The ThrottleStop Guide

Hi ,
Could not see the upload button so i''ve uploaded it here :
http://www.megafileupload.com/d32i/2015-06-20.txt

 

Hello there, Chrome shouts out a virus warning and decline to download the file here:

ThrottleStop 8.00 beta 2
https://www.sendspace.com/file/b4oure

A antivirus scan at Virus total result in this:


Whats the matter with that?

 

One other note im running a 180w power supply and i am receiving a 240w hopefully monday . So i can run the test again and upload some log files and compare the results ?

Thanks for your help in advanced .

 

Thanks for the info. Some antivirus software thinks that any executable file that is stored on SendSpace is either suspicious or it must have a virus. ThrottleStop has been around for years without any problems. Every time I start working on it and I release a new version, an antivirus program sees that the new version is different than the last version so flags it as suspicious. I use Avast and it used to nag me like this so I contacted them and complained. They never wrote back but Avast is not complaining anymore.

http://i.imgur.com/Zf09YF3.png

If you cannot download ThrottleStop with Chrome then try using Internet Explorer. There have been close to 2000 downloads of TS 8.00 beta 2 from SendSpace and so far there has not been a single complaint that this download contains a virus. Most users are happy with TS 8.

@ WillSkiies - If you have a look at the Power column in your log file, you will see that this column is staying steady right around 47 Watts. Your CPU has a TDP rating of 47 Watts so that is why it is throttling. These CPUs are designed to slow down (throttle) so that power consumption does not exceed 47 Watts. In other words, your CPU is running exactly as Intel designed it to run. This limit is built into the CPU so a bigger power adapter is not going to change anything. You can try using a modified bios or Intel XTU or ThrottleStop but I don't think any of these solutions are going to let your long term power consumption exceed 47 Watts.

Edit - To use www.pastebin.com, open up the log file so you can see the data, use the keyboard combo, CTRL+A to highlight the entire contents of the log file, CTRL+C to copy the contents to the Clipboard and then go to that website and use CTRL+V to paste the data into their website and click on the Submit button and it will give you a URL that you can share. Sounds complicated but it is very easy after you have done it a couple of times.

 

So I've been using Throttlestop with Core 2s for a long time, but this is my first time using it with a Core i. (i5 560m in a Latitude E4310 to be exact.) I'm somewhat confused by all the stuff, with the Core 2s it was pretty simple for me to get the best battery life. Undervolt to the lowest possible (.9v in the case of my M4400) and increase the clock until I find the highest stable one. (2.4GHz in the case of my M4400) Etc. Now, for the i5 what can I do for best battery?

 

@ cwazywazy - Adjustable voltage was removed from the original Core i processors. To minimize power consumption, make sure your C States are working properly. Eliminate background tasks that you really don't need and try to increase the amount of time that your cores spend in the low power C6 state. Beyond that, there isn't much you can do.

 

Quick question, how do I set up throttlestop as a monitor only, to check for throttling?

I'm not getting any log file, the check box is ticked and I have started throttlestop.

Do I need to turn it on? Will that not stop the throttling I'm looking for?

Cheers!

 

When you first start ThrottleStop, it should show in the title bar "Monitoring". In this mode it is generally not doing too much to prevent throttling. After you click on the Turn On button, the Monitoring word should disappear which shows that ThrottleStop is now active. What it does in this state depends on what additional boxes you check.

Did you unzip ThrottleStop into its own folder? Is that folder located somewhere on your hard drive where you have read / write access to it? If you do that and you check off Log File, it should create a new folder within the ThrottleStop folder called Logs and in that folder, a log file should be created with today's date on it. I think this log file gets updated once per minute. I am kind of a nut about inefficient software. I did this deliberately so ThrottleStop would not be constantly accessing the hard drive every second. If you have the log file open in Notepad while you are logging data, the new log file data should be written to the log after you close the file and wait a minute. Whenever you exit ThrottleStop, all log data will be written to the log file immediately as long as the log file is not already open.

Let me know if you get this working. I seem to recall that the XPS 15 has a few throttling issues.

 

I wasn't sure where to post this for Unclewebb so he can see it easily, because I couldn't PM it to you for some reason.

Alienware 17 R2 Throttle log file you requested for post 1349 of the alienware 13 15 17 thread

 

@ Mickbt26 - Thank you for that information. The CPU register I was curious about is not being used in your laptop for throttling purposes.

The next thing to check are the CPU power limit registers in the chipset. To do this you would need to use the Read Write Everything program. The 64 bit portable version is available from here.

http://rweverything.com/download/



The 3rd button from the left will open up the Memory window. You will need to click on the Address = button and adjust the memory Address to FED15900. Also click on the dword 32bit button so it is in the same format as the above picture.

The row that starts with A0 is where the power limits are located. You can see in my screen shot a bunch of zeros in this row because my Lenovo Y510P is not using this throttling method. Perhaps Alienware is. You would need to take a screenshot of RW Everything when your CPU is fully loaded and throttling down to 35 Watts. If it does show something interesting in that row, take another screenshot when the CPU is lightly loaded and not throttling to see if anything is different.

