The ThrottleStop Guide

hello all, thanks a bunch for this guide!

but i need some help, i've downloaded and used throttlestop5.0 on my m17x with 920xm; had my multi's set at 24 and my cards undervolted at 1.03v. i set tdp/tdc ratios at 90/75 and i saw a drastic improvement while playing gw2, i was more than happy, everything was stable for about 2 hours or so but then i had a hard shutdown, screen goes black and the game freezes. i've since set multipliers at 15, and power limits at 70/65, and have reflashed my cards to normal voltage settings (1.1v) now however i can't even get 5 minutes before the screen goes black.

*was just monitoring my laptop while it was on to see if it was stable just sitting idle, had tsop running but not on and was only using one card, was about to set ratios to default (62/62) and turn on crossfire to see if i could play, as soon as i turned on crossfire i got bsod.

i've even tried to play or use IE without tstop and i'll get a hard shutdown, not sure if it's a power issue or if i fried something

 

I don't have the c state button in throttlestop. There is a box to choose c state on or off but it makes no difference. With throttlestop on, the cpu makes whining noise under low load. I guess it's caused by deep c state but there is no option in my throttlestop 5 to limit c state.
My processor is Core 2 Duo Mobile, in Intel's document I can't find a MSR to limit c state in core 2 duo. Is that why there is no c state button for core 2 duo?

Intel Mobile Core 2 Duo
Windows Server 2008 R2 Enterprise Edition (Full) 64-bit SP1 (Build 7601)
CPU Arch : 1 CPU - 2 Cores - 2 Threads
CPU PSN : Intel Core2 Duo CPU T5270 @ 1.40GHz (ES)
CPU EXT : MMX, SSE (1, 2, 3, 3S), EM64T
CPUID : 6.F.D / Extended : 6.F
CPU Cache : L1 : 2 x 32 / 2 x 32 KB - L2 : 2048 KB
Core : Merom (65 nm) / Stepping : M0
Freq : 797.76 MHz (199.44 * 4)

 

ThrottleStop has more C State options for the Core i processors because Intel has more information in their public documentation for the Core i processors compared to the Core 2 processors.

Some Core 2 owners have had success reducing CPU whine with the ThrottleStop C States check box but I am not surprised if it does not work for everyone.

Do you have your minimum and maximum processor state set to 100%? Read the included documentation if you need an example of how to do that.

Have you tried running ThrottleStop with Set Multiplier checked and the C States box cleared and have you adjusted the VID voltage appropriately? Playing around with these settings might help to reduce the noise you are hearing.

 

how can i know if throttlestop working or not???i try with my i7 2720,but didnt see anything happen

 

What were you looking to see happen? ThrottleStop is a program that was designed to stop CPU throttling. If your CPU is running as Intel intended it to run then there is not a lot that ThrottleStop is going to be able to help you with. It can't fix a problem if a problem doesn't exist.

You didn't mention what laptop model you have or what type of software that you like to run or what problem you have that you are trying to correct. If you are a gamer then download ThrottleStop 5.00 and run it in monitoring mode and then click on the Log File option so you will have a thorough record of your CPU's performance. Go play a game or run whatever demanding software you like to run and then post a copy of your log file here so I can have a look. Exit ThrottleStop and then Copy and Paste the log file data to Pastebin.com and post a link here.

Core i7-2720QM
4 core: 30
3 core: 30
2 core: 32
1 core: 33

Your CPU has a maximum multiplier of 33 when a single core is active but the maximum will drop down to 30 when all of the cores are active. It can also drop a little lower than this if you exceed the Turbo Boost power limit for your CPU.

With Windows always running, you will frequently have at least 2 cores active so the average multiplier in the log file is probably going to be somewhere between 30 and 32 depending on what you are running.

If the multiplier is stuck at 12 when you are in the middle of a game then that is obviously the sign of a CPU throttling problem. There are 2 types of clock modulation throttling that are typically used on many laptops. The log file will also show that. You should see 2 columns of 100.0 in the log file which shows that your CPU is operating at its Intel designed speed and is not throttling internally. Also have a look at the CPU and GPU temperature columns to make sure neither of these are overheating. You have to turn GPU monitoring on in the options window. This only works if your laptop GPU is supported.

