Sign in with
Sign up | Sign in
Your question

Pentium M Centrino switching frequency

Last response: in CPUs
Share
March 8, 2007 12:54:29 PM

Hi all,
I have an HP laptop (dv4275EA) with a pentium M centrino 1.7GHz. I noticed recently that the CPU changes its frequency from time to time (I know it's normal), so I installed speedswitchXp to have it run at full speed.
However, the CPU still changes its frequency when under heavy load (e.g. games), it switches from 1.7 GHz to 900MHz, but the multiplier does not change (x13). Typically, it would run at 1.7GHz for about 5-10 mins, then gets down to 900MHz for about 3-5 mins, then gets back to 1.7, etc...
So I wanted to know if such a behavior is normal. Does the cpu lower its frequency in response to a thermal event ? Then gets back to full speed when it has cooled down ? Is there a way to have it run at 1.7GHz no matter what, and is it dangerous to do so ?
Thxs in advance,
N_D
March 8, 2007 1:01:05 PM

I don't know what that app does, but if you have any option in the BIOS to disable power efficiency, than it is more likely to be effective.
Also, check the CPU temperatures to determinate if the CPU is not overheated and being throttled. The best is to use Intel TAT, which shows the digital temperature of your CPU.
March 8, 2007 1:07:26 PM

There are no options at all in the BIOS (which I've just upgraded from HP's website) concerning the CPU.
I'll check Intel TAT to see if the CPU is overheated.
So in this case, is it normal that it lowers its frequency ?
Related resources
March 8, 2007 1:15:21 PM

I've tried that already, set it to all settings with no changes. Now it is set to speedswitchXP control, so as to enable full speed.
March 8, 2007 1:29:52 PM

Quote:
I've tried that already, set it to all settings with no changes. Now it is set to speedswitchXP control, so as to enable full speed.


Go into your power options inside windows and go to the power scheme. Change it to desktop (it should be on laptop originally) and that should take care of it. What you can't change in the BIOS.... you might be able to get around inside Windows 8O

Just my thoughts.
March 8, 2007 1:34:02 PM

I've already tried that, doesn't work. It enables the multiplier to always remain at 13x though. But the frequency still gets down to 900MHz from time to time. When it was set to "laptop" the multiplier would get to 7x (if I remember correctly) and the freq. to 400MHz.
N_D
March 8, 2007 1:39:17 PM

Its strange...if the frequency goes down, it does by reducing the multiplier...
How have you figured out that the frequency is dropping to 900?
What util have you used?
March 8, 2007 1:43:06 PM

That is really weird, I almost don't believe it. If the multi is locked the frequency should be locked. The only other way is to change the FSB which would dock the memory and kill overall system performance. I am confused by it all...

As gojdo said, what util are you using? Have you booted up Intel TAT to view frequency or CPU-z or what?
March 8, 2007 2:16:35 PM

I used cpu-Z and speedswitchXP. Both utilities show 900MHz when the CPU is in its "low phase". They show 1700MHz otherwise. But cpu-Z shows that the multiplier remains at 13x.
When in a low phase, system performance is killed indeed. In fact, I figured it out when playing a game. The FPS would drop from 120fps to 60fps.
If I remember correctly, the freq. of the bus is lowered also when in the low phase. I don't quite remember, I'm at work right now and not using my laptop.
I updated:
* graphics drivers (ati mobility radeon x700) - clean update (as in tweakguide.com).
* sound drivers
* intel chipset
* flahed bios
* essential windows XP software update from HP

Anyways, thxs for your answers so far guys, I still hope I can solve this issue. Apparently, it's not a normal behavior...
N_D
March 8, 2007 2:35:46 PM

Quote:
I used cpu-Z and speedswitchXP. Both utilities show 900MHz when the CPU is in its "low phase". They show 1700MHz otherwise. But cpu-Z shows that the multiplier remains at 13x.
When in a low phase, system performance is killed indeed. In fact, I figured it out when playing a game. The FPS would drop from 120fps to 60fps.
If I remember correctly, the freq. of the bus is lowered also when in the low phase. I don't quite remember, I'm at work right now and not using my laptop.
I updated:
* graphics drivers (ati mobility radeon x700) - clean update (as in tweakguide.com).
* sound drivers
* intel chipset
* flahed bios
* essential windows XP software update from HP

Anyways, thxs for your answers so far guys, I still hope I can solve this issue. Apparently, it's not a normal behavior...
N_D


I'll see if I can't look into this speedswitchxp util.

