Buy a battery for my lenovo X201t or buy an android tablet.

[Hattori Hanzo]

Maybe I should do a standard video test, one at 100% and one at 5%, to have actual facts?

P.s. As mentioned above... the 5% was copied from the max battery (Or something like that) setting supplied by Lenovo so I'm thinking that they know how to get the maximum life from the battery. (Then again I cheat by tweaking the living heck out of the machine)

Standard video test?
You mean playing a movie and see how long the system lasts?
How's an absolutely linear workload going to be a valid test for general usage with things like OneNote and webbrowsing, where the CPU load's usually very unsteady?
You need something that resembles the actual workload as good as possible. And if you want hard facts you also need to be sure that nothing else interferes with your testing.

The problem with Windows' setting for CPU performance is that what it actually does is pretty much undocumented. Capping the highest performance states of a CPU might in fact be beneficial to efficiency and hence runtime, if they are a trade off towards performance and against efficiency. So what does the percentage actually do?
Is it just affecting available P-states or also changing the behavior of when they are chosen?

The reason that almost all manufacturers limit the CPU as much as they can for battery life optimized settings is not necessarily common sense. If you limit the maximum power consumption severely the machine will be slower whenever the CPU is really needed, but it runs and your waiting is part of the runtime.
What you actually want is the CPU to work in its most efficient range, though, which is not necessarily the lowest P-states. In fact, in most cases it is better to have the CPU clock higher and go to sleep faster.

If you don't disable SpeedStep entirely during very light load and idle all those settings don't make much of a difference anyway. In such a case the CPU consumes only a fraction of the power the system uses, so even doubling or halving that makes up for only a negligible difference in runtime.

[excalibur1814]

"How's an absolutely linear workload going to be a valid test for general usage with things like OneNote and webbrowsing, where the CPU load's usually very unsteady?"

It's not going to be valid when based on a general use scenario BUT IT WILL deduce if placing the cpu max at 5% saves power over 100%. I get what you're saying... but... I still think that when on battery and wanting maximum life it's better to reduce the max cpu. You have your opinion and I have mine For instance the Dell Latitude XT Tablet would simply be choppy when trying to play a video until I raised the max cpu to 80%... not so with an i5 class cpu.

I'd still prefer dave-d (OP) to at least TRY reducing the cpu to 10% to see if it gets him through an average day (8-10 hour day at university). I max my battery by:
- Putting an ssd into the machine (Selective choice as per the charts: )
Charts, benchmarks SSD Charts 2011, Power Requirement at Idle
Charts, benchmarks SSD Charts 2011, Power Requirement: HD Video Playback
- Turning off ALL services that are not required/essential
- Clean install with the best drivers I can find
- Tweak the power setting so everything is set to max battery (Core i3/i5/core 2 - NOT Atom)
- Lower the max cpu to 10% as that is more than enough for day to day tasks like surfing the web and notes)

Done. I then monitor the machine to ENSURE that the CPU stays at 0%. If not I go hunting again.

P.s. As long as a machine returns to idle fast enough I'd rather limit the CPU than have it race ahead at x Ghz eating battery life for no reason.

[Hattori Hanzo]

It's not going to be valid when based on a general use scenario BUT IT WILL deduce if placing the cpu max at 5% saves power over 100%.

Nope. It will prove that running that DVD with that player is better on this or that setting.
I'd still be interested in the outcome, though. Please add a test at a 50% setting or test at 50% instead of 100%. Could you monitor the chosen P-states and their transitions at both settings? That's going to show a bit more of how that setting works in the first place.

I'd still prefer dave-d (OP) to at least TRY reducing the cpu to 10% to see if it gets him through an average day (8-10 hour day at university).

Can't hurt. But again I'd be very interested if 5% also beats 50%.

I then monitor the machine to ENSURE that the CPU stays at 0%. If not I go hunting again.

Here I'm absolutely with you. Minimizing unnecessary background activity is the most crucial step. If you don't wake the CPU in the first place, there's no arguing that it saves power

P.s. As long as a machine returns to idle fast enough I'd rather limit the CPU than have it race ahead at x Ghz eating battery life for no reason.

