Undervolting. The next thing I'm going to learn about.

T4x series specific matters only
Post Reply
Message
Author
fefrie
Junior Member
Junior Member
Posts: 435
Joined: Wed Jan 24, 2007 3:29 am
Location: Vancouver, Canada

Undervolting. The next thing I'm going to learn about.

#1 Post by fefrie » Sat Dec 10, 2011 9:20 pm

If I were to make a quick summary, as to what I've found out so far....

Undervolting, asides from underUNDERvolting, has no performance downside. It's main benefit is lower CPU temperatures and increased battery life.

The reason undervolting works is that when chip makers are making chips, they produce huge batches of chips, and some chips are better or more efficient than other chips, even within the same batch. So when a batch of chips comes out, instead of testing each chip to find out the optimum voltage to run a chip at, they apply an arbitrary value that all chips should be run at so that 99%ish of all the cpus will function within spec without creating too much heat that the heatsink can't take away.

So..... with that in mind, there are definitley chips out there that are probably pretty efficient ("A-Spec") that get yanked and get to live their lives in super computers at low voltages, and we probably get B and C spec chips, that run a C spec voltages because the cost/profit ratio is not there to justify testing/selling B/C chips with their own voltage specs.

The law of averages being what it is then means that we all probably don't have the lowest 1% quality chip, but falls somewhere in the 2-100% category, so we can therefore undervolt to a certain degree, and since all chips are made in different batches, your experiences with undervolting numbers aren't necessarily in line with what I would find to prevent crashes and cpu performance issues.

And that's why if I want to undervolt, I have to take everyone elses results with a grain of salt?

Am I thinking in the right direction? Because if I can get less heat created because the cpu is running at the right voltage and not bleeding off any excess voltage as heat, then I can get much better battery life, and since the cpu can't use any voltage over it's performance envelope anyways, I will experience NO performance issues if my settings are set properly.

So basically there is no reason as to why I *should not* undervolt. If the settings are set right, then the results are all positive?

So that is what these undervolting programs do then? Given a certain CPU load, the program then delivers the optimum voltage and therefore amperage to maximize every electron in the battery.

It's essentially an O2 sensor for computers? And instead of running the computer rich, with high HC (heat), it runs the changes the amount of gas injected (voltage) so it runs at the proper air/fuel mix and there for most economy. The only thing that has to be considered is to not run too lean (undervolt) or you'll get backfiring and other issues (CPU glitches)

So what's the easiest best background program to do this?
IBM Thinkpad T23 1.13 2647-9LU 640MB Ram 40GB hard drive SOLD!
T42 SXGA 1.7 64mb xp

RealBlackStuff
Admin
Admin
Posts: 17512
Joined: Mon Sep 18, 2006 5:17 am
Location: Mt. Cobb, PA USA
Contact:

Re: Undervolting. The next thing I'm going to learn about.

#2 Post by RealBlackStuff » Sat Dec 10, 2011 10:36 pm

You are working with technology AND with batteries that are by now well beyond their prime.
If you want better life from a battery, get newer technology, both in the laptop AND in the battery, but YMMV.
Have a look at this: http://forum.thinkpads.com/viewtopic.php?f=2&t=100929
Lovely day for a Guinness! (The Real Black Stuff)

Check out The Boardroom for Parts, Mods and Other Services.

fefrie
Junior Member
Junior Member
Posts: 435
Joined: Wed Jan 24, 2007 3:29 am
Location: Vancouver, Canada

Re: Undervolting. The next thing I'm going to learn about.

#3 Post by fefrie » Sat Dec 10, 2011 10:43 pm

Ya, so true. Can't let go, can't move on.

The T4X community is dead.

LONG LIVE THE T4X COMMUNITY!
IBM Thinkpad T23 1.13 2647-9LU 640MB Ram 40GB hard drive SOLD!
T42 SXGA 1.7 64mb xp

Orclas
Sophomore Member
Posts: 190
Joined: Mon May 02, 2011 4:12 pm
Location: Stockholm, Sweden

Re: Undervolting. The next thing I'm going to learn about.