No it's not normal, let's figure out why it does this. 8)

The bus speed really shouldn't change (as you know)... This really makes me nervous, because I have *never* seen a laptop alter the bus speed to hit a lower power consumption state.

My *intial* hunch is that since you locked the multi (which is what speedswitchxp does according to you) that the system still feels the need to drop down to 900MHz, so the only way it can is through lowering the FSB.

This is perplexing....

As I have free time today I'll do some searching.
March 8, 2007 2:48:39 PM

Quote:
Does the cpu lower its frequency in response to a thermal event ? Then gets back to full speed when it has cooled down ? Is there a way to have it run at 1.7GHz no matter what, and is it dangerous to do so ?


Good question. The short answer is that your CPU is effected by thermal events. The most important one is the "catastrophic thermal shutdown," which prevents your unit to be damaged by excessive heat. I would wager that you are being effected by TM2, which is BIOS-controlled. TM2 will change your bus clock /core clock multiplier, as well as the VID to your VRM. TM1 basically just changes your clock duty cycle. Legacy thermal management also exists through the PROCHOT pin mechanism.

What is your CPU temperature when you play this game? People in this forum will tell you to disable SpeedStep. It really doesn't effect performance, just saves you money on your electric bill each month. It's similar to turning off the lights in a room when you leave it. But I wouldn't recommend disabling thermal management in BIOS. The choice is so simple, even a caveman could do it.
March 8, 2007 2:52:38 PM

Thxs mate! I'll update the thread with more information (I'll post exactly what cpu-z and speedswitchXP shows in both phases) as soon as I get back home.
By the way, I'm new to this forum (but I do read some articles at tomshardware.com regularly), and I'm impressed with its reactivity!
N_D
March 8, 2007 2:58:12 PM

Usually, legit questions get good responses. It is the offbeat totally useless threads that piss people off.

This forum is for help, so alot of veterans really do try and help. I am not the best by any stretch, but I can field a fair amount of questions. :D 

Back to doing nothing at work..... slow day :?
a c 99 à CPUs
March 8, 2007 2:59:47 PM

The multiplier of any CPU that can adjust its core speed dynamically (e.g. chips with SpeedStep, Cool 'n Quiet) are not completely locked as the multiplier can be reduced to yield the lower speeds. It is possible to reduce core speeds by keeping the multiplier locked and twiddling with the FSB or LDT clock, but this is generally not done as changing the FSB speed also usually changes things like the AGP/PCIe/PCI bus speeds, among other things. Many desktop enthusiast boards have "locked" AGP/PCIe/PCI slots that have a different clock generator than the FSB uses, but that setup is much more complex than leaving the FSB alone and twiddling with the CPU's multiplier.

The actual speeds that your CPU can assume while running are based on the multipliers that the CPU has available and its bus speed. There is generally an idle frequency that it the same for all chips of a particular line, such as the 1.20 GHz idle speed for Pentium 4-M CPUs and the 1.0 GHz idle speed for Athlon 64/Opteron dual-cores. In modern laptop CPUs, you will generally also be able to access speeds that are yielded by increasing the multiplier above that idle speed, on up to the top speed of your chip. An exception that I know of is Pentium M Dothan 400 FSB, which access speeds that are available every *two* multipliers. They idle at 600 MHz (6x) and then you get 800 MHz, (8x), 1.0 GHz (10x), on up to the top speed of your CPU, which can be an odd multiplier as in the case of the Pentium M 765 at 2.1 GHz. In that case, you'd get 600 MHz, 800 MHz, 1.0 GHz, 1.2 GHz, 1.4 GHz, 1.6 GHz, 1.8 GHz, 2.0 GHz, and 2.1 GHz.