Edit - Off topic but the B0 row in that screenshot shows 800002A8 which is probably the current limit for my laptop. The first 8 shows that this register is locked. The 2A8 part is in hexadecimal.

0x2A8 = 680 decimal
680 / 8 = 85
Current Limit = 85 Amps

As manufacturers keep getting smarter with their throttling schemes, consumers will have to do the same.

 

View attachment 125452

I have done that here are the results:

 

In that register where it shows 8178, the 8 represents the lock bit so this register cannot be changed.

0x178 = 376
376 / 8 = 47
47 Watt power limit

So that looks OK too. The throttling to 35 Watts seems to be at a level that I don't have access to. If I come up with any more ideas, I will come looking for you in the Alienware forum. Thanks for answering some of the questions I had.

 

I have noticed in your throttlestop program if I show the limit reasons before it throttles I only get PL1 flashing red in the core section but when it throttles to 35w the POWER lights up red in the uncore section and PL1 goes off

Any use to you?

 

Hey thanks, I unchecked log file and checked it again and it started working.
I've got an AW15 now and getting some thermal throttling so just wanted to log what was going on.
It's a never ending story!

Thanks for your help!

 

@Mickbt26 while your showing uncore power limiting this does not reduce your package power. The reason the package is 35W is because the CPU is running at 2.8GHz and it is running at 2.8GHz because this is the p-state requested (MSR 0x199), not throttling from higher bins. Check ACPI or run HiMultis program to bypass it.

 

Thanks for responding. I don't even know what these tools are you mentioned. Do you have a link and ill check them out.

 

@ Mickbt26 - To correct MSR 0x199 you need to click on the ThrottleStop - Set Multiplier option and adjust that to the highest value. Also make sure that you have clicked on the Turn On button to take ThrottleStop out of monitoring mode so it can correct this register.

@Dufus - I think even when ThrottleStop is setup correctly, that Uncore - Power is what lights up in Limit Reasons for these Alienware laptops. Hopefully Mickbt26 can do some more testing and post a screenshot of how he has ThrottleStop setup and what Limit Reasons and ThrottleStop shows when the CPU is loaded. My Lenovo never shows POWER lighting up so I am not sure what value in what register actually causes that.

 

@Mickbt26 ACPI - Advanced Configuration and Power Interface , you can check the tables using RWEverything by clicking on the ACPI button.

Try TS first and if you want to try HiMultis extract it and place it in the TS folder as it needs the same drivers. It just sets two Haswell hardware registers once to give maximum CPU performance at the highest bins regardless of OS.

@unclewebb Uncore power is usually integrated graphics but in the OP's case seems to be other uncore components.

 

HIMultis didn't worked for me to fix the throttling to 2.8GHZ. Hope you guys can come with another solution .

 

Me either im afraid. Its starting to look like I have a bad CPU after all.

 

Did you try setting the multiplier with TS, running HiMultis while the multiplier has dropped to 28? Is the battery discharging while on AC?

Would someone make another MSR dump while multiplier has dropped to 28 and HiMultis has been run. Cheers.

 

Hi guys,

I have a problem with my old Fujitsu S751 i5-2520m stock clock is 2,5ghz, max turbo is 3,2ghz with a max tdp of 35watts and a hd3000 onboard.

First of all I dont think its a thermal problem because it only appears in games and its throttling even at 50 degrees. The cpu clocks itself for a second down to around 800mhz and gets back to normal and then happens again back and forth. For split second that causes a lag spike ingame. And somtimes even at 80 degress it runs smooth. I made the test with a game benchmark to make sure its not caused by a random fps drops from the onboard card which is obviously weak but enough to play at 50fps average. It seems to be random. I run the test a few times through. And later in the process the throttling barely happend even though the temps werent going down. If that lag spike happens the cpu clock goes down to around 800mhz for a second
Also shouldnt the multiplier fix the clocks to 3,2ghz in my case? even in the desktop it just all over the place from 8x to 32x
Settings:
http://i.imgur.com/FvqCxrt.jpg
http://i.imgur.com/g1vzgmM.jpg


throttling seems to happend when its going down to 8-9watts from what i can see
http://pastebin.com/Z1HDxJa5

and the one with furmark and intel burn test on:
http://pastebin.com/t4kHCz4t




If I run furmark or intel burn test its steady at around 3,1ghz and throttling at 100 degress down to 3,0-2,4ghz but never even one spike to 800mhz
I tried quite a lot with the energy saving modes in windows and turned stuff like speedstep off and on without any change in performence

Am I doing something wrong?
I appreciate anyones help

 

I set the multiplier to 40x and turned on throttlestop, Then set the test running with heaven extreme in windowed mode and ran 4 threads with throttlestop..
It started ok with 47w CPU usage and 3.3ghz across all cores. But when the prime test got to roughly 11% the cores dropped to 2.8ghz 35w and POWER lit up in both the core and uncore section of "limit reasons" I then ran hi-multis but nothing happened. It stayed at 2.8ghz.
As soon as I stopped the heaven test and kept the 8x prime test running the cores returned to 3.3ghz and 47w usage.
I'm running the 240w adapter and the laptop is pulling 170w when running these tests. The battery is not discharging.

