SpeedStep on a Mobile P4-M: Cannot see the lower speed!

[JesseKnows]

Hello,

I recently received a Presario 2100 with Mobile Celeron 2.4 GHz (SStep SL75J). Since I would like to improve processing and battery life, and CPUs are cheap, I swapped the CPU with a 2.0 GHz Mobile P4-M (SStep SL6FK). I expected the performance at full speed to be a little better, and the additional possibility of running at the lower speed (1.2 GHz) for battery maximization.

I did not run benchmarks so I cannot give evidence to the resulting speed at 2.0 GHz. However, I cannot get the system to go to 1.2 GHz for power conservation.

I installed SpeedSwitchXP, it always shows 2 GHz, even when set to Battery Maximum. The Intel CPU utility shows 2.0 GHz (and states that other speeds are possible). The General Tab of Windows System Properties, on the other hand, shows 1.2GHz all the time. The BIOS shows 2.0 GHz.

I am running BIOS KE.M1.71, two versions behind the current KE.M1.73, could that be a reason? The release notes for the BIOS do not say anything about CPU or SpeedStep: .72 is about support for TI Cardbus Controller which my model doesn't have. .73 is about "HP Battery Optimizer cannot be run with some batteries", could that be related to SpeedStep? The system has a Battery Optimizer when pressing F6 at the POST. I hesitate to flash the BIOS (I bricked a Thinkpad a few months ago flashing the BIOS).

Is there some secret to moving from Celeron to P4 w/SpeedStep? Reinstall Windows (running XP SP3 now)? Reset the BIOS somehow?

Thanks

 

[JesseKnows]

Subscribing to topic.

 

[myocardia]

Mobile Celerons used to not have SpeedStep (not that long ago), so I'm assuming that your BIOS doesn't give you the ability to use it, whether or not the CPU is capable. As far as what speed your CPU is running while in Windows, do yourself a favor, and download software meant for that purpose. We all use CPU-Z.

 

[JesseKnows]

I know that Celerons did not support SpeedStep. My initiative was based on the fact that the Presario 2100 was sold with Celeron or a P4 at different price points, and all used the same BIOS. Also, typically a system that doesn't support SpeedStep would boot into the slower CPU multiplier when presented with a SpeedStep CPU (I have an HP OmniBook 900 that behaves like that). Since my system does go to 2.0 GHz, I conclude that SpeedStep is supported. The issue is how to control for the lower speed.

I tried CPU-z as well, it shows 2.0 GHz when running under battery and SpeedSwitchXP set to maximum battery.

Thanks for the reply.

 

[Pabster]

Have you tried Notebook Hardware Control?

 

[JesseKnows]

I tried with Notebook Hardware Control (2.0 pre-release 6 is what I found). I see the same behavior. Regardless of the CPU Speed setting (Max performance, Max battery) I see the 2.0 GHz speed. NHC gives an option for Clock Modulation as a way to take the CPU speed down, separate from SpeedStep's multiplier change. Setting Clock Modulation to 50% has NHC showing the clock going to 1 GHz, but CPU-Z still showed the system running at 100MHZ FSB with 20x multiplier = 2.0 GHz.

I guess with a Watt meter I could check whether the power consumption actually changes when the Clock Modulation is dialed down. But I don't have one

Plus, my ultimate goal is to set the system for dynamic switching, so I would save power when possible and still not incur the performance penalty when speed is helpful. There is no facility for that in NHC, only to set the Clock Modulation statically for AC power and Battery.

This is puzzling, which is why I am posting

 

[JesseKnows]

Anyone?

 

[JesseKnows]

Help, please

 

[PlasmaBomb]

Another tool to try would be RM Clock

Download

Although I wouldn't hold out much hope if nothing else has worked.

It could be that support just isn't implemented fully...

Edit: Yeah this website doesn't have speedstep listed under features...

nor does this

Going to intels site Link

Under Supported Features:
Is blank it should read

Enhanced Intel Speedstep® Technology (like this)

I also skimmed the manual (if all else fails RTFM)
According to the manual the Presario can do the following to lengthen battery life:
Display is turned off
Hard drive is turned off
Standby is initiated
Hibernation is initiated

Lowering the CPU frequency is not an option...

 

[JesseKnows]

Thanks for the pointer to RMClock. Reading the page, I see this:

NOTE: Intel(R) SpeedStep(tm) technology and the "old" vision of Enhanced Intel(R) SpeedStep(tm) technology found in mobile Pentium III-M and mobile Pentium 4 processors are not supported.
Dynamic on-demand clock modulation adjustment is supported on:
Intel Pentium M/Celeron M, Intel Pentium 4/Pentium 4 Extreme Edition/Celeron, Intel Xeon, Intel Pentium D/Pentium Extreme Edition, Intel Core Solo/Core Duo, Intel Core 2 Duo/Quad/Extreme featuring On-Demand Clock Modulation (ODCM).