Older CPUs like the Pentium III-M and Pentium 4-M have a defined idle speed and then the top speed available. My 2.20 GHz P4-M runs at the common idle speed of 1.20 GHz and then its top speed of 2.20 GHz. Desktop CPUs also have a common idle speed across their line, but the available multipliers can vary. AMD's 90nm and 65nm chips all have a 1.0 GHz idle and then go up to 1.8 GHz, then increase one multiplier (200 MHz) from there. This is because the slowest chip in those lines is 1.8 GHz and AMD didn't feel the need to enable multipliers between 1.0 and 1.8 GHz. All multipliers between 1.8 GHz (9x) and your chip's top speed are available, so my X2 4200+ at 2.20 GHz has the 5x one for 1.0 GHz idle and then 9x for 1.80 GHz, 10x for 2.0 GHz, and then 11x for the top speed of 2.20 GHz. A faster CPU like the 5200+ will have the same multipliers available as my 4200+ does, but adding in 12x for 2.4 GHz and 13x for its top speed of 2.6 GHz.

The speed is controlled by a kernel-space or user-space program such as SpeedswitchXP. The program polls the CPU for current speed and load, and then uses a table of load and speed values to compute what it should do. In SpeedswitchXP, you can set the values where it will switch. That will be the "up" and "down values. You can also set the way that the multipliers will be used, such as going up and down one by one or jumping from top speed to idle if needed. You can also lock the CPU's speed at one value, which is what I think you want to. You can do that by setting the power properties to "performance" and then the CPU will always run at top speed, no matter the load.
March 8, 2007 3:11:31 PM

Yup, agreed.

Lengthy explanation and well written as usual. :D 

That is why I am confused about the OP's laptop. The FSB shouldn't be altered as it adversely affects the entire system, but the multi is free game (from min to max, or any accessible value in between). So, with the multi locked the system should stay at max speed, but his isn't, which is what concerns me. It isn't that it is impossible to do, it just isn't how EIST rolls.

Your comments are always appreciated. :D 
March 8, 2007 3:15:18 PM

Quote:

That is why I am confused about the OP's laptop. The FSB shouldn't be altered as it adversely affects the entire system, but the multi is free game (from min to max, or any accessible value in between). So, with the multi locked the system should stay at max speed, but his isn't, which is what concerns me. It isn't that it is impossible to do, it just isn't how EIST rolls.


SuperFly, Not to beat a dead horse, but I really think it's TM2 (thermal mgmt 2), which is BIOS controlled. This 3rd party app is probably trying to write the clock multiplier MSR, which is promptly being over-written by BIOS again. Please note that TM2 is not EIST!

PB
March 8, 2007 3:21:52 PM

Quote:

You can also lock the CPU's speed at one value, which is what I think you want to. You can do that by setting the power properties to "performance" and then the CPU will always run at top speed, no matter the load.


Right, that's what I want to do. But I set the power properties to "performance" and still, the CPU lowers its frequency from time to time. So speedswitchXP has locked the multiplier to 13x but the frequency of the CPU still gets down to 900MHz (while at 13x) occasionally. Under heavy load (game) It would typically stay at 1700MHz for 5-10mins, then go down to 900Mhz for 3mins or so, then gets back to 1700MHz, and so on. So my guess is that the CPU lowers its freq. because of high temperature, then gets back to normal when the temperature is fine...
btw, my cpu is a pentium M centrino M740 (Dothan).
March 8, 2007 3:33:54 PM

Natty,

I have the same issue with my Dell D600.

If playing UT2K4 it heats up and will "down step" and cripple you while playing. I have found that using something like a cd/dvd case on both edges of the laptop elevates it enough to sometimes not slow down. For further measure (at lan parties) I elevate and ventilate :) ... I take a small desktop fan with me and I almost NEVER have these issues.

____|---------|____

Especailly if you know the location of your processor (this should be easy since it will likely be the hand warmer part of your laptop). (for me it is the lower left of the machine) sometimes just elevating this side a little will help.

You are definitely running into thermal issues. I have tried everything (like you) to stop it from down stepping but to no avail.