 

Sounds to me like what happens for the HQ chips normally... they get a couple minutes or so of their "short power" then drop to their "turbo power". Except that you should be locked to 57W for 2 minutes then drop to 47W; except you're getting 47W for a bit then dropping to 35W. Which means that Alienware has broken your PC. I suggest contacting them and demanding a fix or replacement. And remind them that until you get a properly working PC you will continually demand replacements, no matter how long it takes, or you will demand a refund if they are unable to provide a working replacement. That will hit them where it hurts, because that's now a stack of units they must consider "refurbished" and therefore unable to sell normally I.E. wasted money.

 

The way your CPU works is it is only allowed to use the 32 multiplier when a single core is active. As soon as the second core becomes active, the maximum multiplier for both cores drops to 30. When playing some games, there may be time when only a single core is active but for a lot of the time both cores are going to be active. That is why your log file shows a lot of time where the multiplier is just a hair over 30.00.

Your ThrottleStop screenshot shows that ThrottleStop is in Monitoring mode. That means even though Clock Modulation, Chipset Clock Modulation and the Set Multiplier feature are checked; ThrottleStop is not actively doing anything. You need to click on the Turn On button.

During your Furmark stress test, the CPU is not fully loaded. The multiplier drops down to 25.00 but I am not sure what that means because I do not know if you have ThrottleStop in Monitoring mode or Active mode. Furmark is a little over kill for most laptops and it is definitely over kill for the Intel HD3000. I would stick to trying to play some real world games.

I do not see your CPU running at 800 MHz anywhere in the log files you posted. Are you doing all of your monitoring with ThrottleStop or are you using another monitoring program at the same time?

Try doing another test with ThrottleStop turned on and make sure your log file contains information from when you are gaming. In your previous log file there seems to be sections of time where there is very little load on the CPU like you might have done an ALT+TAB out to the desktop. Generally speaking, when there is a load on your CPU, the multiplier seems to be where it should be.

Your ThrottleStop screenshot at idle shows your CPU package spending 91.7% of its time in C3. When a mobile CPU is in these lower power C States, it is not unusual for ThrottleStop to report the multiplier going up and down wildly as cores and threads continuously enter and exit various C States. That does not mean that there is a problem. That's why I prefer to see a log file while the CPU is actually working and not just sitting idle. A loaded CPU is a lot more meaningful.

Are you using the Windows High Performance profile? ThrottleStop and your computer work best when set to that.

 

Okay Unclewebb, I found something interesting. I believe you had asked me to check earlier, but I can't remember if I told you yet.

TS8 definitely is capable of adjusting the CPU and voltages etc, however it does not save in the BIOS. I currently am basically on a fresh install of Windows 8 which does not have intel XTU running, and settings I save in TS8 do not save in the BIOS, even though while TS8 is running, they do in fact work as intended. Tested with cache ratio multiplier and current limit. Since I didn't have XTU I didn't have a reliable way to force current limit throttle, but I did notice that my BIOS kept the underclock of the cache ratio that I put in it while when I ran TS8, the cache ratio multiplier fluctuates as high as I have it set in TS8.

This leads me to believe that TS8 will override BIOS-level CPU adjustments (such as those made by XTU and in the BIOS) but will not overwrite the BIOS' settings. Is this how it was intended?

 

ThrottleStop has zero interaction with your bios. When you first start your computer, the bios sets up your CPU to whatever settings that you previously saved when you were in the bios. ThrottleStop has no idea and really doesn't care what values the bios set the CPU to. As soon as you start ThrottleStop, it tries to use whatever settings that you previously saved in ThrottleStop. If ThrottleStop is being run for the first time or if ThrottleStop cannot find the ThrottleStop.INI configuration file because you deleted it; ThrottleStop will read the current CPU settings directly from the CPU and ThrottleStop will use those settings as a starting point.

I always thought this was the safest way to do things. If you accidentally Save some unstable settings while you are in the bios, you can end up in an infinite boot loop situation where your CPU might keep crashing over and over again while Windows is trying to boot up. To me it makes more sense to give your laptop a chance to boot up before applying any of the previously saved ThrottleStop settings.

My recommendation has always been not to auto start ThrottleStop until you have a very good idea of what your CPU is capable of and how much voltage it is going to need to be stable. That is also one of the reasons why I have never added a simple auto start with Windows feature to ThrottleStop. I encourage users to use the Task Scheduler so they have to put some thought into what they are doing. If you do something dumb, you can still end up in an infinite boot loop but you should be able to save yourself. You would just need to boot into Safe mode and delete the ThrottleStop.INI file so your unstable settings will be ignored next time you boot up and run ThrottleStop.

No one has mentioned the FIVR - Save Voltage Changes section. I like the option, OK - Save voltages after ThrottleStop exits. This allows a user to set their voltages and turbo multipliers and then you can take these settings for a test drive so to speak with no worries. You can run Prime95 or do some gaming for an hour or whatever and after that you should know if these settings are good or bad. If these settings are good and you want your new settings saved to the ThrottleStop.INI configuration file, all you have to do after testing is exit ThrottleStop, your settings will be saved and then next time your start ThrottleStop, it will use those new settings that you know are safe.