If the numbers in the TPL window are not locked then let me know what those are. Many manufacturers lock this window in the bios so users can not make any adjustments to the Turbo Power Limits. If this is adjustable, you might be able to achieve a little more full load performance by increasing these values but your laptop will run hotter so for your CPU, it is usually not worth it.

Show me some data and ask some specific questions so I can help you out.

 

thanks you very much,ill try out,also how far is i7 2720 can be OC stability??

 

Just ordered a Samsung with i7-3820. I do not believe it is OC'ble in bios, or XTU capable, so will again give TS a nod. Hopefully I can get the +400 MHz and stable...........

 

A Core i7-2720QM can not be overclocked any significant amount using ThrottleStop or any other software. As long as it is not throttling, it's a reasonably fast laptop CPU as is.

TANWare: I don't believe that ThrottleStop 5.00 can overclock the i7-3820QM, yet. I recently purchased a 3570K for testing purposes but without any relevant Intel documentation, I might not ever be able to unlock the extra 400 MHz that is hiding in the 3720QM and 3820QM. I know who I can contact now if I ever get something figured out that needs testing. I was eye balling an Alienware M17x-R4 that has this feature available in the bios. I might have to find a user willing to help to see where the magic XTU register is hiding if there is one.

 

I will be happy with stock clocks but happier with the +400. I don't have the system yet so I am definately jumping the gun. Supposedly with the Clevo's the XTU and +400 are available, Clevo bios is well known for lock downs. Everyone around here has the Samsung with the stock 3610 so it is hard to know what the bios or if even XTU will be available.

More than willing on this end to test whatever. Only thing is the new machine does not use optimus so the IGP will not be accessible. Advantage to giving it over here though is I will have the i7-3820 and C2Q machines to test anything out on...............

 

After much testing with all sorts of settings, system wide and system specific drivers, I finally found what was causing the window popup issue, it was a desktop window manager (not a theming app) that added ThrottleStop to its list and due to the way ThrottleStop works (quite standard, mind you), the main window didn't restore properly. Either removing ThrottleStop from the list or disabling the manager fixed it.

Bottomline is, be mindful of any software that controls how windows are displayed.

unclewebb, are there any new features planned for Sandy Bridge CPUs? I ask because it seems Ivy Bridge allows for more granular control, and I was wondering if any similar situation could be made available.

 

tribaljet: Glad to hear you got that problem figured out.

ThrottleStop 5.00 should work identically on Sandy Bridge and Ivy Bridge CPUs. There are no significant differences, just a die shrink.

I have not done any hands on testing on an Intel 2nd or 3rd Generation Core i CPU so if you see any differences, show me a picture so I can see what is going on.

 

Could anybody help me please

I have the Clevo W370 (Sager Np6370) and the problem apparantly is that the CPU throttles when the GPU is under load. What settings would i have to set exactly to stop this throttling. I think what it disables is turbo boost when the GPU is under load.

Thanks a million!

 

The Clevo W110ER also has this disable Turbo Boost "feature" whenever the Nvidia GPU is active so read the first few posts from this thread.

http://forum.notebookreview.com/sag...np6110-clevo-w110er-first-look-review-96.html

After you get ThrottleStop 5.00 setup, turn on the Log File feature and go play some games, exit ThrottleStop and have a look in the log file to see what multiplier your CPU was using.

You didn't mention what CPU model you have so click on the ThrottleStop TRL button and in that window it will tell you what your default multiplier is and what maximum multipliers you will be seeing when Turbo Boost is working. You can also run a simple benchmark like the built in TS Bench. Play around with ThrottleStop while this benchmark is running. Setting this to 1 thread should give you a good indication of what your maximum multiplier is during normal use. You can adjust the Set Multiplier value up and down while this benchmark is running to better understand what ThrottleStop is all about.

 

I also have a new Sager NP6370 (with an i7-3610 CPU) and I've been noticing some suspicious behavior with it. When I ran a stress test in Furmark and Prime95 with throttling disabled, I was only actually getting about 50% uptime on the turbo. It would be running full blast for maybe around 13 seconds, then the multiplier would drop down to 23 for 13 seconds before going back up. The GPU was at something like 71-73 degrees and the max temp on the CPU was around 91 when the multiplier was maxed.

