The ThrottleStop Guide

10 characters..

 

Please carefully read the following from Unclewebb's ss.



The C-states used by Windows and Linux are NOT the same C-states used by the processor. Windows typically requests up to 3 idle states C1, C2 and C3. For instance when requesting C3 from Windows that may ultimately map to the processor C6 state. IMHO it's a bit unfortunate that the OS idle states were called C1,C2 & C3. Perhaps if they had been called something like I1,I2 & I3 they would be less ambiguous.


Here's a screen shot where I have forced the CPU C3 state on 3 of 4 cores.

​

You can see there is little activity on core APIC IDs 2, 4 & 6. This doesn't show up in windows performance monitor because Windows is only showing its requested states and not actual states. Also note the "task manager" is showing 80% usage while the CPU is less than 12% active. Many people believe task manager shows CPU work load but it actually shows a scheduling load. Most of the time the scheduling load and actual CPU work load can be fairly close but it is better IMHO if you use something like the C0% shown in TS to get a truer idea of what the CPU is doing rather than what windows is doing.

 

I'm sure you wont regret the purchase! However, that does seem a steep price for one, I know they're becoming rare, but still.

I found mine from an electronics compant who were getting rid of about 100 of them. They didn't have a clue about the potential of these CPUs, and just thought they were 2.33ghz, without an unlocked multiplier.

Just keep looking around, I'm sure you will find cheaper ones around and about.

Good luck

Matt

 

Hey Dufus

Thanks for clearing that up.
You should know Unclewebb has verified the same wandering multiplier at idle just as I have reported and discussed with you in other threads before.


Unclewebb

One more thing if now I'm not disturbing.
I can't get my CPU to respond either to ThrottleStop multiplier setting or Power Management minimum and maximum states at idle. Multiplier keeps changing as seen here.

Could you still verify that with any computer of yours?

Just as soon as any process takes place, either ThrottleStop multiplier setting or Power Management controls are funcitonal.


By the way if not a bug then why no wakeups from C3 and the steady multiplier? What can C3 do that C6 can't?
Why does C6 transition to C0 instead of demoting to C3 if C3 can perform as needed?

 

What are our options to get ThrottleStop to load with Windows during start up either during login or better yet, before user login? I am not skilled in Registry editing but willing to learn if with detailed quality tutorials.

 

I just use the:

Task Scheduler Method

or you can drag a link to ThrottleStop into your Startup folder. Click on the Windows logo and it should be easy to find that folder.

If you use UAC, you need to use the first method.

 

You can put a shortcut to throttlestop in the startup folder.

EDIT: I'm too slow.

 

I did the create a start up short cut but it would not start, even if I put the short cut in the folder manually (C:\ProgramData\Microsoft\Windows\Start Menu\Programs\Startup) Here I have other apps that start such as my Bluetooth connection manager.

I got it to work with the task schedul method but I set it different so I had no control of it as mentioned in that linked tread. I corrected it and it now loads in task bar so I can change as desired or I can set it to load minimized .

I do want it to load as Windows starts no matter who logs in or not, but I want to have control of it in task bar. Is there a work around to this in Windows 7 x64?

 

What does checking the start minimized box in the ThrottleStop options menu do?

 

When you start TS it will go straight to the taskbar and not show you the TS window.

 

lol, I thought you'd forgotten about this. IMHO not really worth losing sleep over but if you remember the test program I sent...



What that program did IIRC is set the lowest multiplier in MSRs 0x199 (IIRC that was 7x), turned off EIST via MSR 0x1A0 then verified the x7 multi in MSR 0x198 before monitoring performance. The Multi's shown for Core 0-3 are the calculated multi's from the performance counters. However the interesting one is shown in the Title Bar next to the Dufus Icon which shows the current MSR 0x198 status multi which is 11x. Therefore something has changed the 7x multi to 11x even though EIST was disabled.

You could try investigating if your power management drivers are doing this or check what microcode update you are using and see if there is a later one or perhaps even an earlier one to try. Other than that you might have to just resign yourself to the 11x multi being set and rely on the higher c-states to do their power saving job which they generally do very well.