On the other hand, if the settings that you came up with were not stable and lets say Prime95 stopped with an error or your computer crashed with a BSOD, those crappy settings you were testing will automatically be lost and they will not be saved. After you reboot and ThrottleStop starts up, it will use the last safe settings that you previously saved. I think this gives overclockers an extra layer of protection.

When you are first starting out and you have absolutely no idea what your CPU is capable of, maybe the option, OK - Do not save voltages might be best. This is intended for the noobs or for users that are really taking some wild guesses and don't want those settings to ever be saved.

The last setting, OK - Save voltages immediately is more for advanced users that know exactly what they are doing. If you have a really good idea of what your CPU is capable of and you have reached the point where you are just making some minor adjustments of a mV or two, maybe then it is time to take off the training wheels and just save your changes immediately.

The goal with all of this was to try and create something that was safer and easier to use than Intel XTU. When using XTU, after my laptop crashes, it takes forever for it to boot up again and then finally XTU starts up and it says that it crapped the bed and has returned the CPU to default settings. That's fine but the boot time when this happens on my Y510P is brutal. It takes forever for me to test out anything. I have not had any issues like this when using ThrottleStop.

I have got zero feedback about this so I am not sure what is best. Maybe users prefer when XTU botches their bios settings and they get to waste the day trying to recover from the mess.

 

Good to know =D.

I have only ever used "save voltages immediately" actually XD.

I told you I use TS8 for OCing and such right? Usually if I want to test an OC I just pull up TS8 and change the multipliers and the voltage and leave it be. It's worked pretty well for testing overclocks and such in games like GTA V, as I use borderless windowed and I just can hit "ok" and suddenly I've got an overvolt and a 400MHz OC and I'm testing things, then when I'm done I can return it to normal and such =D. I like that bit, especially whether you want the bar to be 125mV or 250mV increments etc.

It's a good goal, for sure. I don't get the slow boot issue with XTU due to plextor M6Ms being a really fast SSD, but I can see why from your experience it'd make sense. It always annoys me that it would reset all my settings (current limit and power limit etc) and I'd have to re-do everything (though granted I knew what my settings were so it was not hard).

I think people may not understand how they should use that feature, probably. I really couldn't say.

 

@Mickbt26 Thanks for the report. Once Limit Reasons shows "power" on the core then you are being power throttled. This wasn't coming on before according to your previous post so while HiMulti's is overriding p-states it may be that there are counter measures such as "limiting p-states does not work so enforce power limiting". Other methods may be to disable turbo boost or set the maximum performance override. Wouldn't be surprised if they are learning ThrottleStop tricks and providing counter measures (TSCM). You could check those memory locations again but most likely the limit is issued by EC. You could also try temporarily disabling the IntelPPM service to see if it makes a difference. If it's EC controlled then not much you can do unless there's an override in the EC data (RWEverything), EC firmware mod or trick the CPU into thinking it's using less power which isn't ideal.

Your "guaranteed" frequency is the HFM (base frequency) while turbo's are "opportunistic". IOW you can not accuse the manufacturer of the CPU not working correctly if it is limited to the base frequency of 2.8GHz. About all you can do is complain with your disappointment and/or make your opinion known through your wallet or fix it yourself and probably void your warranty / possibly break the laptop attempting a fix.

@D2Ultima TS would need to know where to save settings on the BIOS chip and that will vary across different systems and sometimes same system different BIOS version. We briefly discussed XTU on another forum where I said it may not be a bad thing not to have XTU in control of your settings. For instance running unlocked this is what XTU shows.

Note it refuses to display/accept the real set ratio's and instead displays the default max ratio limits ie core ratio is shown as 40/39 while it is actually 47/47. IOW XTU has it's own agenda, that's only fine if you can work within it.

 

First of all thank you for the fast response.

1.About the screenshot i usually turn it on before gaming and turn it off while i am not. I guess when I made the screenshot I just forgot to turn it on before taking the screenshot i guess.
Here another one with being turned on.:
http://i.imgur.com/J7HL4LW.jpg

2.And about the power profile yeah it is on high performence with maximum state at 100%.
About the first test. Well its a benchmark test within the game. And the test isnt long like 2.30 mins and then it loads the map again if its set on loop. That might be the reason why its not on load for a few seconds every 2.30 mins. Also the load ingame isnt even close to 100% i dont understand why its throttling because it doesnt even reach 80 drgrees. At prime 95 i get to 100% load with 90 degrees and there it barely throttles just to 25x with stock clock 2,5hghz

