The ThrottleStop Guide

Discussion in 'Hardware Components and Aftermarket Upgrades' started by unclewebb, Nov 7, 2010.

  1. aldam

    aldam Notebook Evangelist

    Reputations:
    39
    Messages:
    446
    Likes Received:
    1
    Trophy Points:
    30
    920xm oem @22x133
    TDP/TC 85/70

    3dmark06 test.

    logs from trottlestop

    Code:
    DATE     TIME    MULTI   C0%   CKMOD  CHIPM  PWR_mW  TEMP
    08/02/11 00:41:28  20.88   96.9  100.0  100.0       0   67
    08/02/11 00:41:30  20.68   99.8  100.0  100.0       0   69
    08/02/11 00:41:30  20.62   99.8  100.0  100.0       0   69
    08/02/11 00:41:32  20.51   99.8  100.0  100.0       0   70
    08/02/11 00:41:33  20.46   99.8  100.0  100.0       0   70
    08/02/11 00:41:34  20.38   99.8  100.0  100.0       0   72
    08/02/11 00:41:35  20.33   99.8  100.0  100.0       0   72
    08/02/11 00:41:36  20.32   99.8  100.0  100.0       0   73
    08/02/11 00:41:37  20.33   99.8  100.0  100.0       0   74
    08/02/11 00:41:38  20.14   99.8  100.0  100.0       0   74
    08/02/11 00:41:39  20.20   99.8  100.0  100.0       0   75
    08/02/11 00:41:40  20.16   99.8  100.0  100.0       0   75
    08/02/11 00:41:41  20.06   99.8  100.0  100.0       0   75
    08/02/11 00:41:42  20.07   99.8  100.0  100.0       0   76
    08/02/11 00:41:43  20.03   99.8  100.0  100.0       0   77
    08/02/11 00:41:44  20.02   99.8  100.0  100.0       0   78
    08/02/11 00:41:45  19.99   99.8  100.0  100.0       0   77
    08/02/11 00:41:46  19.96   99.8  100.0  100.0       0   78
    08/02/11 00:41:47  19.96   99.8  100.0  100.0       0   79
    08/02/11 00:41:48  19.92   99.8  100.0  100.0       0   78
    08/02/11 00:41:49  19.90   99.8  100.0  100.0       0   79
    08/02/11 00:41:50  19.90   99.8  100.0  100.0       0   81
    08/02/11 00:41:51  19.84   99.8  100.0  100.0       0   80
    08/02/11 00:41:52  19.85   99.8  100.0  100.0       0   81
    08/02/11 00:41:53  19.81   99.8  100.0  100.0       0   82
    08/02/11 00:41:54  19.78   99.8  100.0  100.0       0   82
    08/02/11 00:41:55  19.80   99.8  100.0  100.0       0   82
    08/02/11 00:41:56  19.75   99.8  100.0  100.0       0   82
    08/02/11 00:41:57  19.73   99.8  100.0  100.0       0   83
    08/02/11 00:41:58  19.74   99.8  100.0  100.0       0   83
    08/02/11 00:41:59  19.70   99.8  100.0  100.0       0   83
    08/02/11 00:42:00  19.71   99.8  100.0  100.0       0   83
    08/02/11 00:42:01  19.68   99.8  100.0  100.0       0   83
    08/02/11 00:42:02  19.66   99.7  100.0  100.0       0   84
    08/02/11 00:42:03  19.69   99.8  100.0  100.0       0   84
    08/02/11 00:42:04  19.65   99.8  100.0  100.0       0   85
    08/02/11 00:42:05  19.65   99.8  100.0  100.0       0   84
    08/02/11 00:42:06  19.65   99.8  100.0  100.0       0   85
    08/02/11 00:42:07  19.63   99.7  100.0  100.0       0   85
    08/02/11 00:42:08  19.62   99.8  100.0  100.0       0   85
    08/02/11 00:42:09  19.61   99.8  100.0  100.0       0   85
    08/02/11 00:42:10  19.59   99.8  100.0  100.0       0   86
    08/02/11 00:42:11  19.60   99.8  100.0  100.0       0   86
    08/02/11 00:42:12  19.56   99.8  100.0  100.0       0   86
    08/02/11 00:42:13  19.56   99.8  100.0  100.0       0   86
    08/02/11 00:42:14  19.57   99.6  100.0  100.0       0   86
    08/02/11 00:42:15  19.53   99.8  100.0  100.0       0   86
    08/02/11 00:42:16  19.57   99.7  100.0  100.0       0   86
    08/02/11 00:42:17  19.52   99.8  100.0  100.0       0   87
    08/02/11 00:42:18  19.51   99.8  100.0  100.0       0   86
    08/02/11 00:42:19  19.56   99.8  100.0  100.0       0   86
    08/02/11 00:42:20  19.51   99.8  100.0  100.0       0   87
    08/02/11 00:42:21  19.50   99.8  100.0  100.0       0   87
    08/02/11 00:42:22  19.52   99.8  100.0  100.0       0   87
    08/02/11 00:42:23  19.48   99.8  100.0  100.0       0   87
    08/02/11 00:42:24  19.51   99.8  100.0  100.0       0   86
    08/02/11 00:42:25  19.47   99.8  100.0  100.0       0   87
    08/02/11 00:42:26  19.46   99.8  100.0  100.0       0   88
    08/02/11 00:42:27  19.47   99.8  100.0  100.0       0   88
    08/02/11 00:42:28  19.44   99.8  100.0  100.0       0   88
    08/02/11 00:42:29  19.45   99.8  100.0  100.0       0   87
    08/02/11 00:42:30  19.46   99.8  100.0  100.0       0   87
    08/02/11 00:42:31  19.43   99.8  100.0  100.0       0   87
    08/02/11 00:42:32  19.41   99.8  100.0  100.0       0   87
    08/02/11 00:42:33  19.44   99.8  100.0  100.0       0   88
    08/02/11 00:42:34  19.41   99.8  100.0  100.0       0   87
    08/02/11 00:42:35  19.43   99.8  100.0  100.0       0   88
    08/02/11 00:42:36  19.40   99.8  100.0  100.0       0   89
    08/02/11 00:42:37  19.39   99.8  100.0  100.0       0   89
    08/02/11 00:42:38  19.42   99.8  100.0  100.0       0   89
    08/02/11 00:42:39  19.62   94.8  100.0  100.0       0   89
    Do you have any idea why the cpu didn't maintain 22x ratio?
     