 

Lappie: Unfortunately, ThrottleStop can only be run and controlled from a single user account. If it is running in one account, it should control the CPU for all accounts but you would only be able to adjust it from the one account. I'll put more multi-user functionality on the things to do list but it might be a while before this happens.

 

Great job onwhat you did so far. Cool app!!!

 

Great app unclewebb--thanks!

ThrottleStop is running superbly on my T9300-based M1330. I've configured things to work together with RMclock, for finer control of power options.

My configurations is as follows:
* ThrottleStop
-- Options: "AC profile = 1", "Battery profile = 4", "battery monitoring", "battery monitoring", "start minimized", "minimize on close", "unlock bidirectional PROCHOT", "Nvidia GPU", "start dual IDA", "profile 1"
-- Profile 1: "set multiplier = 13.5", "VID = 1.0750", "slfm6"
-- Profile 4: "slfm", "EIST", "C states"

* RMClock:
-- Profiles: "ac power = no management", "battery = performance on demand"
-- No management, OS settings: "use OS power plan = high performance"
-- Performance on demand, OS settings: "use OS power plan = power saver"
-- Performance on demand, CPU settings, battery: "PST = 1 (12x @ .9500v), 6 (11x @ .9500v)"

As I mentioned above, things have been running beautifully. For my purposes, the only thing that would make this better are the ability to automatically remember/reapply settings temporarily disabled when enabling Dual IDA.

I'd like to enable DIDA, yes, but I'd also like to enable any DIDA-compatible power saving options, when possible. I've tried using the alarm settings to switch to Profile2 (w/ c states) once DIDA is enabled, but TS then sometimes switches to Profile2 while still trying to enable DIDA, which then disables c-states again.

Any thoughts on this?

EDIT:

Setting "OS powerplan = balanced (min cpu = 100%)" seems to help a bit, though DIDA takes longer to lock. I'm getting temps 4-6C lower than before, which I'm assuming equates to less power draw.

 

AruOjisan: Can you read this thread in the IBM forum? Especially my first post in that thread.

http://forum.notebookreview.com/lenovo-ibm/556473-t500-undervolt-not-sticking.html

I just put a few posts in that thread that I think will answer some of your questions and will help explain why some of the "power saving" settings, really don't save that much power when you are using a 45nm Core 2 mobile CPU. At idle, I can run Dual IDA with the highest VID voltage and power consumption is identical to when I run the lowest multiplier and lowest VID. It doesn't make too much sense until you realize that these CPUs enter the C3/C4/C6 sleep states when idle so most of the time, they don't use the VID settings that ThrottleStop and RM Clock let you adjust. At idle, the cores spend ~98% of their time in a sleep state where they are using a much lower VID internally so the VID reported by software is also meaningless.

You can run ThrottleStop and RM Clock together like you are doing but I don't think you are any further ahead. The tiny power savings, if any exist, are usually more than offset by the extra load you are putting on the CPU by running two separate programs.

With ThrottleStop, after Dual IDA is enabled, it's possible to click on the C States button which makes the multiplier wander randomly up and down at idle but doesn't save any significant amount of power for the same reason as above. If you decide to do this, ThrottleStop sees that the average multiplier is less than the maximum multiplier so thinks that Dual IDA is no longer enabled even though it still is. I haven't found a way around this so if you do this, you will have to manually manage the CPU.

If you switch profiles; when you go back to the Dual IDA profile, ThrottleStop will automatically disable C States so it can get back into Dual IDA. You will have to manually enable the C State option after this.

There aren't a lot of options for you when using Dual IDA. In my testing, you don't really need them. I've never found that power consumption was significantly higher when using Dual IDA. When you are testing, run a ThrottleStop log file while on battery power so you can have a look in the log at the PWR_mw column which keeps track of actual battery power consumption.