#4 Post by Orclas » Sat Dec 10, 2011 11:06 pm

fefrie wrote:The reason undervolting works is that when chip makers are making chips, they produce huge batches of chips, and some chips are better or more efficient than other chips, even within the same batch. So when a batch of chips comes out, instead of testing each chip to find out the optimum voltage to run a chip at, they apply an arbitrary value that all chips should be run at so that 99%ish of all the cpus will function within spec without creating too much heat that the heatsink can't take away.

So..... with that in mind, there are definitley chips out there that are probably pretty efficient ("A-Spec") that get yanked and get to live their lives in super computers at low voltages, and we probably get B and C spec chips, that run a C spec voltages because the cost/profit ratio is not there to justify testing/selling B/C chips with their own voltage specs.
Well, I have rather understood it that the CPUs from the same family are all done in batch and then tested to see which frequency they can handle, i.e. a 1.5 and a 2.0 Ghz Dothan comes from the same wafer, but since the circuit error % vary, some are deemed perfect enough for the highest speeds, others only suitable for the lower range. If voltage at all plays a part here I don't know - except that high frequency = higher voltage - but my guess would be that a 1.5 Ghz CPU running at 100% and a 2.0 CPU "speedstepped" to 1.5 runs at the same voltage.

I don't know if you ever studied statistics, but I think that it from the manufacturers point of view has to do with things in the direction of probability, normal distribution, standard deviation and such. Or, in plain English, something like "how do we specify CPU X (frequency and voltage) so that the probability of error/instability is less than 0.x% for all possible usage patterns, component combinations (MB circuits, power supply, voltage regulator circuitry/PWM controller etc), cooling etc?".

In the case of specified voltage, that would for example make the CPU stable also in some crappy build where power supply and voltage regulating circuitry lets the voltage fluctuate by perhaps +/-20%, whereas a high quality build would allow for settings (both voltage down and/or frequency up) many standard deviations out on the curve.

I feel I should by now once again point out that above is a combination of loosely understood facts and some hunch based assumptions (I'm no engineer, I have a business degree... :roll: ).
fefrie wrote: It's essentially an O2 sensor for computers? And instead of running the computer rich, with high HC (heat), it runs the changes the amount of gas injected (voltage) so it runs at the proper air/fuel mix and there for most economy. The only thing that has to be considered is to not run too lean (undervolt) or you'll get backfiring and other issues (CPU glitches)
Interesting analogy and roughly correct I guess.
fefrie wrote:So what's the easiest best background program to do this?
I know there are others, and tested some, but after having tried NHC I don't really see why one would use other programs, since it's (in my opinion) pretty much a one-stop-shop for clocking (also ATI GPU clocking), volting, fan control, battery management and HD monitoring.
T61 (8895-2FG) T8100/Intel X3100/SXGA+/Intel 320 SSD/4GB/Win7 x64 Pro
T400 (retired)
T40 (semi-retired)

fefrie
Junior Member
Junior Member
Posts: 435
Joined: Wed Jan 24, 2007 3:29 am
Location: Vancouver, Canada

Re: Undervolting. The next thing I'm going to learn about.

#5 Post by fefrie » Tue Dec 13, 2011 2:14 am

I ended up using RM clock simply because I already use TPFancontrol, and I'm a big fan if TPF. It's simple and I don't want to use another program that would replace it.

RM has a bit of a learning curve, and after reading the tutorial, I decided to go for the full undervolt with 0.7v at the minimum and 1.03 at the max.

Since in reality the laptop rarely ever sees max cpu useage, I'll leave the settings as is for now and if I get a BSOD, then I'll tweak the voltage upwards. But again since it can't go any lower than 0.7 I'll b ehappy with it and leave it at that.

I can't be bothered to find the perfect profile. It's fine already as it is.

I'm noticing that the cpu temps at word processing and just leaving static webpages open have the CPU temps at 31c (GPU 33c) which is extremely low and unusual. I'm used to seeing it around 50C, if not around the 45ish zone.