Second test:
http://pastebin.com/6wKUvAgj
Lags start at around 13:53 just when i started the benchmark
at 13:56 i forgot to set the loop so there is not a much load for like 30 seconds
Second benchmark is at around 13:57. Whats weird temps are higher but there is close to no throttling what so ever. Just a 13:59 there are a few ones and at 14:00 the test was over and started again the loop so there are like 10 seconds low load
Third benchmark at 14:00 No lags what so ever. Just once at the ending at 14:03.
Fourth benchmark 14:05 small spikes again
Last one at 14:06 again small spikes at the end 14:09


prime95 full load with throttlestop being on:
http://pastebin.com/taQUEDs8

it just throttles down to 2,5ghz stock clock

3. About the log and why there is no 800mhz shown. I have no clue why it does not monitor that drop. I see that drop in cpu-z and hwmonitor. The multiplier goes to 8x and the clock rate to 795-798mhz and that causes the lag spike ingame. In the throttlestop log i see that drop only with the wattage dropping down as far as i can see that.
Should i try a real benchmark software like heavenbench 3dmark or what should i use?
I mean i experienced that kind of behaviour in other games too
And i dont talk about graphic intensive games. Obviously the hd3000 wouldnt do the job


Anything i should change in throttlestop?

 

Hi, im new to throttlestop. I have a lenovo y50 i7-4720, 16gb ram, 4gb 960m. Can someone with the same laptop as mine help me with what is the best optimabl settings to have throttle stop run in, othanks.

 

Code:

   DATE       TIME    MULTI   C0%   CKMOD  CHIPM   BAT_mW  TEMP    VID   POWER
2015-06-28  13:52:15  29.70   27.3  100.0  100.0        0   61   0.7805   16.4
2015-06-28  13:52:16  29.92   41.0  100.0  100.0        0   61   1.2209   21.9
2015-06-28  13:52:17  30.07   36.2  100.0  100.0        0   65   1.2209   19.9
2015-06-28  13:52:18  29.54   23.9  100.0  100.0        0   64   1.2209   14.9
2015-06-28  13:52:19  29.68   29.3  100.0  100.0        0   63   1.2209   17.3
2015-06-28  13:52:20  29.67   33.7  100.0  100.0        0   59   0.7855   18.8
2015-06-28  13:52:21  29.55   35.9  100.0  100.0        0   61   1.2209   19.6
2015-06-28  13:52:22  29.88   33.8  100.0  100.0        0   68   1.2209   18.7
2015-06-28  13:52:23  29.07   26.4  100.0  100.0        0   63   1.2209   15.2
2015-06-28  13:52:24  29.65   25.3  100.0  100.0        0   66   1.2209   15.3
2015-06-28  13:52:25  29.01   26.6  100.0  100.0        0   58   1.2209   16.0
2015-06-28  13:52:26  29.14   26.6  100.0  100.0        0   57   0.7805   15.7
2015-06-28  13:52:27  29.66   36.0  100.0  100.0        0   66   1.2209   19.1
2015-06-28  13:52:28  29.16   27.5  100.0  100.0        0   62   0.7805   16.0
2015-06-28  13:52:29  29.34   20.1  100.0  100.0        0   62   1.2209   12.2
2015-06-28  13:52:30  29.82   31.5  100.0  100.0        0   63   1.2209   18.2
2015-06-28  13:52:31  30.07   33.3  100.0  100.0        0   57   1.2209   18.8
2015-06-28  13:52:32  30.14   34.0  100.0  100.0        0   59   1.2209   18.9
2015-06-28  13:52:33  29.34   28.0  100.0  100.0        0   62   1.2209   16.2
2015-06-28  13:52:34  28.95   13.0  100.0  100.0        0   59   1.2209   10.1
2015-06-28  13:52:35  29.71   29.0  100.0  100.0        0   65   1.2209   16.3
2015-06-28  13:52:36  27.86   13.9  100.0  100.0        0   60   1.1909   10.5
2015-06-28  13:52:37  28.16    6.3  100.0  100.0        0   60   1.2209    6.9
2015-06-28  13:52:38  28.04   15.5  100.0  100.0        0   62   1.2209   10.5
2015-06-28  13:52:39  29.42   26.4  100.0  100.0        0   57   1.2209   15.6
2015-06-28  13:52:40  27.89    6.8  100.0  100.0        0   56   0.7855    7.1
2015-06-28  13:52:41  28.37    9.8  100.0  100.0        0   61   1.2209    8.2
2015-06-28  13:52:42  30.11   33.8  100.0  100.0        0   57   1.2209   18.6
2015-06-28  13:52:43  28.04    6.3  100.0  100.0        0   54   0.7855    6.9
2015-06-28  13:52:44  28.01    6.5  100.0  100.0        0   54   0.7805    6.9
2015-06-28  13:52:45  29.86   34.8  100.0  100.0        0   60   1.2209   19.0
2015-06-28  13:52:46  27.70    6.5  100.0  100.0        0   53   0.7855    7.2
2015-06-28  13:52:47  28.30    6.2  100.0  100.0        0   54   1.2209    6.8
2015-06-28  13:52:48  29.03   24.3  100.0  100.0        0   61   1.2209   14.5
2015-06-28  13:52:49  28.85   18.7  100.0  100.0        0   55   1.2209   12.4
2015-06-28  13:52:50  28.02    6.2  100.0  100.0        0   53   0.7805    6.8
2015-06-28  13:52:51  28.35   17.9  100.0  100.0        0   62   1.2209   11.4
2015-06-28  13:52:52  29.42   27.9  100.0  100.0        0   56   1.2209   16.0
2015-06-28  13:52:53  27.92    6.1  100.0  100.0        0   55   0.7855    6.7
2015-06-28  13:52:54  28.06    6.4  100.0  100.0        0   56   1.2209    6.8
2015-06-28  13:52:55  29.53   34.3  100.0  100.0        0   61   1.2209   18.3
2015-06-28  13:52:56  28.13    7.9  100.0  100.0        0   56   1.2209    7.8
2015-06-28  13:52:57  28.32    7.6  100.0  100.0        0   56   1.2209    7.5
2015-06-28  13:52:58  29.84   34.8  100.0  100.0        0   62   1.2209   18.2
2015-06-28  13:52:59  28.21    9.1  100.0  100.0        0   58   1.2209    8.6
2015-06-28  13:53:00  28.06    6.8  100.0  100.0        0   55   0.7855    7.1
2015-06-28  13:53:01  29.21   27.0  100.0  100.0        0   65   1.2209   15.3
2015-06-28  13:53:02  29.09   16.2  100.0  100.0        0   58   1.2209   11.9
2015-06-28  13:53:03  28.18    6.4  100.0  100.0        0   57   1.2209    7.0
2015-06-28  13:53:04  28.36    6.7  100.0  100.0        0   58   1.2209    7.2
2015-06-28  13:53:05  29.59   34.2  100.0  100.0        0   58   1.2209   18.8
2015-06-28  13:53:06  28.19    6.9  100.0  100.0        0   56   1.2209    7.2
2015-06-28  13:53:07  28.68    6.5  100.0  100.0        0   57   1.2209    7.0
2015-06-28  13:53:08  28.78   22.5  100.0  100.0        0   58   1.2209   13.9
2015-06-28  13:53:09  28.62   10.2  100.0  100.0        0   57   1.2209    8.6
2015-06-28  13:53:10  28.00    6.6  100.0  100.0        0   56   1.2209    7.0
2015-06-28  13:53:11  29.76   33.7  100.0  100.0        0   59   0.7805   18.1
2015-06-28  13:53:12  28.21    7.2  100.0  100.0        0   56   1.1909    7.4
2015-06-28  13:53:13  28.12    6.6  100.0  100.0        0   56   0.7855    7.0
2015-06-28  13:53:14  29.25   27.9  100.0  100.0        0   61   1.2209   15.7
2015-06-28  13:53:15  28.73   13.3  100.0  100.0        0   56   1.2209   10.1
2015-06-28  13:53:16  28.40    6.5  100.0  100.0        0   55   1.2209    7.0
2015-06-28  13:53:18  28.45    6.6  100.0  100.0        0   55   1.2209    7.0
2015-06-28  13:53:19  29.96   33.4  100.0  100.0        0   56   0.7855   18.5
2015-06-28  13:53:20  28.48    6.7  100.0  100.0        0   56   1.1909    7.0
2015-06-28  13:53:21  28.37    6.4  100.0  100.0        0   54   1.2209    6.9
2015-06-28  13:53:22  29.39   34.1  100.0  100.0        0   55   1.1909   17.7
2015-06-28  13:53:23  28.45    6.5  100.0  100.0        0   54   0.7855    7.1
2015-06-28  13:53:24  28.42    6.4  100.0  100.0        0   55   0.7855    6.9
2015-06-28  13:53:25  29.15   18.2  100.0  100.0        0   55   1.2209   11.6
2015-06-28  13:53:26  28.75    6.4  100.0  100.0        0   54   0.7855    6.9
2015-06-28  13:53:27  29.49   29.3  100.0  100.0        0   63   1.2209   15.9
2015-06-28  13:53:28  29.67   28.8  100.0  100.0        0   56   0.7855   16.2
2015-06-28  13:53:29  28.16    6.6  100.0  100.0        0   55   0.7855    7.0
2015-06-28  13:53:30  29.20   15.2  100.0  100.0        0   61   1.2209   10.3
2015-06-28  13:53:31  29.39   19.9  100.0  100.0        0   58   1.2209   12.7
2015-06-28  13:53:32  29.88   36.5  100.0  100.0        0   61   1.2209   19.2
2015-06-28  13:53:33  29.48   19.4  100.0  100.0        0   64   1.2209   12.7
2015-06-28  13:53:34  28.84   24.1  100.0  100.0        0   61   1.2209   14.1
2015-06-28  13:53:35  29.49   29.6  100.0  100.0        0   56   1.2209   16.2
2015-06-28  13:53:36  29.48   30.6  100.0  100.0        0   56   0.7855   17.6
2015-06-28  13:53:37  29.37   25.3  100.0  100.0        0   58   1.2209   14.8
2015-06-28  13:53:38  29.60   26.1  100.0  100.0        0   63   1.2209   14.8
2015-06-28  13:53:39  29.60   31.1  100.0  100.0        0   61   1.2209   16.5
2015-06-28  13:53:40  29.20   22.2  100.0  100.0        0   57   1.2209   13.6
2015-06-28  13:53:41  29.77   35.0  100.0  100.0        0   58   1.2209   18.6
2015-06-28  13:53:42  29.50   32.9  100.0  100.0        0   57   1.2209   18.5
2015-06-28  13:53:43  29.33   27.9  100.0  100.0        0   61   1.2209   16.2
2015-06-28  13:53:44  28.72   20.4  100.0  100.0        0   59   1.2209   12.8
2015-06-28  13:53:45  30.40   33.2  100.0  100.0        0   61   1.2209   19.6

@ VxO - I copied about 90 seconds of the log to have a better look at it.