You have not understood the concept of "rush to sleep".
And I'm fine with the CPU using a little more power for a short time if it saves twice that by sleeping longer and also enabling the chipset (or now its peripheral controller parts) to sleep.

[excalibur1814]

You have not understood the concept of "rush to sleep".

I do get what you're saying... I do... Just don't believe that we need a core i3/i5 or even high Ghz Core 2 to do general day to day stuff.

As soon as Insight decide to send the 4-cell battery I'll get onto testing (Media played from hard drive) at 5%, 50% and 100%. Might even create a batch file to open and close ie9/opera/one note etc for a few hours)

[dave-d]

hi thanks for the feedback. just going to let you know what ive tried. and what im trying based on your recommendations

Putting an ssd into the machine

ive been running an ssd for months. replaced the old mechanical hard drive when it broke.

Turning off ALL services that are not required/essential

im looking at msconfig now. not really sure what services to get rid of. might have to just use trial and error. any ideas?

Clean install with the best drivers I can find

my last clean install was about a month ago. Im running mainly the windows and lenovo drivers.

Tweak the power setting so everything is set to max battery

done

Lower the max cpu to 10% as that is more than enough for day to day tasks like surfing the web and notes)

I had it currently on 20%. ive just lowered it to 5%. will let you know if i see improvements.

without disabling any services cpu usage is stable at about 3% . a good sign?

"bit off topic" but does anyone know if the panning hand in one note can be bound to my digitizer buttons? im thinking of picking up a new digitizer for that purpose but i have a feeling that the panning option in the wacom driver isnt the same thing that im looking for.

[Hattori Hanzo]

I do get what you're saying... I do... Just don't believe that we need a core i3/i5 or even high Ghz Core 2 to do general day to day stuff.

That's why I think you didn't get it.
You surely do not need an i3 for office and browsing. That's besides the point, though.
The thing is that efficiency and clock speed are not in an easy and direct relation. Just fixing the clock of a modern CPU at its lowest operating point is a fundamentally bad idea. With the modern power gating you have a basic leakage current once it has a clock signal. It also needs power to keep its cache running. Once you access memory, the memory controller needs to be awake and hence consumes more power.

If you power up the CPU to a rather high clock speed, you can do work fast and shut off all that stuff after a very short time, minimizing the time you have leakage currents and non sleep states.
Imagine it like bringing groceries down your basement. You can bring down each potato piece by piece and walk 20 times or just pack yourself up with three bags and go once. In the latter case you omit dragging your own weight up and down the stairs all the time.

If you have a CPU running at a low clock speed the ratio between the base power consumption to have it running and all peripherals active and the actual power added for calculating stuff is pretty bad. If you raise the clock you surely consume more power, but can increase actual efficiency.

I've seen a lot of people shooting their own foot when overclocking or undervolting their CPUs, because they forced the CPU to low power states, keeping it awake and active all the time. In the end their undervolting didn't do much outside of torture tests or their overclocking made the system not any bit more responsive although benchmarks where better.



Edit:
3% is 3% is too much imo. Get Process Explorer and look where those 3% come from if the task manager is not giving it away easily
http://technet.microsoft.com/en-us/s...rnals/bb896653
PE is way more powerful than the built in task manager, way more accurate in displaying CPU usage and provides much more details about where actually the processing time gets sunk.
An occasional spike in CPU usage is ok, but it should be 0% most of the time or caused by something you want to keep.

[excalibur1814]

@dave-d

MSCONFIG: On my x220 (Which is obviously a different machine) Ticked
-Intel Common user interface
-Smart Audio Control panel application (Probably don't need this on)
-Microsoft security essentials

That's all.

Services on this machine here (X220): There are consequences for disabling a few of these but nothing that cannot be enabled again.

[excalibur1814]

With luck you'll get your pc running at 0%

[dave-d]

cheers for the list. working on implementing it. out of curiosity how did you get a service description in msconfig?

[excalibur1814]

Snipping tool