Dell has a utility called QuickSet "set profile using this to "Always On, or Maximum Performance" (no go)

Try using Windows power scheme set to "Always On" or "Home/Office Desk" (no go)

Try disabling Speed-Step in the Bios (no go)

So the only thing that works for me is the elevate and ventilate. I would get one of the notebook coolers that are for sale at Comp or on the net but it is my Work Machine first and my Play machine second :) 

Hope this helps!!
a c 99 à CPUs
March 8, 2007 3:43:36 PM

ches111's got it. Newer Intel CPUs will skip clock cycles to reduce their heat output. It doesn't affect the FSB speed or the multiplier, the CPU just "ignores" the clock tick. You can see that in action in Tom's article on the P4 560: http://www.tomshardware.com/2004/11/14/the_p4/. This video shows the CPU skipping clock cycles to cool itself, which is exactly what your CPU is doing.

At least your CPU will skip clock cycles and not just shut off on you. My old laptop's ACPI is starting to become flaky and sometimes the CPU fan will not come on when it's supposed to. The old P4-M will run at its normal speeds and not throttle until it hits the critical shutdown temp of 93 C, then the computer clicks off. Notebooks are NOTORIOUSLY bad at overheating, so something like Antec's Notebook Chill Pad would be help as it blows air around the bottom of the notebook and greatly reduces temperatures of things inside the notebook that are ill-ventilated, like hard drives. I got my brother one for Christmas and his laptop runs a bit cooler with it, so I'd recommend you get one if you game much on your notebook.
March 8, 2007 3:46:06 PM

Thxs ches111, It does help! It's always good to know you're not the only one :) 
That behavior sounds logical but still strange to me. Indeed, an M740 CPU is supposed to be stable at 1.7GHz. The fan is here to cool it down (mine is quite noisy btw). So if it feels the need to lower its frequency because it's too hot, that means it's not stable, right ? I fear what will happen this summer!
March 8, 2007 3:54:02 PM

Quote:
ches111's got it. Newer Intel CPUs will skip clock cycles to reduce their heat output.


Not exactly. Older Intel CPUs did this with #PROCHOT assertions from the ICH. New Intel chips handle thermal events by 1) changing PWM or 2) changing multiplier/VID setpoints. These are called TM1 or TM2, respectively.

MU, I disagree with you simply because the thermal control is much more nuanced than just gating BCLK.

I recommend reading http://www.digit-life.com/articles2/intel-thermal-featu... or the Intel Software Development Manual Vol 3B Section 13.2.2

PB
March 8, 2007 3:54:17 PM

Natty,

I it is not a matter of it NOT being stable. I would venture a guess that for the most part it works properly almost ALL the time while doing non-graphics/non-IO intensive work.

It is a built in protection mechanism for the CPU... I think most vendors do not allow you to disable this since they know the poor ventilation characteristics of the notebook style case.

A buddy of mine was upset that his machine would shut down all the time (Desktop Replacement P4 Notebook) when he stepped away from it.

Turns out he used it mostly at night in bed surfing the net/working (yeah right ;) ). When he would get up and leave the machine he would leave it setting on the bed. Guess where the vents were for the machine? When I explained it to him he was able to set it on the night stand and not have that issue. (frankly I would be glad that a "P4" would shut down) COuld you imagine the fire!! Bed sheets and a P4 just do not mix :) .
March 8, 2007 3:56:34 PM

Be glad yours only steps to 900mhz.

Mine is the Centrino 1.6ghz and it steps down to 600mhz and on rare occasion down to 450mhz.
a c 99 à CPUs
March 8, 2007 4:00:24 PM

The CPU can be very stable and still overheat. A stable CPU will execute all instructions and give the correct answer at the given clock speed and voltage. Unstable CPUs will give incorrect answers as some transistors may not switch properly- that's why overclocked CPUs are tested by working on complex math problems with a known answer, such by using the fast Fourier transform (FFT) or searching for prime numbers.

Intel ships the CPUs out the door when they will give correct answers at their clock speed and voltages AND be within their listed thermal properties, which are given in the whitepapers on Intel's website. Unlike desktops, where the CPU maker ships a heatsink that's guaranteed to work, it's up to the notebook maker to ensure that the CPU is properly cooled. My guess is that Dell didn't count on somebody running the CPU at full speed for long periods of time and thus skimped on cooling to save weight and cost. Thus the cooling is fine for low to moderate load situations but insufficient for running flat-out for long stretches. Most laptop makers do this- it's not something specific to Dell.

