T500 Undervolt not "sticking" ?

So I am using CPUgenie v1.5 (trial version if that matters)
with only manual voltage undervolting (NO cpu power management enabled in CPUgenie)

P8700 & 8GB DDR3 1066

These are the following voltages that I have manually found to be stable and working using CPUgenie's built-in stress tester.

(Current V) (Stock V)
6 (Super LFM) 0.875V 0.875V
6 0.912V 1.000V
6.5-9.5 1.025V 1.163V

Sometimes I still observe the CPU pushing at 1.163V at 9.5 multiplier, which I observe through CPU-Z as well as the Power Monitor in CPUgenie.

Why does it not stick to the 1.025V that I set it at?

 

The difference in what you actually set the voltage as and what CPU-Z reads it as is the result of vdroop.

By any chance, are you running CPU Genie while you have CPU-Z running? I use RMClock on my Vostro and if you don't hit apply and run when Windows starts up, none of the voltages "stick".

 

The Windows 7 - Control Panel - Power Options - Power Plan might be interfering with CPU Genie. It has nothing to do with vdroop.

The 45nm Core 2 mobile CPUs work differently than the previous versions. Check out the pictures in this post.

http://forum.notebookreview.com/6369641-post5060.html

The 45nm CPUs are designed to drop down to the deeper low powered sleep states like C3 at idle. They will use a lower VID internally that can't be controlled by CPU Genie or RM Clock, etc. so measure battery power consumption using the Windows Performance Monitor when you are done.

 

But I am picking up the non-sticking readings in both CPU-Z AND Power Monitor in CPUGenie. And yes...I have them both running at the same time, as well as hwmonitor and performance tab of windows task manager.

 

Those readings really are happening. That shows that your Windows power settings are interfering with the CPU Genie settings.

 

Hmm yes you are probably right. I just enabled "CPU Power Management" in CPUgenie and now the voltages and respective multipliers show the correct voltages, with 1.025V max @ x9.5 as I have set it to.

Now, I will look into your suggestion regarding Intel Dynamic Acceleration (IDA). So it seems like battery life is negligible so your reasoning is that its better to just have it on "high" all the time.


EDIT: Hmmm...just noticed something funny. CPU-Z now shows a multiplier of x3.0 at the lowest/idle state ....I thought x6.0 (Super LFM) was the lowest (pretty sure it only hit x6.0 lowest before the undervolt...but oh well. I don't notice a difference as the Core V is still 0.875V @ speed of 798Mhz, just like before.

 

CPU-Z does not correctly report SLFM. Inside the CPU, SLFM cuts the bus speed in half but for convenience, CPU-Z reports the LFM bus speed but cuts the multiplier in half instead.

CPU Genie does not support IDA mode on my T8100 so performance immediately decreases by 200 MHz which is almost 10% when running single threaded tasks. I knew there was something about that program I didn't like. The P8700 supports a 10.0 IDA multiplier so if that is not available in CPU Genie, it will only cost you 5% in performance.

There is also a second problem. It doesn't let the CPU get up to full speed quick enough in some situations so when running single threaded apps, it might end up running them with the CPU at about 50% of its maximum speed which just kills performance.

Edit: The RealTemp XS Bench is a great test to show this problem.





The Extreme High Performance setting in CPU Genie is not that Extreme. The best way to save power is to let a CPU run fast so it can get its work done quickly and then it can spend a higher percentage of time in the deeper C3/C4/C6 sleep states where the real power savings are. Running a CPU at a fraction of its rated speed is inefficient. It wastes power and will make a laptop feel sluggish.

I think you need to set Windows 7 to the High Performance profile so it doesn't interfere with CPU Genie. Also make sure that the Minimum processor state in Windows is set to 100%. That will let CPU Genie manage things for you.

RM Clock can still be used on some of the newer 45nm Core 2 Duo CPUs. Sometimes all it takes is a simple registry modification so it supports the half multipliers better.

http://img89.imageshack.us/img89/6684/rmclock45nm.png

Obviously I'm biased towards ThrottleStop but I'd pick RM Clock ahead of CPU Genie any day. RMC works OK in Windows 7 as long as you set the Minimum processor state to 100%.

 

(3) Yes...min. processor state in Win is already set to 100%. I switched profile from Balanced to High Performance...didn't seem to make a difference because at this point I believe CPUgenie takes control of power management as I enabled "CPU Power Management" in CPUgenie.

(2) Yes, please do...It is actually interesting that you say this because there have been some rare occasions where I feel that it is underperforming or "stuck" on a low performing multiplier...a Restart/Shutdown seems to solve the problem. (rare occasion)

(1) Nope...seems like CPUgenie maxs my P8700 out at x9.5. However, it seems that CPUgenie is the only utility that properly addresses cpus with .5 multipliers (half step) such as mine. RM clock & CrystalCPUID I believe don't handle that as well as they've been reported to be unstable?


Also...don't think my BIOS supports dual IDA either..

 

With everything running...

http://i53.tinypic.com/20r7l09.jpg

 

X86 cpus' Guide - View details on Intel Core 2 Duo P8700 BGA

The above site shows that a P8700 has an IDA multiplier and can run single threaded tasks at 10.0 x 266 = 2666 MHz.l

When you use the IDA multiplier, your CPU is running faster so it's not unusual for it to need more core voltage. When partially loaded, these CPUs are in C3/C6 so that's not really a big deal.

If you want to see your CPU using single IDA mode where it speeds up a single core; reboot and don't run CPU Genie. Start up ThrottleStop but leave it in monitoring mode only. Run a single thread of HyperPI.