    duttyend likes this.
  2. maximinimaus

    maximinimaus Notebook Evangelist

    Reputations:
    468
    Messages:
    635
    Likes Received:
    0
    Trophy Points:
    30
    Maybe the BIOS throttles the CPU because of the temps.
    Do you have the latest BIOS?
     
  3. aldam

    aldam Notebook Evangelist

    Reputations:
    39
    Messages:
    446
    Likes Received:
    1
    Trophy Points:
    30
  4. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,096
    Messages:
    5,867
    Likes Received:
    5,022
    Trophy Points:
    681
    Bobmitch: Blocking the CPU from using some of the low power package C states can improve some SSD benchmarks. I've heard that this new benchmark is more consistent than Crystal Disk Mark so you might want to do a comparison with that.

    AS SSD Benchmark 1.6.4194.30325
    Alex Intelligent Software - Downloads

    error-id10t is the person that introduced me to the C State Demotion stuff so if it works for you, thank him!

    aldam: As a CPU heats up, power consumption also goes up. Your log file that shows the average multiplier slowly throttling down is usually a sign that you are hitting the CPU's TDP / TDC limits. You might be able to raise those limits a little and maintain full turbo boost longer but then you are going to be running into the thermal limits of your CPU. You need better CPU cooling to get the most out of the XM CPUs. The best performance I've seen is with the Alienware M17x-R2. With a few tricks I have seen users running the 26 multiplier across all 4 cores when fully loaded. With your cooling, that's not ever going to be close to possible. Power consumption can also vary from one CPU to the next depending on how much leakage that CPU has.

    Edit: Here's a good example of 5150Joker's 940XM running 8 threads of the demanding wPrime benchmark with the 26.00 multiplier on each thread and not a hint of throttling or over heating.

    [​IMG]
     
    duttyend likes this.
  5. tilleroftheearth

    tilleroftheearth Wisdom listens quietly...

    Reputations:
    4,965
    Messages:
    12,347
    Likes Received:
    2,348
    Trophy Points:
    631
    unclewebb,

    On my i3 350M U30Jc on the High performance profile, I can't seem to make the cpu stay at 2.26GHz without dropping down to 1.1GHz (for example...).

    Is this a limitation of the chipset, or am I not setting up ThrottleStop 3.2 Beta2 correctly as indicated in post 728 on page 73?

    See:
    http://forum.notebookreview.com/7755405-post728.html


    What else must I do to make sure the cpu clock doesn't vary?

    (On my Q9450 it behaves properly...).
     
  6. T1mur

    T1mur Notebook Guru

    Reputations:
    0
    Messages:
    55
    Likes Received:
    3
    Trophy Points:
    15
    Try disabling C1E, that should stop the clocks from dropping down when idle.
     
    duttyend likes this.
  7. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,096
    Messages:
    5,867
    Likes Received:
    5,022
    Trophy Points:
    681
    Core i mobile CPUs have a wandering multiplier at idle regardless of how ThrottleStop is set up. I have heard of a Windows registry mod possible to lock these CPUs but it results in more power consumption and more heat as it blocks the CPU from using any of the power saving C States. Most users that have tried this have not found it to be a desirable long term mod for a laptop.

    The previous 45nm Core 2 processors could be locked to the highest multiplier at idle without causing any problems or excessive power consumption. I asked a friend about the above mod so if I hear more, I'll post some more info so you can test it yourself. This mod is not practical for most Core i mobile CPUs because blocking the C3/C6 C states will limit the amount of turbo boost to a multiplier increase of only +1. Your i3 does not use turbo boost so maybe the mod might be just what you're looking for.