During the first part of this section of the log, the CPU core temperature is only in the 60C to 65C range so the CPU is definitely not overheating.

At first, power consumption is fairly consistent until about 13:53:35. Then it goes through a cycle of being OK and then dropping down for about 2 seconds and then back to OK and it does this over and over again. During each 1 second monitoring interval, the average multiplier dips a very small amount during these cycles but it is not dropping significantly. The interesting part is that the C0% is dropping significantly. A CPU is in the C0 state when it is actively working on something. When the CPU has nothing to do, it exits the C0 state and drops down to one of the low power C States like C3 or C6.

My first thought was maybe it is a GPU driver issue. Are you using the most up to date Intel GPU driver? Have you tried to log your GPU performance with MSI Afterburner or GPU-Z if either of those work?

My next thought was that your CPU is having these momentary pauses because it is sitting idle, waiting for data to be accessed from the hard drive. When a mechanical hard drive is getting ready to fail, it is not unusual for them to have brief moments of sluggishness. This theory also fits with your observations. If you are doing a benchmark loop, the first time through this loop, information will need to be loaded from the hard drive into memory. The second time through the same loop, as long as nothing else Windows related needed that memory, most of the benchmark info should already be in memory so there will be less accesses to the hard drive needed and less stuttering.

Check out HD Tune.

http://www.hdtune.com/

There is a free trial version but you can also download an older free version (2.55) to see how healthy your hard drive looks during its benchmark.

I am not 100% sure what the problem is but it doesn't look like ThrottleStop is going to be able to help you. If you have some time to kill could you try running this same benchmark after booting up so there is nothing in memory. Leave ThrottleStop in Monitoring mode and check off the More Data option. It will create a pile of numbers that might show better what is really going on.

@ Jmunandar - I posted a ThrottleStop pic in the Y50 thread where you also asked this question. Let me know if you have any further questions.

 

Sure, go ahead and set it to 112.00A, or even to 130.00A if you want to. Do this with both Core Current Limit and Processor Current Limit. These are only a limiter, not a static value like voltage. The CPU will only draw as many amps/watts as it needs, (or is capable of,) so there is no harm in raising it to a level beyond what the CPU is capable of. (Obviously, this does not apply to voltage settings.) Anything other than a desktop X, K or mobile Extreme CPU is going to have fixed TDP. This is controlled by the CPU, and to whatever extent the OEM implemented draconian controls to block overclocking, further limited by the BIOS. Settings for higher power limits will be ignored beyond what those fixed limits are.

 

My intel drivers are up to date

I am gonna test the gpu unit tomorrow
Is there anything particular I should look into ?

Here the hard drive results. Dunno how bad it is if some sectors are reallocated.
http://imgur.com/a/Q2SMm

 

This is true now, but remember that older machines don't have this limitation unless the BIOS does it. I've made my 4800MQ pull 94W consistently with Linpack before.

 

The guy I was responding to has an Alienware 15 with 4710HQ. There have been a few exceptions in socketed CPUs, but not that many. The exceptions are mostly limited to Haswell.

 

@unclewebb
I did some checking with the cache ratio functions of TS8.
Apparently I can set my cache ratio via TS8 to whatever I want really (I set it to 8x in my OC profile) but I noticed that as long as the min/max setting had a higher value (it's 8 to 35 by default for me; probably because I set my BIOS to have core #1 limited to 35x instead of 37x like default) it would still bounce higher than what I set it to. Is this how it should be? Like, if I tick "unlock adjustable cache" in the FIVR setting then set my cache ratio manually to say... 10x, but the limit on the side is 8x to 35x, should it bounce between 8x and 35x? This is what was happening, and it made me wonder what the difference in the settings were. I was doing reading and found that lowering cache ratio (or ring bus ratio) improves stability with truly negligible performance changes, so I was looking, but I noticed it had that behaviour.

Ah. I was just being general about it. I know none of the HQ chips do it, I just jump on general statements XD. Did the old QM chips from Sandy/Ivy do that in compatible machines (like Clevos and your M18x R1/R2)? Or was it a Haswell thing?

 

That sounds like the kind of feature that I added to ThrottleStop.

The Cache Ratio slider lets you adjust the maximum cache ratio. The Cache Ratio Min / Max feature gives you a lot more control over the cache ratio.

For many applications, changing the cache ratio doesn't make that much of a difference to overall performance. If you are overclocking your CPU to its limit, it is usually a good idea to reduce the cache ratio so you have some more headroom to overclock the CPU. Desktop Haswell CPUs are generally happy with a 40X cache ratio when the CPU multiplier is around 45X.

