Use APIC Cluster Model (V4) : 0
     Use APIC Physical Destination Mode (V4) : 0
                       Hardware Reduced (V5) : 0
                      Low Power S0 Idle (V5) : 0
S0ix	Modern Standby,[34] or "Low Power S0 Idle". Partial processor SoC sleep.[35][36] Sub states include S0i1, S0i2 and S0i3. Known to ARM and x86 devices.

Don't know if a sleep state influences package states.
https://en.wikipedia.org/wiki/ACPI

10 hours ago, lois said:

So you mean that on intel 15th gen there is only one package c-state?

Because the info from powertop and turbostat matches my windows findings with throttlestop.
The idle power (package W from hwinfo) matches the reading from turbostat. And also power from the wall is identical.
I guess Linux pstate handling has nothing in common with Windows but shows the same results.
image.png

Go read the link and what I actually wrote: "All of the c-states are just condensed into 3."

7 hours ago, JanTenner said:

Im in the same boat as lois. Tested on windows with throttelstop and on linux with turbostat. Never reach higher than PKgC2.

My theory is that intel_pstate would expose hardware states via sysfs, but it is relying on ACPI for 15th gen CPUs so no extra package states exposed. I would consider it as as interface tool. Other tools like turbostat query hardware states too. They show package states.

knotgerman, you seem to be very much into the topic. Please correct me if Im understanding this wrong.

It wonder what mobile variants e.g. Core Ultra 235T would do. Or is it a differently configured motherboard/bios. Can't imagine that those have the same power draw.

Same thing happened for a short while for 13th gen. The c-states work; the reporting doesn't because Intel does not care to fix intel_idle and instead use acpi_idle.

https://www.energystar.gov/productfinder/product/certified-computers/details/3999745

Workstations: Long Idle (watts) Field details: 6.6w
Workstations: Short Idle (watts) Field details: 44.5w

https://www.energystar.gov/productfinder/product/certified-computers/details/3798818

Category I2: Long Idle (watts) Field details: 2.2w

Category I2: Short Idle (watts) Field details: 22.6w

https://www.energystar.gov/productfinder/product/certified-computers/details/3892312

Category D2: Long Idle (watts) Field details: 3.7w

Category D2: Short Idle (watts) Field details: 22.6

OEM's have 12VO psu's and likely S0ix enabled. The best I can do is 12w under the same "Long Idle" conditions i.e. let your desktop idle with monitor blanked.

5 hours ago, lois said:
S0ix	Modern Standby,[34] or "Low Power S0 Idle". Partial processor SoC sleep.[35][36] Sub states include S0i1, S0i2 and S0i3. Known to ARM and x86 devices.

Don't know if a sleep state influences package states.
https://en.wikipedia.org/wiki/ACPI

Once S0ix is enabled, it should drop most arrow lake desktops down to 10w. The problem is turning it on.

  1. Asrock is the only OEM I know of that you can beg for beta BIOS with stuff turned on. Beg them to provide you with beta bios and it turned on.

  2. Find a version of SCEWIN that actually works to turn it on.

  3. Manually edit efi variables and risk bricking your MB

Don't have #1, #2 didn't work, and I'm not risking my MB to attempt #3.

1) I tried a ASRock beta firmware with a supposed fix on Ubuntu/Win11, but no success.
BTW: Just saw that Lois asked for an updated status in the ASRock forum thread.

2) Tried a few SCELNX, the issue is that the relevant option is not read. Seemed to have worked on LGA1700 boards.


If its Intels fault, why there some PCs which have low idle (e..g the energystar links). Also hard to believe that windows Arrow Lake laptops have the same issue.

Doe really all major Z890 motherboard vendors not optimise or test this? There is ASRock, Asus, and I think also Gigabyte, MSI (no sure these two). Also not only Z890.

Im just baffled that all those vendors seem to miss this. Linux and Windows. Arrow lake/LGA 1851 is not that new.

1 hour ago, JanTenner said:

If its Intels fault, why there some PCs which have low idle (e..g the energystar links).

OEM's have S0ix enabled for such low "Long Idle" + 12VO. "Short Idle" = intel's bad microcode chasing benchmarks. 22W desktop idle for OEM's is IMHO bad.

Dell USFF desktop:

https://www.energystar.gov/productfinder/product/certified-computers/details/3892093

300T (13th gen)

Category I1: Long Idle (watts) Field details: 2.2

Category I1: Short Idle (watts) Field details: 6.4

265T (15th gen/ARL desktop)

Category I2: Long Idle (watts) Field details: 2.1

Category I2: Short Idle (watts) Field details: 5.8

Like I said, "Low Power S0 Idle" enabled should get all arrow lake boards down to ~10W.

I was able to enable "Low Power S0 Idle Capability" via WINSCE.

Could not observe any improvements.
cat /sys/kernel/debug/pmc_core/package_cstate_show

Package C2 : 67638118

Package C3 : 0

Package C6 : 0

Package C7 : 0

Package C8 : 0

Package C9 : 0

Package C10 : 0


GitHub
No image preview

GitHub - ab3lkaizen/SCEHUB: SCEWIN binaries and solutions...

SCEWIN binaries and solutions for various error messages - ab3lkaizen/SCEHUB

Then i used the 6.18-rc2 kernel on Ubuntu and found /sys/kernel/debug/pmc_core/substate_requirements
https://github.com/gregkh/linux/commit/a22bc8643889ef5a3c032700f69001fdc457413c

Edited by lois

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

Account

Navigation

Search

Search

Configure browser push notifications

Chrome (Android)
  1. Tap the lock icon next to the address bar.
  2. Tap Permissions → Notifications.
  3. Adjust your preference.
Chrome (Desktop)
  1. Click the padlock icon in the address bar.
  2. Select Site settings.
  3. Find Notifications and adjust your preference.