 

If you wanted the best AC performance AND most power savings on DC, then consider adding "ExitTime=10" into Throttlestop.ini, meaning it will setup your AC or DC profile then exit after 10 seconds. Then add your full path to throttlestop in setpll's resume-setpll.vbs AC_script and DC_script variables so that upon startup, resume-from-standby/hibernate OR whenever you do a AC<->DC or DC<->AC switch, Throttlestop will run, setup your CPU then exit after 5 seconds. I found Throttlestop added anywhere between 0.4-0.8W monitoring overhead otherwise.

unclewebb may consider adding this script to his package??

 

Ok guys this is weird... I installed my new T7600G today. I opened up ThrottleStop and it still won't allow any clock adjustment past 14. It's as if I didn't even change the CPU... it's the same as having my old T7600. The new CPU was clearly physically marked T7600G and CPU-ID shows it as being a T7600G.

What do you think?

 

DeeRock: As always, post a screen shot of CPU-Z and ThrottleStop so I can have a look.

When you install a T7600G, two new buttons should automatically show up on the main ThrottleStop screen. UNLK stands for Unlock and you need to push that one to unlock the maximum multiplier. Beside that one is supposed to be a Reset button which lets you reset your CPU back to its default specs. If these two new options are not showing up then that's a problem and I'll tell you how to fix it. Make sure you are using the most recent version of ThrottleStop from my sig.

Edit: Here are the magic buttons that should show up if you have a Core 2 Extreme or T7600G unlocked CPU.

 

Ok, here it is, thank alot!

 

Ok, here it is, thanks a lot!

 

As mentioned, try the version of ThrottleStop in my sig. There was a bug related to this that I recently fixed but I've kind of lost track when I fixed it.

 

I think it was 2.99.6 which had this bug fix. I personally just added "Extreme=1" to the throttlestop.ini, which I read in another post of yours on notebookforums.

Hopefully this will work out right.

Matt

 

It's working now. I downloaded the newest version of TS (v2.99.6) like you said and the UNLK and RESET buttons showed up. I experimented for a while and the fastest stable configuration I could come up with is 18.0 clocks / 2.99 Ghz / 1.3000 volts which is GREAT!!! You're awesome!

Thanks again for your patience!

 

That's great news! What temps are you getting under full load?

Matt

 

are there any known issues between ThrottleStop and SIS Chipsets?

I tried ThrottleStop 2.89 and 2.99.6 on a clean Vista installation and both make my laptop freeze as soon as the application shows on the screen.

 

naton: I've only tested ThrottleStop on my two computers which both have an Intel chipset. I haven't heard of any problems when it is running on an Nvidia chipset but I've had zero feedback from anyone with a SiS chipset so there very well could be a problem specific to that chipset.

Try adding this option to the ThrottleStop.ini configuration file.

NoChipset=1

This will bypass the chipset code in ThrottleStop so let me know if that solves your freezing problem.

 

Thanks for the tip. I'll try it and let you know how it goes.

 

I haven't been around for a while, but I wanted to stop by and say thank you unclewebb. I just got rid of 2.85 and started configuring 2.99.6, and all is going well thus far. I really like the built in benchmark. Now I don't need another program to test if TS is working the way I had intended.

Keep up the great work!

 

I got a high temp of 78 degrees, but it seemed to be hovering around 70. I was doing some audio today and it locked up so I backed it off to 17 clocks @ 2.8 Ghz / 1.250 volts. I'm not sure if the lockup was due to the overclocking. If it will hold at this setting I'll be satisfied. Thanks for asking!

 

@UncleWebb Have you released 3.00 yet? I saw the picture on the front page and thought I may have missed something.

 

Not quite yet. I did the easy part first. Someone complained that the first post needed updating so I thought it would be best to go a little into the future.

There's still a couple of minor things I'm working on. When it's official, my sig will change to 3.00 so it will be easy to find. Not enough time at the moment to work on this project as much as I would like. The kids are off school this week and want my computer to play games on.

Sirhcz0r: You're welcome.

 

Hey unclewebb... i got a few questions for you

First: if TS is on, and i restart/shutdown and reboot... are the effects of TS still on? even though the app isn't running? I remember reading the answer to this question but i recently became confused and wanted to make sure... sorry for asking... yet again.

