Issue with Intel SpeedStep settings

[MacUsers]

Hi there,

I up/downgraded my pfSense box from a Xeon SuperMicro to a Dell i3-540 R210, mainly to minimize the power consumption. Since then I'm seeing this during the boot:

CPU supports enhanced SpeedStep, but it's not recognized

What does it telling me? Some of the Google searches suggest that FreeBSD has issues recognizing SpeedStep for some Intel CPUs. How do I fix that?

-Santanu

[MacUsers]

any/no one having this issue?
Not sure, if i can call it an "issue" but nice to have a clean startup.

-S

[show-p1984]

I haven't seen that on my Xeon, however I get this on bootup (Supermicro X8SIL-F w/ x3470):

est0: <enhanced speedstep="" frequency="" control="">on cpu0
est1: <enhanced speedstep="" frequency="" control="">on cpu1
est2: <enhanced speedstep="" frequency="" control="">on cpu2
est3: <enhanced speedstep="" frequency="" control="">on cpu3
est4: <enhanced speedstep="" frequency="" control="">on cpu4
est5: <enhanced speedstep="" frequency="" control="">on cpu5
est6: <enhanced speedstep="" frequency="" control="">on cpu6
est7: <enhanced speedstep="" frequency="" control="">on cpu7</enhanced></enhanced></enhanced></enhanced></enhanced></enhanced></enhanced></enhanced> 

So I assumed that it worked, but sysctl reports max freq + 1 Mhz (== Turbo boost) permanently and doesn't seem to lower the clock in idle.
PowerD is enabled and set to Hiadaptive.

EDIT:
Just checked IPMI, the sensors show vcore adjustments. This means Speedstep should work_s_ as intended. Is this a bug in sysctl?

EDIT:
Works as intended for me. Hiadaptive was just too aggressive. Adaptive shows scaling:

sysctl dev.cpu. | grep freq
dev.cpu.0.freq_levels: 2934/106000 2933/95000 2800/82000 2667/70000 2533/62000 2400/53000 2267/46000 2133/39000 2000/33000 1867/28000 1733/24000 1600/20000 1467/17000 1333/14000 1200/11000
dev.cpu.0.freq: 1333

[A Former User]

Can you enable power management in the BIOS?
Also this will show if speedstep is recognized:

grep -i speedstep /var/run/dmesg.boot

[MacUsers]

Power management is set to "Active Power Controller", which

• Enable Dell System DBPM (BIOS will not make all P states available to OS)

• Memory frequency = Maximum Performance

• Fan algorithm = Minimum Power

This is what I see in the dmesg:

[2.3.3-RELEASE][admin@pfs17.littlehome.co.uk]/root: grep -i speedstep /var/run/dmesg.boot
est0: <enhanced speedstep="" frequency="" control=""> on cpu0
est: CPU supports Enhanced Speedstep, but is not recognized.
est1: <enhanced speedstep="" frequency="" control=""> on cpu1
est: CPU supports Enhanced Speedstep, but is not recognized.
est2: <enhanced speedstep="" frequency="" control=""> on cpu2
est: CPU supports Enhanced Speedstep, but is not recognized.
est3: <enhanced speedstep="" frequency="" control=""> on cpu3
est: CPU supports Enhanced Speedstep, but is not recognized.</enhanced></enhanced></enhanced></enhanced>

[A Former User]

The Maximum Performance may be stopping the CPU's from any energy efficiency.
Is there any other option available other than Max Perf.

[Sergei_Shablovsky]

@macusers

This is common (and VERY OLD) well known FreeBSD pesky problem with powerd and est drivers correct work in Intels cpus.

The solution are described here https://www.ateamsystems.com/tech-blog/increase-freebsd-performance-with-powerd/

Because the pfSense are

• build on top of FreeBSD components (so some NIC and hardware drivers able to using several cores/several CPUs, some - no https://forum.netgate.com/topic/149735/how-pfsense-utilize-multicore-processors-and-multi-cpu-systems)
• FreeBSD have some issues for a years (!) (especially with drivers);

In case of using multi-cpu and multi-core hardware some issues come back :)

[stephenw10]

When you see that in the boot log, for example:

est0: <Enhanced SpeedStep Frequency Control> on cpu0
est: CPU supports Enhanced Speedstep, but is not recognized.
est: cpu_vendor GenuineIntel, msr 233500002000
device_attach: est0 attach returned 6

It's almost always because Speedstep is disabled, and often locked, in the BIOS. (The CPU MSRs).

Steve

[Sergei_Shablovsky]

@stephenw10 said in Inssue with Intel SpeedStep settings:

When you see that in the boot log, for example:

est0: <Enhanced SpeedStep Frequency Control> on cpu0
est: CPU supports Enhanced Speedstep, but is not recognized.
est: cpu_vendor GenuineIntel, msr 233500002000
device_attach: est0 attach returned 6

It's almost always because Speedstep is disabled, and often locked, in the BIOS. (The CPU MSRs).

Steve

Thank You, Steve!

Are You reading the link https://www.ateamsystems.com/tech-blog/increase-freebsd-performance-with-powerd/ ?

There are very interesting dependency between SpeedStep and TurboBoost... Isn’t it?

[provels]

Gotta love this!
login-to-view
I have an ancient (like 2004) P4 server that I run Xigmanas on and get the same message.
I believe it's because the BIOS has no idea WTF the CPU is talking about.

[Sergei_Shablovsky]

@provels said in Issue with Intel SpeedStep settings:

Gotta love this!
login-to-view

If You Googling that realize that this problem come again and again even in new machines.

I have an ancient (like 2004) P4 server that I run Xigmanas on and get the same message.

If I able to buy 2012-2014 year produced IBM branded ServerX series 2 CPU 48Gb ECC RAM 2 hot spare PSU 900W each w/ separate mgmt card for $300 and with good Intel 40Gb card he able to deal with 70-80Gb of traffic (with ssl, vpn, balancing, blocking, etc full stack) why I need to spending more? For what reason ?
By the way IBM Server X is from “non-Lenovo” era, when IBM have responsibilities for EACH server.

So in particular this case “old” not equal “trash”. :)

I believe it's because the BIOS has no idea WTF the CPU is talking about.

Please read articles that I pointed above.

[provels]

@sergei_shablovsky
FWIW, my XigmaNAS box that reports "est: CPU supports Enhanced Speedstep, but is not recognized." , still downclocks as expected with PowerD enabled. So there's that...

[stephenw10]

You're probably seeing acpi throttling or p4tcc rather than true CPU frequency changing. I never found either did much.

Steve

[provels]

@stephenw10
Possibly, the box is 2004 (not a typo). But you're right, it makes no difference in operation. The box sucks 100w at idle or full load (P4, 3.60GHz).

[stephenw10]

Yup those later P4s were hungry hungry beasts!