    Edit: Here's the trick. Open up the registry and navigate to here:

    HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\intelppm

    Inside this folder is a REG_DWORD value called Start. Change this from 3 to 4 and reboot and see what happens. Play around with ThrottleStop and make sure C1E is disabled. Let me know if this works for you and if there are any 24/7 downsides like too much heat. I believe this blocks C3 and C6 being used so the multi doesn't wander around when lightly loaded. It can also give a boost to SSD benchmark scores since the CPU is ready for action. I'm not sure if you will need to disable EIST in ThrottleStop for this to work correctly.
     
    pressing, Vasudev and duttyend like this.
  8. natekron

    natekron Notebook Enthusiast

    Reputations:
    0
    Messages:
    19
    Likes Received:
    0
    Trophy Points:
    5
    I noticed that the new highest multiplier is permanently available after setting it in TS, since I can access it in CPU Genie and RMClock, as well, without starting Throttlestop.

    That's awesome since I'm using RMClock to control voltage/p-states (I set it to switch at 85% load, so it stays at lower states most of the times, unlike in TS or CPUGenie, where it jumps to the highest for every light/moderate load), and I was afraid I would lose the ability to use multipliers higher than the default 9x.

    Also, sometimes the FSB frequency is reported wrong - it increases randomly to 277-350 MHz. I noticed it when updating the BIOS and also when running Prime95 at frequencies above 2.8 GHz - what could be causing that? All of the other utilities report it fine at 266 MHz, and the temperature doesn't go over 81 degrees for the CPU... Could it be the north bridge overheating?
     
  9. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,096
    Messages:
    5,867
    Likes Received:
    5,022
    Trophy Points:
    681
    As soon as ThrottleStop unlocks a new maximum multiplier, that multiplier will be available for any other software to use. The bios usually resets this when you reboot. If you like RM Clock then you can use the ThrottleStop ExitTime=5 INI option so TS will run, unlock your processor and then automatically exit 5 seconds later. It's best to only have 1 program trying to control your multiplier at a time.

    The problem I had with RM Clock is that it didn't seem to work correctly with Core 2 Quad processors. Even when my CPU was fully loaded, it still left 2 of the 4 cores locked at the lowest 6 multiplier. I thought that was more of a bug than a feature. ThrottleStop has an INI option called SplitQuad. You can read about it in the included docs. When you are using this option you should be able to double click on ThrottleStop and have it run either side of your Core 2 Quad CPU at the highest multiplier while the other 2 cores are running with the minimum 6 multiplier. You can continue to double click on ThrottleStop and cycle through this.

    In the Options window there is a value called PowerSaver C0%. This lets you adjust the sensitivity of the PowerSaver feature. When you increase this value, it will take a bigger load before your CPU goes up to its maximum speed.

    ThrottleStop was designed to be a little different than RM Clock because I found for the newer 45nm Core 2 CPUs, there isn't always a big difference in idle power consumption even when there is a huge difference in FID and VID. Here's an example of my T8100.

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

    The reason for this is that these CPUs will enter the lower power C sleep states automatically and the VID drops down to a lower value automatically. The value that RM Clock or ThrottleStop or any software lets you set is ignored 95% of the time at idle. For this reason, coming up with the perfect FID / VID combos for every multiplier doesn't always translate into less power consumption. There have been some studies where the opposite is true and running a CPU at a slow speed only makes it inefficient. Running a CPU in short fast bursts lets it get its work done and then quickly get back into one of the deeper C3/C6 sleep states for longer periods of time where the real power savings are found. I'll try to post a link to a paper about this later on.

    Every time you click on the FSB button in ThrottleStop, the bus speed is recalculated. Are you using SetFSB or a similar program to adjust your bus speed? There is a Windows 7 feature / bug that when you adjust your bus speed this way, it can screw up the high performance timer that ThrottleStop uses. If you are a regular SetFSB user, there is a simple fix for this. My code to try and correct for this issue might need some improvement. How often does this happen and does clicking on FSB correct it or does it stay screwed up?

    Thanks for the feedback.
     
    duttyend likes this.
  10. natekron

    natekron Notebook Enthusiast

    Reputations:
    0
    Messages:
    19
    Likes Received:
    0
    Trophy Points:
    5
    Thanks for the reply, Unclewebb! I knew there were more settings to Throttlestop, but I can't open the Throttlestopdocs file for some reason (it's just blank). Is there an online copy of the documentation?

    I'm not using setFSB or anything like it, so the wrong FSB reading is weird...

    I did use PowerSaver for a while, and it works fine, but even at C0% 90% it's still too trigger happy, going to the maximum voltage and multiplier for much longer times than with RMClock, and the temperature on average is higher. RMClock seems to work well - it just doesn't exit sometimes (which leads to the settings not being saved), I haven't noticed any other weird behavior.
     
Loading...

Share This Page