Second: about the C-states button that when you click on it swtches from C1, C3, C6, C7, and MAX ...

What does this do exactly? Does leaving it on C1 prevent the cpu from going into any other state rather than active? and then c3 would let it only go as far as that state? and c6 as far as that state? ......and max would mean it would allow cores to go into the deepest sleep states? Is this what it does?

From my own observations this seems to make sense, because when c1 is selected... C0% is the only one showing percentages as the rest are 0.0% ... and then only once i switch it to c3, c6,c7,max it would seem that the c-states go into a deeper and deeper sleep. (notice the percentages getting higher in C3% and C6% and lower in C0% as i toggles through C1,c3,c6,c7,max in the attaches screenshots)

If this in fact what it does and what is happening... would it mean that leaving this setting at c1 would mean that the cores never go to sleep and are readily available to jump into action, therefore reacting quicker and translating into slightly better performance?

Would greatly appreciate it if you can clear this up for me. Thanks bro, and for the record... TS is simply the best software EVER. Your Genius. Thanks a mil for all your time and effort putting TS together and perfecting it along the way. Been a fan since the early M15x + 260m throttling days.... now i just love the overclocking capabilities for my 920xm

 

You're right. ThrottleStop is a 920XM's best friend.

ThrottleStop needs to be started at least once after booting up or else it won't be doing anything to your CPU. On a 920XM, ThrottleStop writes the TRL and TPL information directly to your CPU so those values are usually maintained even after you exit ThrottleStop until the next time you reboot.

Some laptops like the M11x-R2 automatically reset the TPL values when ThrottleStop exits but I'm pretty sure the M17x-R2 maintains them. If you do some wPrime testing with ThrottleStop active or turned off, it should be obvious based on your times whether the TPL settings are maintained in the CPU or not. As soon as you reboot, all settings should go back to their default values.

As for the C1/C3/C6/C7/MAX button, your understanding of it is about as good as my understanding.

What ThrottleStop reports for C3% and C6% refers to the entire package state. That button lets you block the entire package from going into C3 or C6 but the individual cores / threads can still use these sleep states. Of all the ThrottleStop features, I'll admit that this might not be very useful for the majority of users.

When testing, one user with a mechanical hard drive felt that the C1 setting made things feel much snappier for the reasons you mentioned. If you can keep a CPU in the C1 state, then it is instantly ready for action. A CPU sleeping in the C6 state has to wake up first before being ready to work.

When hard drive benchmark testing with an SSD, I did not see any measurable difference in performance. I'll have to wait for some feedback to see if anyone else can confirm any sort of performance difference, even if they can't prove it with a benchmark test. Most benchmarks bring a CPU out of C6 at the start of the test and spend most of the benchmark time out of C6 so any improvement will likely be difficult to measure.

When adding this feature, I was testing on a desktop Core i5-650. The C1 setting resulted in a power consumption increase at the wall at idle of about 12% compared to when that button is set to C6/C7. If you have a laptop running on battery power, that wouldn't be a good idea but if you are plugged in, maybe the C1 setting will help something. The MAX setting is similar to C6/C7 but tries to convert C3 requests to C6 so I was usually seeing 0% in C3 when set to MAX.

I'm still playing around with this new setting. My friend Alex has done some testing and with his help, I've found a way to increase the C6% package state number considerably at idle, hopefully without hurting performance. Most users don't run ThrottleStop while on battery power but if they do, this new feature might be useful and could improve battery run time.

Thanks for the feedback (((STEREO))). It's always good to hear from happy users.

 

Man thanks for that reply dude... really appreciate it!

Ok this is particularly interesting to me... because ssds have a serious problem with the HM55/PM55 chipsets (the ones the r2 uses) and don't get their full speed potential, supposedly because of the c-states putting the cores into deep sleep at idle... which is why initiating a light cpu load (like playing music on itunes or something) and disabling core-parking helps

I'm going to test quick quick if leaving this setting in c1 also helps at all

 

This new feature was designed with that problem in mind but it didn't show any benefit. To get full turbo boost out of a Core i CPU, cores and threads need to be able to go into the C3/C6 sleep states. If you block that, no more full turbo boost. You can block the package C3/C6 states but that doesn't solve the SSD issue. Oh well, I gave it my best shot.

