New Centrino Software to reduce fan noise and heat

Message

beeblebrox · #1 Post by **beeblebrox** » Thu May 12, 2005 1:19 am

I am not sure, whether this already has been posted here somewhere, but I just want to let you know about the most perfect notebnook tool I found so far.

It is called Centrino Hardware Control (release 1.7 currently), developed by a hardware control systems company and has been working on my T40p without problems for weeks.

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

I was able to successfully reduce the heat by up to 20°C and overclock/underclock the ATI dynamically (ATITool didn't work on my TP).

I found that overall voltage for the Banias 1.6Ghz could be reduced by up to 0.3V, thus drastically reducing heat (you know that voltage goes in squared into heat production).

Whereas the Banias produced 80°C at tests before, it ran at 60°C afterwords. During office work the temp is around 37°C.

The Dothan /Sonoma should be even better suited to undervolting.
You can NOT damage the Pentium, but you have to find your min voltage. Intel Pentiums are within a safety zone, you can find out your lowest working voltage.

The TP is dead silent now, I completely forgot what a fan is

Regards,
Zaphod

beeblebrox · #2 Post by **beeblebrox** » Thu May 12, 2005 1:26 am

I forgot to say, that I effectively ended up with 40 minutes more battery time !!

Unfortunately the option to intriduce the HLT state for the CPU is not yet included (what rmclock does), so there is still the option to even increase battery life. But anyways, the software works without problems and is a very good replacement for SpeedSwitchXP.

Kenn · #3 Post by **Kenn** » Thu May 12, 2005 2:30 am

Wow, great find! Thanks for the heads up, Zaph.

Currently downvolted my 1.8 Dothan from 1.340v to 1.084, and it can probably go a little lower. When running at 600MHz, I'm having no problem running at the minimum 0.700v allowed by the program (down from 0.988)

It makes a huge difference in temperature and power draw too. At 1.8Ghz, CPU at load went from 49o to 43o, battery drain went from 24w to 19w(!).

So far, it looks like a total win/win. I'm very impressed!

Plinkerton · #4 Post by **Plinkerton** » Thu May 12, 2005 2:57 am

Seems like a great program, and would be quite helpful, but I really have no idea what I should change. I have a 2378-FVU. There seem to be so many choices on the voltage levels. And I don't quite understand "clocking".

Anyone want to shed some light?

Thanks,
Plink

beeblebrox · #5 Post by **beeblebrox** » Thu May 12, 2005 3:49 am

Plinkerton:

the analogon would be the rpm on your car.
The factory set them to something like 2500 to 3000 rpm for idle time (like waiting at the red stop light) just to make sure the engine doesn't die off.

if you fine tune the electronics / mechanics, you easily find out that the motor runs perfectly at 1600 rpm in idle state.

For the factory to make sure they never ever have any problems they just stay well above the minimum requirements.

Same with Pentiums, there are so many different versions with newer releases, that they prefer to give guidelines for computer manufacturers. That's makes perfecr sense.

Again, you can fine-tune your car to increase power and reduce consumption. That's what you do here with the Pentiums. You don't overclock, but just put the Centrino on "Diet" and let it do the same work.

The other way round it using the voltage to overclock, however the chipset does not allow this (T40 series). don't know about the newer chipsets. Some have succeeded to boost the Pentium by 200 Mhz (didn't work on mine!), by increasing the FSB clock and/or voltage (Aopen boards).

So, you have a bandwidth of option to improve your Centrino without doing damage.

To be honest, I am quite happy with 40 minutes extra battery time... not much, but for free!

mlykke · #6 Post by **mlykke** » Thu May 12, 2005 4:50 am

Beeblebrox - you rock dude

This is very cool. I will be testing this on the three thinkpads i will receive today. But i swear - if we get a SINGLE post about how annoying the fan is on this forum again we should all just throw this link in their face.
Cause if it works just as good for everyone we could solve some problems on our own instead of waiting for IBM to do it.

And 40 minutes more battery time...thats ALOT

Im thinking something in the lines of a T42P or T43P with a 6 cell battery instead of the 9 cell and still retain almost the same runtime as a complete stock T4xP with the 9 cell battery - Very nice!

#7 Post by **K. Eng** » Thu May 12, 2005 9:54 am

Very nice program. It's much slicker than Mobile Meter...