So SpeedStep would not be manipulated by RMClock, but I can try to get Dynamic on-demand clock modulation adjustment to get me the desired result. Maybe

I guess it is possible to read the omission in Intel's site as evidence to lack of support for SpeedStep. It's very strange though, for a Mobile P4-M to not support SpeekStep but rather always run at the higher multiplier.
I see on cpu-world that the SL6FK is said to support 2.0 GHz and 1.2 GHz. It's only Intel's site that omits the reference.
I can try running the CPU on a desktop system and see what multiplier I get. If the higher, that proof positive that the CPU only runs at the high speed. If the lower, then I am back at square one. If it doesn't even POST, then the motherboard is not compatible with the CPU...
I could be brave and flash the latest BIOS...
At worst, I spent ~$20 to go from Mobile Celeron 2.4 to Mobile P4-M 2.0. Probably not a great improvement there

 

[PlasmaBomb]

I don't think ODCM saves much power (at least I haven't seen much with P4s) but give it a go. If the worst comes to the worst you have spent $20 and had some fun/learned something new

 

[JesseKnows]

Curiouser & curiouser...

I just turned the machine on after a 2-week trip. It had been hibernated. When it came back, Notebook Hardware Control (which was loaded) showed 1.2GHz! CPU-Z agreed that the machine is running 1.2 GHz, 12X multiplier!
To make sure I ran the Intel Processor ID utility. When that came up, it showed 2.0GHz
I said "(&^%(" and Alt-Tabbed to CPU-Z. It showed 2.0 GHz (20X), and NHC also showed 2.0 GHz.
At this point I am back at the original situation, no way to select 12X multiplier again.

So, somehow the machine is capable of getting to 12X, but I do not seem to be able to control it.

PC Wizard 2008 states that the CPU supports Low/High multiplier of 12x/20x. But, the "Features" section says that EST (Enhanced SpeedStep Technology) is "no".
Intel Processor ID shows Enhanced SpeedStep Technology "yes".

Update:
Installed some Windows updates and rebooted. SpeedSwitchXP reported 1.2GHz again, with the system is set to Dynamic Switching and the CPU pegged at 100% according to Task Manager (probably running the boot-time AV scan). Asking for Max Performance did not change the speed to 2.0 GHz.
Ran the Intel Processor ID utility, and it's also showed 1.2 GHz.
PC Wizard is also showed 1.2 GHz.
After a while the CPU was now down to under 10% (boot scan over) and suddenly the system switched to 2.0 GHz!

I think a BIOS update is in order.

 

[BlueAcolyte]

You can't control speedstep... When it works right, it will change depending on processor load.

 

[JesseKnows]

So...

I updated the BIOS to the latest. The BIOS shows Mobile P4-m 2.0 GHz as expected (I believe it showed that before the update as well). This suggests that the BIOS recognizes the CPU correctly and should support the CPUs capabilities. Unlikely that my unit than came with Mobile Celeron somehow does support Mobile P4-m.

The system again booted at 1.2 GHz, and switched to 2.0 GHz soon after boot was completed (based on SpeedSwitchXP's display). First came up as 1.2 GHz and showed Maximum 1.2 GHz. Then switched to 2.0 GHz, and actually still showing Maximum 1.2 GHz. Clicking Detect max. speed then caused Maximum to update to 2.0 GHz.

Tried another reboot, with SpeedSwitchXP graphing the speed. I see 1.2 GHz right after boot, then going to 2.0 GHz soon thereafter.

I'll try again with SSXP set to Maximize Battery when running from DC, and boot without the AC power. That's really where I need the power saving.

 

[JesseKnows]

Originally posted by: BlueAcolyte
You can't control speedstep... When it works right, it will change depending on processor load.

So what's SpeedSwitchXP all about? I have used it on a ThinkPad T30 with Mobile P4-m, and I see the speed changing when I change between Max Battery and Max Performance.

Also, the system is now running at 2.0 GHz with no load.

 

[BlueAcolyte]

Well I suppose that switches speed manually to save power. I dunno, I'm using it on a desktop system.

 

[JesseKnows]

You have a desktop with Speedstep? Using a P4?

 

[myocardia]

Every Core 2 Duo and quad support SpeedStep, as well as the last generation (maybe last two) of P4's.

 

[JesseKnows]

I tried running on DC.

First, running at 2.0GHz, hibernating, then waking from hibernation, the system stayed at 2.0 GHz.

Next, shutdown and restart with SpeedSwitchXP set to Dynamic Switching. The system first booted at 1.2 GHz, with the CPU running at 100% while the boot was completing. About 10 seconds after the CPU was close to idle, SSXP showed the speed going down to 400 MHz, then up to 2.0 GHz, and after a couple of short-span transitions (one speed reading by SSXP) settled at 2.0GHz where it is now. System idle and 2.0GHz speed.

I believe the situation is still as it was when I started the thread: the system does not lower the multiplier when idle or when set to Maximum battery.
We do have additional information, in that the system actually can run at 1.2 GHz, and does so for a short time after boot.

 

[JesseKnows]

I tried using RM Clock. As mentioned above, it doesn't deal with the multiplier. It does perform Dynamic on-demand clock modulation adjustment, and that seems to be doing The Right Thing. I do not have enough data to say whether the clock adjustment actually has impact on the power consumption.