I know the deep C states get the blame for this problem but I think it could also be because the CPU can be using a much lower multiplier when lightly loaded compared to the maximum turbo multiplier that is possible when a single core is active.

 

It works... thanks unclewebb

 

How's this for first go? My CPU is OEM from dell not sure why showing in wPrime as ES??

 

I found small bug at the interface ... insignificant though.

I've got ThrotteStop (v2.99) to load in the startup folder and to start minimized. The first time I open it from the task bar icon to the right (XP here) - the button for the DTS shows as DTS when showing DTS temperatures in the window, and it should show Temp on the button while showing DTS temperatures, letting the user know that when pressed it would show Temp in the window.

though when I press it at least once it fixes itself, i.e. the button shows Temp when showing DTS temps in the monitoring window, and vice verse.



also, can throttlestop be made to load with registry mod as opposed to being in the startup folder ?

 

Thanks for the bug report. I've actually fixed that one in 2.99.7 but I haven't released it yet. Maybe this weekend.

I'm not a fan of any utility that adds anything to my registry. For Vista and Windows 7, I prefer the Task Scheduler.

Task Scheduler

For XP, the startup folder works. I don't have access to any Intel computers running XP for testing purposes so you will have to use the startup folder or add ThrottleStop.exe directly to this registry key.

HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\Run

This is the key for Windows 7 but I think XP is the same.

DanXbix: The Specifications box in CPU-Z does not show ES so your 940XM is definitely an OEM CPU and not an ES CPU. HWiNFO32 shows your laptop model as an M17x-R2 EC. I'm not sure what that means but it doesn't mean ES if that's what you were thinking.

 

hey uncle, whats your opinion about getting errors in prime95 when cranking up the multipliers on my 920xm...? I've ran the wprime 1.55 and TS 1024 test stable on 27x.... but prime95 reports errors in seconds...

What do they mean anyways? Anything that can seriously mess up things in the OS or something?

Thanks I wanted an answer from a pro

 

Well, something is wrong or i just don't understand how this works.
When i set multiplier to 5, it starts randomly change from 5 to 9 (as i see it in TS). Disabling C1E state and/or enabling power saver doesn't help.


But if i generate CPU load, then everything goes right and multiplier remains constant.


So what's the problem?

 

The TS Bench is not a very stressful benchmark. wPrime is more stressful and Prime95 is more stressful yet. Beyond Prime95 is software that uses the Intel Linpack libraries. LinX uses these libraries and will create some serious stress and heat.

http://www.xtremesystems.org/forums/showthread.php?t=201670

An overclock that is stable in one test might not be stable in a more stressful test. This is where you get to decide how stable is stable enough for you. If you only play single threaded video games then you will be able to get away with a much higher overclock compared to if you need to run a LinX like program 24/7.

An unstable overclock can in theory mess up Windows or any other program on your computer. You definitely don't want random data being written to your data files or being written to Windows.

GreyFlux: The Core i mobile CPUs use deeper sleep states like C3/C6 so you might not see ThrottleStop reporting a steady average multiplier when the CPU is lightly loaded or idle. This is normal. When you put a load on your CPU, then it might show a steadier multiplier.

Try setting the C6 button to C1 or C3 and see if that makes any difference.

 

unclewebb
Thanks. So when CPU uses C3/C6 states it has the lowest multiplier (& voltage i guess) and there is no software that can properly show me it, right? I just want to be sure that CPU is really in C6 state and don't consume power like with x9 multiplier.

Yes, in C3 state multiplier remains constant. But does it force CPU to be in C3/C1 state always? I mean when i set C-state in TS, CPU can't switch between them automaticly, right? That's not good.