Hyper Pi - The Super Pi Front-End

This program locks SuperPi mod to a single core so the other core can spend most of its time asleep This allows the core running SuperPI to use the IDA multiplier the majority of the time. ThrottleStop makes this very easy to see while most other software does not report this correctly. Whenever the second core needs to wake up to process some background tasks, IDA mode ends but when that task is completed, the second core goes back to sleep so the first core can re-enter IDA mode. It's free performance.

HWMonitor is probably using the incorrect TJMax value for your CPU. Intel says it should be using 105C so you might need to edit the hwmonitorw.ini config file.

http://ark.intel.com/Product.aspx?id=37006

CPU_0_TJMAX=105.0

Edit: Can you post thumbnails of your screen shots? That will help keep the mods in a good mood. Not everyone has a wide screen.



ThrottleStop is one of the few tools that can accurately monitor IDA mode.

 

Results:

**CPUgenie**

http://i55.tinypic.com/16kw12o.jpg


**ThrottleStop**

http://i52.tinypic.com/14ieete.jpg

*TinyPic doesn't provide any thumbnails


So yea...from the results I can tell the difference. ThrottleStop does produce better performance allowing my P8700 to hit its x10 multiplier. However, as you mentioned, the performance increase is a small percentage. I don't really like the fact that my cpu is running at x10 @ 2.66ghz constantly. I notice the increase in CPU whine as well as just the fact that I'm stressing it more.

EDIT: My P8700 hits the wall at lowest vid possible of 1.0500 V using ThrottleStop set Multiplier at 10.0

I find that both programs are pretty different and unique in what they provide.
CPUgenie is more of a set it and forget it program...which I can purely use to undervolt the respective multiplier levels of my CPU.
On the other hand, ThrottleStop gives you the manual ability to switch your CPU into max performance mode constantly, which is useful when I want to squeeze every ounce of performance out of my system, or force a battery saving mode to squeeze every ounce of battery life.

So, two different programs.

 

Have you tried the ThrottleStop Power Saver option or the C States option? Either of those will reduce the multiplier at idle. If you don't want to use IDA mode then you can also check Disable Turbo. You can also lower the maximum VID voltage.



At idle, these CPUs spend over 98% of their time in the C3 state or a deeper sleep state. When running like this, the CPU uses voltages lower than you can select or monitor with any software. Here are the Intel specs:

C4 - VID = 0.650-0.850 V
DC4 - VID = 0.600-0.850 V
C6 - VID = 0.35-0.70 V

 

Yes...Power Saver option helps but it seems like it jumps from idle @ x3.0 straight to x10.0 with the slightest cpu activity (ie. such as moving my mouse around). Seems to be very sensitive. At least this is what I observe according to CPU-Z.

BTW: 8 GB of DDR3 so I disabled my paging file...not sure if this plays a role.

 

In the Options window you can adjust the sensitivity of the Power Saver feature by increasing the PowerSaver C0% number. The intermediate multipliers can make a CPU run slower but they don't do anything to reduce power consumption so ThrottleStop avoids using them.

Here's an example. Let's say you have a task that is going to take 60 minutes to complete if you let your CPU run at the 6.0 multiplier. I'm not sure if your CPU will be 100% stable at this speed with the lowest possible VID voltage but best case scenario, lets say it is and can run this speed with VID=0.875

If you were to use the 10.0 multiplier instead, you would need a voltage VID=1.05. The advantage would be that it would take 60% (6.0/10.0) as many full load CPU cycles so worst case, the other 40% of the time it could drop down to the C6 sleep state where it uses a VID=0.5 approximately.

First Example:

60 minutes X 0.875 VID = 52.5

Second Example:

( 36 minutes X 1.05 VID ) + ( 24 minutes X 0.50 VID ) = 49.8

That is just a very rough example using some actual numbers from your CPU but it helps to show that a slow CPU might not be saving you anything. The second example consumes slightly less power, not more. You could also turn your computer off after 36 minutes and save some more power since it has already finished the job it was working on. A small increase in VID voltage so your CPU can run faster and spend more time in C6 or be turned off is a win / win situation. The VID number that CPU Genie or ThrottleStop or RM Clock lets you select isn't used at idle when the CPU is in the C3/C6 sleep state so raising this slightly in those programs doesn't have the negative impact that you might think it would.

I did some real world testing like that with my T8100 and compared actual battery power consumption data at all the different multipliers and didn't see any savings when trying to run the CPU at half or three-quarters of its rated speed. A slower CPU can give you the illusion that you are saving power but when you look at overall power consumption, the savings are much harder to find because of C3/C6.

If you run ThrottleStop in monitoring mode and let Windows manage the multiplier, you will see that Windows 7 gets the CPU up to full speed much quicker when lightly loaded compared to previous operating systems like XP. Why? I guess Microsoft finally did some testing and discovered that a slow CPU is not an efficient CPU.

When lightly loaded, SpeedStep transitions can be happening hundreds of times a second within a CPU. Programs like CPU-Z sample the multiplier once per second and do not accurately report what the CPU is really up to inside. CPU-Z is also designed for consistent validation purposes so what it shows at idle or when lightly loaded may not accurately reflect what speed the CPU is really running at internally. The average multiplier that ThrottleStop shows is an extremely accurate measurement of what multiplier your CPU is really using. ThrottleStop reads high performance timers within the CPU and the method that Intel recommends in their November 2008 Turbo White Paper so all of the SpeedStep transitions are properly accounted for.