Plinkerton · #8 Post by **Plinkerton** » Thu May 12, 2005 12:06 pm

So in the CPU Voltage section, why are there so many things to change, even though they all have the same numbers? They all have a multiplier of 6, and a voltage of 1.340V. And the first box has a 6, and a volatge of .998. What does this part mean, and what should I change on this part?

Thanks again,
Plink

BoostedA4 · #9 Post by **BoostedA4** » Thu May 12, 2005 1:27 pm

Yes, I have heard a lot of success with this, though also a few who receive BSoD every now and then. I think I will try the undervolting tutorial over at www.notebookforums.com and compare it with this. My new T43 arrived today, so I'll play around this weekend.

Just as a side note, I think it is really beneficial for thinkpad members to check out www.notebookforums.com on a regular basis. It is an extremely popular message board (though largely for the gaming crowd), but they find the best tweaks to notebooks, i.e. this above posted link, as well as other undervolting tutorials, fan/temperature effects, overclocking, and just any general care. Though this website is great for thinkpad specific questions, when it comes to items such as the CPU, the info at notebookforums is amazing.

emorphien · #10 Post by **emorphien** » Thu May 12, 2005 1:36 pm

Very intersting, I'll have to try this when I get my centrino notebook.

Kenn · #11 Post by **Kenn** » Thu May 12, 2005 3:35 pm

Plinkerton wrote:So in the CPU Voltage section, why are there so many things to change, even though they all have the same numbers? They all have a multiplier of 6, and a voltage of 1.340V. And the first box has a 6, and a volatge of .998. What does this part mean, and what should I change on this part?

Thanks again,
Plink

That lets you change the voltage for each supported Speedstep variation of the CPU. Pre-sonoma, 6X multipler is 600MHz, which by default runs at 0.988V. 18X is 1.8GHz, which defaults to 3.40V for a P-M 745. You can change the voltage for each setting individually. One of the text labels on the page is clickable - it will show you an HTML table of all the different speedstep speeds available for each CPU and their corresponding default voltage.

TakiusProteus · #12 Post by **TakiusProteus** » Thu May 12, 2005 5:08 pm

Can this be done on an X series?

jodo2017 · #13 Post by **jodo2017** » Fri May 13, 2005 4:13 am

TakiusProteus wrote:Can this be done on an X series?

Yes, and it works really well.

The Wattage got reduced from 10.5 to 8.7 in 600 MHz (that's what I usually use when on battery) that’s ~20% giving me 1h+ extra battery-time on my 8-cell.

beeblebrox · #14 Post by **beeblebrox** » Fri May 13, 2005 6:48 am

So, Jodo
you have beaten me. I managed to get "only" 40 minutes more battery time.
I need to figure out how to get that 1+ hour.

Actually, I suggested to the author of that program some hints for improving the software. If CHC 1.7 would also implement the HLT instruction of the centrino we would get probably 2+ hours more from the same battery.

my reasoning:
in idle state (e.g. between two key strokes), the CPU does nothing but polls the USB and other ports for power saving etc.

If CHC 1.7 would implement the HLT (halt) instruction, similar to rmclock (the other centrino software), the CPU would stop (!!) completely and be awakened again by an interrupt (e.g. the next key stroke).

Now, a Pentium that is stopped completely even uses far less power than a Pentium that lazily scans the ports.
If you type 2 characters/second, the 600 Mhz CPU still makes 300 million wasted cycles in between two key strokes.

So, I am looking forward to CHC 1.8

Mack · #15 Post by **Mack** » Fri May 13, 2005 10:44 am

So this program only works for the centrino, not for a thinkpad X24 with intel speedstep?

Edit: I checked the website, and cannot find the author's email.

Kenn · #16 Post by **Kenn** » Fri May 13, 2005 1:40 pm

beeblebrox wrote: If CHC 1.7 would implement the HLT (halt) instruction, similar to rmclock (the other centrino software), the CPU would stop (!!) completely and be awakened again by an interrupt (e.g. the next key stroke).

How fast does it go into and recover from HLT? I ask because there is definitely a noticeable delay in the clocking up and down of the speedstep mechanism, and adding even more jerkiness to that would probably suck

mlykke · #17 Post by **mlykke** » Fri May 13, 2005 4:16 pm

Has anybody been able to use the fan control in the software?

This could help alot on making the machine silent.

Or does anybody know of any other software that can control the fan?