Should i have TS running whole time or using it once after OS loading is enough? Do i need to run TS after hibernation/standby?
Will Intel Speed Step automatically control the multiplier after i uncheck "Set multiplier" in TS?
Are there any conflicts with Windows XP power schemes?

Sorry for bad English.

 

Is there a way to undervolt the i3-370m or overclock? i looked through all the options in throttlestop and can't really find anything.

 

conscriptvirus: Intel changed the design so their Core i CPUs can not be easily undervolted like you used to be able to do with the older Core 2 CPUs. A Core i3-370M does not use turbo boost so about the only way to overclock it is to use SetFSB if your motherboard supports that utility. For your CPU, ThrottleStop can accurately monitor it but that's about it.

GreyFlux: ThrottleStop is correctly reporting your CPU. In your first picture it shows that the CPU package is spending 38.3% of the time in the C6 sleep state. Using C6 minimizes power consumption but it will show an average multiplier randomly going up and down at idle.

When you use the C1/C3/C6/C7 button in ThrottleStop, you can block the CPU package from using some of the deeper sleep states. If you set this to C3, that will block the CPU package from using C6. That might make the multiplier steady at idle but actual power consumption usually goes up. That would be a bad thing if you are trying to maximize the amount of time your laptop can run on battery power.

When using ThrottleStop with Windows XP, it is best to set your power scheme to desktop so Windows and ThrottleStop don't fight over control of your CPU. I have not done any ThrottleStop testing or development when using XP so you will have to see what it does when using that operating system.

You have to have ThrottleStop running all the time if you want it to control your CPU.

 

Just back on the C1, C3, C6 and C7 modes to allow higher 4k writes on the SSDs on mobile platforms.

Someone posted at another thread a tool called Intel BITS.

BIOS Implementation Test Suite

I used this tool and it showed that both the C1 and C3 auto-demotion actions were disabled (reading on this it seems this is standard). There is an option to enable this which works and afterwards when I run the benches I'm getting similar 4k write speeds as I was when I disabled C6 via BIOS (at around 63MB/s).

I'm wondering if this is something you could look at and see if it makes sense to you? I haven't seen other people commenting on this but it works for me everytime. The only issue is that I have to boot via the USB initially, toggle this to enabled and then choose to boot into OS from there.

If this is something that could be put into TS as a function and it's confirmed it actually works (not just my benches), it would be a great thing to add. I haven't found anything negative from this.

 

@uncleweb: I downloaded your 2.99.6 TMastiffeS version from your sign and I tried to open it with my lap in my sign, but it showed mornitoring only, so i couldn't touch anything in there? so that means TS is no worked with 720qm at all?

I used the bios A010 mod by TheWiz

Thanks in advance!

 

djhuydx: Click on the "Turn On" button to go from Monitoring Mode to Active Mode. An i7-720QM has most of the fun features locked by the CPU so ThrottleStop won't be able to do anything too exciting with one of these. Start saving up for a 920XM ES CPU from Ebay. For about $400, your full load CPU performance can increase by over 100% as soon as you install one of those in your M17x-R2.

error-id10t: Can you do some testing for me? MSR 0xE2 bits[28:25] control the C States Auto Demotion stuff. Can you boot up without running BITS and show me a screen shot of this MSR and Crystal Disk Mark on the same screen. After that, reboot, run BITS and show me another screen shot of this same MSR and what Crystal Disk Mark reports.

I played around with this but I don't have a laptop that has the SSD problem so I didn't notice anything too exciting when toggling these bits. Once I see how these bits are set, I'll send you a test version of ThrottleStop so you can see if toggling these when in Windows makes any difference. Don't run ThrottleStop when testing just to make sure it doesn't interfere with anything.

MSR Tool

Enter 0xE2 in the MSR Number box and click on the Read MSR button.

 

thanks bro for guiding quickly

I know that the 720qm is not strong enough but it takes sometime for save up some money for 920xm, lol a lot of things to do first...

 

Darn.. I wish I could use this software..

Found that I can't really tweak anything yet with the 2820QM..

I should save up for a 2920XM..