The sad reality is that if you want to do real hard work with a computer, get a desktop. I learned that one after having a hot-running laptop as my only machine for many years. A decent normal-sized desktop can be run at 100% load for years on end and not complain because it's made for it. (Note that some of the SFF units aren't meant to be run like this- the Dell OptiPlex SX280 with a Pentium D is one example.)
March 8, 2007 4:04:51 PM

I get it. So if I understand correctly, my CPU is clocked at 1.7 GHz but the fan cannot cool it down properly so in practice it runs a bit at 1.7GHz and a bit at 900MHz. So my laptop should either have a lower CPU or a better fan. I feel a bit cheated there...
March 8, 2007 4:21:13 PM

You could try a program called Centrino Hardware Control. I use it along with Speedfan to control my laptops freqency settings and to monitor temps. There is a setting in the program to fix your frequency and voltage at whatever you want. Just goole it and you should find it no problem.
March 8, 2007 4:36:57 PM

Thxs kriztoff, but I'm a bit afraid to fry my laptop with those tools. If the CPU lowers its frequency because it's too hot, I guess it's not advised to force it to increase it, right ?
Anyways, looks like my problem has found its solution (quite not what I expected, but well..). So thxs to all for your answers,
N_D
March 8, 2007 4:39:04 PM

This is just thermal protection at work. The PROCHOT is registering temps too high and passing that info to the DVS to down clock/down volt your CPU.

Strictly a protection mechanism.

Your processor is just fine.. It is the thermo dynamics of the case/cooling system of your notebook that is at issue. This is a known/common issue with notebooks. People want their notebook to be smaller, with less weight, but MORE powerful.. This means the device case also gets smaller but the proc capabilities also grow.

Intel and AMD both did a fine job with the Pentium M and Turion processors. They run on much less voltage then predecessors, with better performance and with better thermals.

Can't blame either of the two companies for others case designs.
March 8, 2007 4:41:00 PM

You could take it apart and clean the dust out of it, that'd likely help. Depending on how old it is, it likely has some pretty thick dust in there.
March 8, 2007 4:44:52 PM

Kriztoff,

I will try Notebook Hardware Control (the new version of Centrino Hardware Control) found here:

http://www.pbus-167.com/chc.htm

I am hosting a Lan party in "the Cave" ;)  Friday night and I will test it out. I will have one of my buddies use the laptop for a while to see if it down modes!!
March 8, 2007 4:53:35 PM

Kriz,

Playing with that app now... Looks interesting...

Just wonder if its settings are JUST the same thing only different.

I will find out tomorrow night!
March 8, 2007 5:19:17 PM

Yeah it sounds like thermal throttling is activating on your laptop. you should try using the program rmclock to undervolt the cpu and manage power management. you should be able to lower heat output a little or if you have to, you can set the max p-state to the multiplier below max and see if it can run at that speed under load without overheating.
March 8, 2007 5:26:31 PM

That app. (rmclock) looks better than speedswitchXP (no offense to the author). I'll have a look at it. Thxs!
N_D
a c 99 à CPUs
March 8, 2007 8:28:56 PM

Thanks for the link! I guess my knowledge of Pentium 4/M thermal management was a little bit out of date.
March 8, 2007 8:54:34 PM

I have confirmation that the low phase is caused by the bus lowering its frequency and not the cpu itself. In a low phase, the bus frequency gets down to 70MHz instead of 133MHz. Here are two screenshots that I took in each phase:
http://www.geocities.com/natty_dread78/LowPhase.JPG
http://www.geocities.com/natty_dread78/NormalPhase.JPG

That explains the cpu frequency going down to 900Mhz with a 13x multiplier. But is this normal behavior ? Is there a way to go around this and is it safe ?
Thxs,
N_D
March 8, 2007 8:58:26 PM

What are your temps? Could you get a screen shot of that too? Just so we can see if it is a throttling issue.

Side note: XP in different languages is funny, your start button is nuts.