And it's never too late to undervolt. A milliwatt saved is a milliwatt saved, and if I can take out all the heat out of this system as best I can, I can reduce my anxiety of a Motherboard failure due to heat....
IBM Thinkpad T23 1.13 2647-9LU 640MB Ram 40GB hard drive SOLD!
T42 SXGA 1.7 64mb xp

eric99
Posts: 11
Joined: Tue Nov 29, 2011 11:12 am
Location: Wilkes-Barre, PA

Re: Undervolting. The next thing I'm going to learn about.

#6 Post by eric99 » Tue Dec 13, 2011 2:26 am

fefrie wrote:I ended up using RM clock simply because I already use TPFancontrol, and I'm a big fan if TPF. It's simple and I don't want to use another program that would replace it.

RM has a bit of a learning curve, and after reading the tutorial, I decided to go for the full undervolt with 0.7v at the minimum and 1.03 at the max.

Since in reality the laptop rarely ever sees max cpu useage, I'll leave the settings as is for now and if I get a BSOD, then I'll tweak the voltage upwards. But again since it can't go any lower than 0.7 I'll b ehappy with it and leave it at that.

I can't be bothered to find the perfect profile. It's fine already as it is.

I'm noticing that the cpu temps at word processing and just leaving static webpages open have the CPU temps at 31c (GPU 33c) which is extremely low and unusual. I'm used to seeing it around 50C, if not around the 45ish zone.

And it's never too late to undervolt. A milliwatt saved is a milliwatt saved, and if I can take out all the heat out of this system as best I can, I can reduce my anxiety of a Motherboard failure due to heat....
Spooky; I was just about to reply to your original post when your most recent message appeared.

Congrats on figuring out the under-volting tango. Looks like you and I have similar-spec TPs and we both tried the same programs to under-volt. I gave up on RM Clock some time ago, when it did not bond with my desktop Win7 system. To under-volt my T42 I used NHC and took the easy route: I set the lowest level at 0.7 volts---a significant reduction---and set the highest at 1.24, which is the default setting for the second-highest level.

Those are the two most common settings at which my T42 runs---with the throttled-down 600 MHz minimum setting more than any other. I am not a power user, though I run a powerful chess analysis engine at times; that eats CPU but good.

So: Care to share the profile, step by step, that ranges from 0.7 to 1.03? Betcha a stogie you can do it faster and easier through NHC....

Cheers.
T60, 1953-P3U, 1.66 GHz T5500 (Merom) CPU, 3 GB RAM, Intel 950 graphics, 14.1" XGA, 160 GB SATA 5400 rpm hdd, Win 7 Pro SP1 32-bit and Linux Mint 17.3 Mate (x64)

fefrie
Junior Member
Junior Member
Posts: 435
Joined: Wed Jan 24, 2007 3:29 am
Location: Vancouver, Canada

Re: Undervolting. The next thing I'm going to learn about.

#7 Post by fefrie » Tue Dec 13, 2011 2:52 am

I couldn't be bothered to test voltages for a certain multiplier, so I chose the bottom at 0.7 and the top at 1.03 and let RM pick all the ones in between and set them automatically.

I'm getting cpu temps as low as 28c now and GPU temps as low as 30 in idling mode.

My laptop now runs at slightly above room temperature. Doesn't even get warm anymore sitting on top of a table of couch.

Unreal.
IBM Thinkpad T23 1.13 2647-9LU 640MB Ram 40GB hard drive SOLD!
T42 SXGA 1.7 64mb xp

FrankL
Freshman Member
Posts: 59
Joined: Mon Dec 29, 2008 8:28 am

Re: Undervolting. The next thing I'm going to learn about.

#8 Post by FrankL » Tue Dec 13, 2011 6:55 am

You'll come back on that statement when you start experiencing plenty of 'unusual' behaviour.