The best thing would be something along the lines of the CPU stepping control just with the fan so you could turn down the RPM's.

Torque · #18 Post by **Torque** » Fri May 13, 2005 6:25 pm

mlykke wrote:Has anybody been able to use the fan control in the software?

This could help alot on making the machine silent.

Or does anybody know of any other software that can control the fan?

The best thing would be something along the lines of the CPU stepping control just with the fan so you could turn down the RPM's.

Sadly, no.

I would really like to override and manually adjust the fan rpms.

How low have you guys gone, in terms of voltage at 6x multiplier?

jmt · #19 Post by **jmt** » Fri May 13, 2005 8:41 pm

Will this program work on the Ultra Low Voltage 1.10GHz processor?

mlykke · #20 Post by **mlykke** » Sat May 14, 2005 4:50 am

Hey Torque - Another dane huh

At the 6x multiplier i went down ALL the way on all of the three Thinkpads im currently playing with.
On the T42P with the 9 cell battery - The batterystatus actually claimed to give me almost 7 and a half hours of battery time. [censored]! Thats ALOT more than according to the specs from IBM.

dvorak · #21 Post by **dvorak** » Sat May 21, 2005 9:17 am

CHC 1.7 interferes standby, cold-boot, standby works nicely, after I start CHC, wait a minute or so and try to standby: "System Standby Failed", device driver for the MS ACPI ... is preventing the machine from entering standby.

I'll try CHC 1.8 beta 01, perhaps it fixes this.

Kenn · #22 Post by **Kenn** » Sun May 22, 2005 11:41 pm

I'm running stable at 600Mhz 0.700v as well. However, I just read that some people were having problems with random crashes playing DVDs at this setting, though everything else seems to work fine. Might be something to check out.

Dvorak, let us know if 1.8b1 adds any new functionality!

dvorak · #23 Post by **dvorak** » Mon May 23, 2005 6:13 am

Kenn wrote:I'm running stable at 600Mhz 0.700v as well. However, I just read that some people were having problems with random crashes playing DVDs at this setting, though everything else seems to work fine. Might be something to check out.

Dvorak, let us know if 1.8b1 adds any new functionality!

Doesn't add anything new, and didn't fix the standby problem either. Does standby work on your system while CHC running?

I'll continue using RMClock, takes less resources and does the same job :)

Kenn · #24 Post by **Kenn** » Mon May 23, 2005 1:13 pm

dvorak wrote:Doesn't add anything new, and didn't fix the standby problem either. Does standby work on your system while CHC running?

I'll continue using RMClock, takes less resources and does the same job

Yep, I use standby, hibernate, and power-off all the time, and have never had a problem with CHC running...Have you tried updating to the latest BIOS and Embedded Controller Program versions (3.14 and 3.04 I think?)

I think I'll give RMClock a try though, it sounds like it has some additional controls and I'm interested in seein if the HLT command results in additional battery life savings...

dvorak · #25 Post by **dvorak** » Mon May 23, 2005 2:22 pm

My problem was only related to standby, while CHC was running Windows notified of something that prevented the machine from entering standby, hib. worked fine. Afterwords closing CHC didn't help either, only a reboot.

RMClock's HLT app increased the battery drainage on mine, it's more of a workaround than feature, prevented RMClock from reading CPU load wrong, hence on adaptive it sometimes stayed at full speed, on battery I use minimal, so it doesn't matter there. And even if it stays on max all the time, at ~1.1V the temp's always around 37 degrees. Close to .988V of default voltage to the minimal speed too.
Currently I'm nicely sitting on 600MHz, and all has been working nicely for the past days.

Kenn · #26 Post by **Kenn** » Mon May 23, 2005 3:44 pm

Hmm, I take it you can standby just fine using RMClock or if CHC never loads on startup?

I haven't heard of anyone else with a thinkpad running into this problem, perhaps you could try terminating processes or stopping services one by one and seeing if any of those changes let you go into standby?

I just tried RMClock by the way. You're right, HLT doesn't seem to do any good (in mobilemeter the power drain remains about the same regardless of whether it's on or off). Also, for some reason it seems like making changes to CPU voltage in CHC is more "responsive" as far as power drain goes - within 2 seconds of going idle the CPU drops to idle wattage drain, whereas with RMClock, it slowly settles to the same point over the course of 15-20 seconds. I need to run some more tests, but it looks like there may be some differences under the hood of these applications.

dvorak · #27 Post by **dvorak** » Mon May 23, 2005 4:05 pm

Yes, standby works fine when CHC is not loaded before (either at startup or manually). I can't imagine why, perhaps somewhere deep in Windows' settings there's something related to .NET apps, haven't had time as I'm not so needy to use CHC :).