Yes... I realize it does say start. :p 
March 8, 2007 9:41:02 PM

For some reason, rmclock fails to read the core temperature. So I installed speedfan and monitored the temperatures in both phases. Here are the screenshots:

http://www.geocities.com/natty_dread78/temperature_normal.JPG
http://www.geocities.com/natty_dread78/temperature_lowPhase.JPG
Apparently, the low phase is not caused by overheating ??? (temp. stays at 62C)

ps: yeah, my windows is in french! 8O I don't like it, but it has to, for my girlfriend doesn't speak english.
March 8, 2007 9:55:24 PM

The only reason I have to think it is a thermal event is because the CPU load stays in the upper 80% range. I don't know if that is the system idle process or what, I don't use rightmark.

Since it is a laptop the CPU temp won't change quickly unless the load is taken off. If it is a straight change in CPU speed without a change in load, the temp will likely stick around until the change in heat output actually makes it through the heatsink.

I really wonder about those loads, what are you doing on the laptop? That is what concerns me.

So in short... I am still perplexed lol.

p.s. my gf's aren't allowed to touch my computers... I am slightly anal about them, plus I haven't met one smart enough to trust to not get viruses on there. :lol: 
March 8, 2007 10:20:54 PM

Yes, I think it takes into account the system idle process. I don't know either how rightmark works exactly, but the windows taskmanager shows 98 or so for the idle process. But to make the CPU go in its low phase, I have to load it with a demanding game (e.g. farcry). Otherwise, it remains in the normal phase (1.7GHz).

ps: lol your computerS! How many do u have ? 8O As for me I only have one (well not exactly, but I fried one :? ), so I don't have the choice, I have to share with my girlfriend who doesn't own one...
March 8, 2007 10:42:55 PM

Natty,

I am guessing that this is still a thermal event.

The way speedstep works is:

It will lower the clock speed/bus when it detects that (pay attention here) the CPU is no longer needed at the higher level. So when it detects a down time on the cpu where you no longer need it, the cpu will downclock.

On the other hand when loaded it is supposed to give the full processing power.

You are seeing the opposite of this... You are hitting a thermal event guaranteed.

Do the same test you just did with a fan on the laptop and with it slightly elevated as I told ya before.

I bet you then are able to play for a far greater time. I would be willing to put MUCH money on it as I can demonstrate this same reaction myself.
a c 99 à CPUs
March 8, 2007 11:12:36 PM

That and your Vcore and multiplier is still at maximum. If it was anything with SpeedStep, then your Vcore and multiplier would be lowered. How about you read the link about Intel Thermal Monitor that phanboy posted and use the same RightMark software and look at how your CPU behaves. What I'd do is lock it at maximum speed and then let it idle, watching the Vcore temp, and CPU speed over time. If it does not throttle, then put about 50% load on it (again at full speed) and see if it throttles at all, again, watching the Vcore. Finally, load up the CPU at 100% load and then watch for throttling. If the CPU throttles only under heavy loads, then it's almost certainly the thermal monitoring affecting the CPU.
April 29, 2009 8:19:23 AM

Guys, I may have turned up a couple of years too late, but I think I know what the problem is - I have the same issue myself with an Intel Core 2 Duo Centrino Mobile. And I STILL haven't found a way to fix it.

The problem was introduced by Intel to tackle power consumption - it's called Dynamic FSB Frequency Switching, and it's an in-built feature of these processors!

According to Intel:
Quote:
Dynamic Front Side Bus Frequency Switching technology allows the FSB to be underclocked in order to save power / battery life. Previously, Core™ 2 Duo Mobile processors operated in Low Frequency Mode (LFM) when set for reduced-power operation, but the FSB was kept at full speed. By cutting the FSB speed in half... the Santa Rosa notebooks should be able to save more power than ever before. Intel refers to this combination of chip and FSB underclocking as Super Low Frequency Mode, or SLFM.
Intel Santa Rosa Platform


I've tried a lot to get around this problem! The last release of RightMark Clock (mentioned above) has some rudimentary support for DFFS, but it can't even detect that my CPU has this feature, and development of the tool died over a year ago.

Am I doomed to watch my games slow to a crawl after 10 minutes of play???
!