Have you ever noticed a sudden drop in the battery life of your laptop, and suspect it’s the result of the application you installed last week?

Have you ever wondered whether your application satisfies that 2-hour while-on-battery runtime requirement?

Have you ever tried to use Perfmon to gather power related statistics and given up after suffering serious information overload?

PowerInformer is a tool developed to provide basic power relevant statistics to a developer. They can use these statistics to optimize their application such that it matches battery life constraints (e.g. the game must run for at least 90 minutes on a notebook with a 56 W hr battery) while also meeting performance requirements.

Some of the main features include calculating the average residency and the percentage time of the C1, C2 and C3 states of all logical processors.  Also, the screen layout has been redesigned to accommodate more data and to make it look better.  Finally, important data such as disk and file I/O operations have been added to provide more information about the system power consumption.

PowerInformer Features (full list):

• Battery and power status indicators
• Processor (physical, core and logical) detection
• Support for both Windows* XP and Vista*
• Percentage time of C1, C2 and C3 states of the system and of all logical processors
• Average residency of C1, C2 and C3 states of the system and of all logical processors
• Percentage time of the system and of all logical processors
• Percentage time of the P stages of the system
• System calls per second
• Interrupt rate of the system
• Disk I/O operation (access, read and write) rates
• File I/O operation (control, read and write) rates

What is in the PowerInformer Package:

• The application
• User’s guide
• Release Notes
• End user licenses

Comments (12)

December 30, 2008 6:55 PM PST johnsonzhang Total Points: 10 Registered User

i install this tool on my laptop, when running this tool, the error information tells me that "add Pdh counter (% of maximum frequency) failed with -1073738823

December 30, 2008 7:31 PM PST johnsonzhang Total Points: 10 Registered User

i find out the root cause, that is the tool only support English OS, my OS is Chinese version.

February 25, 2009 10:49 PM PST Neftali Reyes

Looking for a tool that neasures C3 states. I wrote my own, but it differs greatly in results when compare to original Intel C3Resmon utility. Seeking to see if my utility was right all along.

February 26, 2009 10:19 AM PST Highlander20 Total Points: 10 Registered User

I also got the error message the previous guy posted "add Pdh counter (% of maximum frequency) failed with -1073738823. However, My system OS is english. Does anyone have any answers to this?

February 27, 2009 1:24 AM PST Allen_W

I meet this issue too on my side, OS: EFI Vista 64 bit English

February 27, 2009 1:41 AM PST allen.w Total Points: 5 Registered User

Please tell me if you guys have fixed this issue and thank you in advance!

February 27, 2009 7:28 PM PST Highlander20 Total Points: 10 Registered User

I think the developers of this tool have been laidoff form Intel reason why there is no answer. Too bad, I thought this was a very nifty tool. I guess I will continue to develop my own.

March 6, 2009 4:10 PM PST Taylor

Unfortunately, due to changing priorities (now where have I heard that phrase before) the support of this tool has been sidelined for the foreseeable future. If I recall, the tool was developed for XP. I think that the original developer had some problems with Vista. "Pdh" commands are the APIs that apps (such as Perfmon and PowerInformer) use to get at OS provided counters. The problem here is two fold: (1) MS likes to change the names of the counters from one OS generation to the next, and (2) MS is attempting to deprecate the pdh commands (as I understand it) in favor of a newer performance monitoring architecture within Vista. Here's a side note: the pdh APIs, and so perfmon and PowerInformer, provide performance values obtained from the OS. This means that they are only roughly the same as their HW equivalents. For example, the information about C3 is an approximate aggregate of all lower C-states. As another example, the OS talks about "Logical processors", not physical processors.

March 25, 2009 5:02 AM PDT wsung Total Points: 5 Registered User

In order to test CPU C3 state

March 28, 2009 11:45 AM PDT neff

Tailor: I read your other blog about logical vs core, etc.. Very good article. I'm developing a similar tool for approximations only so the Performance Monitor results are good enough for me. Just in case you didn't see this question in your other blog, I figure that the formula used for power Informer is (C3 Time / C3transitions) X 100 = C3 Residency. Is this close enough??

March 30, 2009 11:36 AM PDT taylor

It's SUM{ (dSamplePercentIdle[i]/100) * (dSamplePercentCn[i]/100) }*100 where the SUM is over the delta period that you're looking at, e.g. 1 sec. The pdh values are obtained from PDH_FUNCTION PdhGetFormattedCounterValue( PDH_COUNTER, DWORD, LPDWORD, PPDH_FMT_COUNTERVALUE). See MS's pdh.h. The counter name (at least in XP) is "\Processor\% C3 Time". (Remember that "\" is an escape command so you need "\\".) Double check this as the counter name can be picky. If you're on Vista, the name could very well be different. The way you use the pdh commands seem a little strange until it sinks in. You add the counter first, then at some later point, you sample it. Basically, you're trying to minimize overhead. So you tell the OS what counters you want to sample before you actually issue a get counter command. To make this even more accurate, some of the pdh commands are "array", meaning that you collectively sample several counters at once. I know I'm being boringly pedantic about this but remember that the HW counters are not the same as the pdh counters.

November 18, 2009 10:45 PM PST Hamin

Hi all, For test c-states, 1.we can use PowerInformer from Intel or Perfmon.msc /s from OS. When I used PowerInformer, Win Vista Business SP2 show me "add Pdh counter (% of maximum frequency) failed with -1073738824". What's the problem? 2.How to test C4(E) and C6? Does Intel have any tools for using? 3.In Win7, are there any tools for testing c-states?

Trackbacks (0)

Leave a comment