What do you mean by responsive? Like one second you're draining 14w (from using max speed for example) and second later it's down to 10w?

Only typing and using lvl 3 brightness, it'll stay near 9.5-10w on mine, when turning brightness to 0, I've even see as low as 8.8-9w, with RMClock, no net, no wifi, hdd spinning as far as I remember.

Kenn · #28 Post by **Kenn** » Mon May 23, 2005 9:47 pm

On my 15" running 0.700v 600MHz, lowest brightness and fan off (38*C), no net and only MS word and various 0% CPU processes running, I get about 11.45watt discharge. Thank the Flexview screen for that. That's about 5.5hours on a 9cell with 10% wear. Sounds like you'd be getting over 7 hours with the hi-cap battery

Yeah, by responsive, it seems like on my machine, via mobilemeter, changing the voltage from CHC results in a pretty sudden drop (within 2 seconds) of wattage drain, whereas from RMClock, it seems to slowly taper down over 10-20 seconds. Probably just a fluke, but it was something I noticed while switching between the programs.

beeblebrox · #29 Post by **beeblebrox** » Tue May 24, 2005 8:33 am

I did not have much time to play with the tools, but it seems to me that the programs are using some software routines either based on .Net or some windows parameters, which don't work quite often.

I used to play with mobilemeter and SpeedSwitchXp, but their readings about CPU speed is random more or less, since the Pentium M does not tell Windows about its internal states, and most software just reads Windows data. Example: if the CPU ist stopped, the meter shows 100% CPU utilitzation, because Windows, or other software, wait for the results of their initiated routines. The longer they wait, the busier the CPU. In theory. However, if the CPU is stopped, the results never come, until the next interrupt wakes up the CPU. I am sure, that a stopped CPU that does nothing consumes less than an idling CPU.
In SpeedSwitchXP I never knew about the real state of the CPU, not even on Pentium III notebooks. Windows, internal and calculated speed differed dramatically.

Similar, I have not had time yet to find out how the wattage parameters are obtained by the tools. I guess, they are calculated as well. Some programs read out quite frequently, others take a longer term statistical approach. Thus the differences. It takes the Pentium only a few micro seconds to switch between different states, so software always shows an average over a few seconds.

Therefore I started to trust my own results. With CHC turned on (my Banias only goes down to 0.78 to 1.276V) and a brandnew 9 Cell (3 full cylces so far), no Wifi, TFT level 3, HDD to 3min and working on Word and Powerpoint, I got yesterday... 7h48 min.

I think, for me that's ok. I checked with the Centrino LV (Low Voltage) CPUs of the X-series and it seems to me that because of production variance I ended up with a Pentium CPU that is very similar to the LV series. Good to know.

I wonder, if there is a program to control the Pentium III for the T2x series as well.

BTW: Does anybody know what software is useful to overclock the T40?
I tested the ones that are suggested for the HP nx6000 and the TP X31, but they always crashed immediately, telling me it's the wrong chip set.
I thought 855GE and 855PM were quite similar... ?!
Anyone?

#30 Post by **K. Eng** » Tue May 24, 2005 8:12 pm

I'm getting 3 hours on 31 Watt/hrs

WiFi is on and brightness is set to medium.

My 1.30 GHz Banias is set to ULV voltages at the lower multpliers and LV voltages at 1.20 and 1.30 GHz. No stability problems in the past week!

beeblebrox wrote: Therefore I started to trust my own results. With CHC turned on (my Banias only goes down to 0.78 to 1.276V) and a brandnew 9 Cell (3 full cylces so far), no Wifi, TFT level 3, HDD to 3min and working on Word and Powerpoint, I got yesterday... 7h48 min.

I think, for me that's ok. I checked with the Centrino LV (Low Voltage) CPUs of the X-series and it seems to me that because of production variance I ended up with a Pentium CPU that is very similar to the LV series. Good to know.

New Centrino Software to reduce fan noise and heat

New Centrino Software to reduce fan noise and heat

Who is online