What I did to assess the minimum voltage required for each speed step:
- run burncpu/cpuburn/prime95/etc. for at least an hour at the set voltage. Prime95 will report if there is an error in the calculation if the cpu voltage is too low (if you don't get a BSOD)
- up the voltage until prime95 no longer report errors
- up the voltage at least 2 notches above that point to have some safety margin (prime95 catches far less than 100% of all cpu instability)

btw. on my MS Windows Vista system, RMClock makes speed stepping much smoother than the default Windows speedstep implementation. If only for that reason it's worth using.

Johan
Moderator1
Moderator1
Posts: 1977
Joined: Mon Mar 07, 2005 2:00 pm
Location: Copenhagen, Denmark

Re: Undervolting. The next thing I'm going to learn about.

#9 Post by Johan » Tue Dec 13, 2011 1:11 pm

... optimizing T4x performance thorugh CPU undervolting and/or GPU downclocking has been discussed in many threads here of the forum, and to those not fully acquainted with all the many benefits of "Notebook Hardware Control", especially its ability to undervolt the CPU, which will reduce power consumption, which will extend battery time and also trigger the fan to go "on" less often, have a look in the thread NHC 2 is out! and, if in a hurry, just review from this post and onwards. Examples of using NHC in T43/p's are discussed many places, e.g. Research w/ NHC on T43p High Temps. The very fine guide by visionviper --> GUIDE: How to max out your battery life *56K warning - Pics* will explain how to set up NHC.

I have been using NHC for years (on T42's and T42p's) under XP, and this has significantly helped cool and quiet down the laptop, and extending the battery time. I cannot emphasize enough what a highly useful program NHC is (on a side note I find NHC much simpler to use compared to e.g. RM Clock). Moreover, I also appreciate that I don't have to use a seperate fan control program, e.g. the Shimodax/Troubadix "ThinkPad Fan Control", since this is already integrated in NHC.

Bottom line... in my opinion: Those who want a more cool and quiet T4x/p, go get NHC and read the above threads about how to set it up! :-)

Johan
IBM T42p's (2373-Q1U & -Q2U): 2.1 GHz, 15" UXGA FlexView, 2 GB RAM, 128 MB FireGL T2, 128 GB 1.8" SATA SSD, IBM a/b/g, BT, Win 7 Ultimate
IBM T42 (2373-N1G): 1.8 GHz, 15" SXGA+ FlexView, 2 GB RAM, 64 MB Radeon 9600, 64 GB 1.8" SATA SSD, IBM a/b/g, BT, Win 7 Ultimate

Raceboy
Senior Member
Senior Member
Posts: 779
Joined: Sun Feb 10, 2008 7:17 am
Location: Tartu, Estonia

Re: Undervolting. The next thing I'm going to learn about.

#10 Post by Raceboy » Tue Dec 20, 2011 3:49 am

Johan wrote:...
Bottom line... in my opinion: Those who want a more cool and quiet T4x/p, go get NHC and read the above threads about how to set it up! :-)

Johan

+1. I've been using NHC on all of my laptops since T40 and X40 and using it on my T60p also under W7 and it works PERFECT. I managed to find a T7200 that works on lowest possible voltage 100% stable on ALL multipliers sans the highest, where it requires slight bump. I bet there's not many people who can say that they have cool and quiet T60p with V5200 graphics :) Or cool and quiet T42p with 1.8 Ghz Dothan and FireGL T2.
X61s:L7500,4GB,128GB SSD,IPS
X32s:PM 758 LV CPU mod,2GB,64GB microSATA SSD,COM mod,IPS
701c,240,380,X60s,560X,570E,600/E,T20,T21,T30,TR451,T42p
Past:560/E/Z,600E,R30,T21,T23,T30,T40,TR451,T40p,T41,T41p,T42,T42p,T43,X20,X22,X23,X24,X31,X40,X41,X60/T,X61/s,X201,T60,T60p,T61,T400,T601p

Post Reply
  • Similar Topics
    Replies
    Views
    Last post

Return to “ThinkPad T4x Series”

Who is online

Users browsing this forum: No registered users and 7 guests