For benching, the higher the cache ratio you can get away with, the better. Here is a simple Super Pi comparison, first with the cache ratio locked to 8, followed by the cache ratio locked to 36. Who says tweaking the cache ratio doesn't make a difference? CPU MHz was identical for both benches.

Cache Ratio 8


Cache Ratio 36

 

Well this is the thing. If I set the cache ratio slider to "10" and leave the min/max feature at "minimum 8" to "maximum 40" for example, my cache ratio will hit 40. It felt a little... weird? Redundant? I don't know how to describe it.

And yeah, I know some things benefit a lot, but I was just setting things up for a quick profile and I noticed it. Since I get black screen shutdowns at 3.9GHz I wanted to minimize potential errors so I set a specific voltage and lowered the cache ratio. If that proves stable in testing for most things I'll simply keep it and try to up the cache ratio some, but it was the baseline XD. For all I know my original problem was that I used to set the cache to 39x and my CPU would have been fine with it at 35 or 36 XD.

 

@D2 Ultima - I agree that the cache features in ThrottleStop are redundant but we can blame Intel for that. They use multiple registers to control the cache so I added features to ThrottleStop to control both of the cache registers independently. As I mentioned, the Cache slider only seems to control the maximum cache ratio when overclocking the cache. The Min / Max feature over rides the slider, except when overclocking. It is easy to play with these settings and then you can watch what cache ratio you are actually getting in the monitoring table in the FIVR window.

If you can monitor the GPU, see if the MHz are steady or throttling back.

Reallocated sectors are a sign that your hard drive is failing. It might fail tomorrow or it might keep going for months or years. I had a sick drive like that earlier this year and luckily I was able to clone it to a new drive without any significant loss of data. All my ThrottleStop code was sitting on that drive so I thought it was going to be game over. I tend to forget doing proper backups for months at a time.

I would like to see a ThrottleStop log file with ThrottleStop in Monitoring mode. If you have a lot more throttling when setup like this, that would prove that ThrottleStop might be helping a little but is not helping enough to stop the throttling that you are seeing. If that's the case, there is a hidden option in ThrottleStop that might be able to help.

If you don't have access to a spare laptop drive for testing purposes, I was thinking that if you had enough RAM, maybe you could setup a RAM drive, install a game to it and see what you get for throttling. This might help bypass your sick drive.

 

Makes sense to me about the number of registers. Better to have both than only one and that one is the wrong one for some instance or other. Thanks for clearing it up!

 

So I have a little bit of a struggle with my Latitude 3540. The cpu seems to enjoy dropping the boostclock when thermals hit 75C, to me this is quite bizarre as my i5-4200u has a T Junction of 100C. http://pastebin.com/jqVnpyxg. I have tried to remedy this by unticking BD Prochot and setting a static voltage, but to no avail. Is there any way to keep the 4200u at boost clocks till around 95C, and prevent this early throttling from destroying gaming performance?

 

Thanks for the log file but can you also post some screenshots of how your have ThrottleStop setup? With this CPU, I would not use a Static Voltage. I think you will be better off if you use Adaptive voltage combined with some negative offset voltage instead.

The 4200U is a low watt processor. It is too bad that manufacturers install these things in so many laptops. I have heard that they are not all bad but lets just say that I am not a big fan of the U. Intel designed this line of processors to throttle and they do a great job at that. They would be great if you could lock them at their maximum possible speed but depending on what a manufacturer has done in the bios, they can end up running at far less than that. Most of the throttling is usually triggered by power consumption and not by temperature.

Post some screenshots so I can have a look but there is probably not much ThrottleStop or any software is going to be able to do.

 

I won't call myself any expert when it comes to how this program works, but do these pics work? And I knew going in that the 4200u was a low power chip, but I was hoping that coupled with the 8850m on my laptop it would be enough for csgo. It comes so close but I get some stutters off 60 fps that really hurt gameplay. I still can't complain about the $470 price tag though, if there were some more powerful machines with 1080p screens at that pricepoint, I would have jumped on them. Especially if I knew how much of a hell it was for those running dual graphics with Intel and AMD gpu chips, out of the box my laptop would hang on sleepmode and couldn't run cs above 15fps, just praying now that I can push it the extra few meters to have the perfect machine for my on the go needs.

 

Amazing work unclewebb, thank you so much. Just replace my Intel XTU with throttlestop and I love it!

I have only one issue, I added throttlestop to the windows startup folder, but it doesn't start automatically on boot, I am on windows 8.1.

What should I do to make it start automatically?

 

You need to check TDP level control and set it 2. After cpu throttle problems should be gone...

I would like to know that also. I have Windows 7 and tried to put Throttlestop icon to startup folder, but somehow I m not allowed.

 

Set it as scheduled task - triggered on user login

 

Or you guys could do what I do and pin it to the start menu and just click it after you load in

 

Actually, I didn't see the second post...my fault.
I set up throttlestop in the task scheduler, working perfectly!