I even tried gradually reducing the multiplier in search of a clock speed that would be able to maintain 100% uptime, and similar behavior persisted with max temps as low as 87 degrees. This seems extremely odd to me, because the maximum safe temperature on the CPU is supposed to be 105 degrees, according to Intel.

Disabling throttling in Skyrim kept at least a couple of cores at maximum full-time without the temperature ever going above the mid-70s, so for all I know the issue may be entirely academic, but I'd still like to know what the deal is.

 

Do you have access to a Kill-a-Watt meter or similar? Running Prime95 + Furmark is a big load for a laptop. I am wondering if you are pushing your power adapter to its limit with that load.

What is your adapter rated at?

As CPU core temps, etc. go up when stress testing, power consumption can also start to go up. This could lead to it starting to throttle at a lower and lower temperature. That's why it would be interesting to keep an eye on power consumption to see if there is any correlation.

 

I'm not entirely sure how I would go about checking any of those things. I certainly don't have a physical meter I could measure anything with.

Here's some important-looking gibberish printed on my power brick, if it means anything to you:
Model: PA-1131-07 (the brand is Lite-On)
Input: 100-240v ~ 2.15A 50-60Hz
Output: 19V === 6.3A

I just did the Furmark + Prime thing under the assumption that that was the common thing to do when measuring temperatures and the like. Is it typical for a laptop to not be able to handle that load but still be fine under realistic circumstances?

 

Power = I x E

Output Power = 6.3 A x 19 V = 119.7 ~ 120 watts

Ask around in the forums if anyone has a Kill-a-Watt meter. They cost about $15 on EBay. They are not the most accurate devices but give you a general idea about power consumption and I think your stress testing might be pushing up against what your adapter can deliver. Some Dell laptops have been designed so they start throttling before they get too close to the rated power output of the adapter so something similar might be going on here.

Check in the forums for your laptop to see if anyone has a meter or a higher output power supply to test this.

If you have an external monitor, try testing with that with the laptop screen turned off. If you can save a few watts from the screen it might help out. Some older laptops could pull about 180 watts during this kind of stress testing. Not sure how much your laptop would like to use.

Edit: My opinion is that Furmark + Prime 95 is an unrealistic load for a modern laptop. If your laptop can handle this then that is great but if it can't, it still might be 100% usable when running normal games or apps.

 

That 660m with i7-3610 under turbo would probably oversaturate a 120w brick easilly. A Fermi core. like 670m or 675m, would have been even worst. The 120w though is probably fine for either overstressing either the CPU or GPU as you would do in real world usage. as UW mentions. A side point is beware over overclocking the video card you may be towards the limit of what your brick wants to supply cleanly...........................

 

hello unclewebb,

I have some trouble figuring out what is locking down my system.
I have a Macbook Pro 2011 with sandy bridge 2720QM CPU.
here is the screen shot of TS when it is throttling
http://i.imgur.com/m6kmQ.png
CMod and Chip stay at 100% in the log file.
Temperature seems to be fine.
What might be the reason why it is locked down?

 

First off; your temps are throttling your cpu it seems?

Second; you need to check the set multiplier box and also click the turn on button.

Third; you need to be in the maximum performance profile with both the minimum and maximum processor states are 100% (in the processor power management section in the advanced power options).


Also; you're running TS in your MBP in a VM or by dual booting?

 

Yes. It seems like it is the case. But I thought I7 can run to as hot as 100F without damaging the component.

I checked the set multiplier box and turned it on, nothing happened.

Already in performance profile

I run TS on windows 7 64bit using bootcamp (osx isn't running)

Thanks in advance

 

If you're physically booting up into Win7x64 and the cpu has a load on it with both the multipler box checked and TS turned on and it is still staying at 800MHz - then it may be an interplay between the mac hardware and bootcamp software translation that is the issue.

While SNB cpu's can work up to 100 degrees celsius, that doesn't mean they won't trottle to stay at/near that temp.

 

Here are some of the logs

first session
2012-09-24 04:56:11 23.53 5.7 100.0 100.0 0 84 1.2260 10.5
2012-09-24 04:56:12 26.09 6.0 100.0 100.0 0 85 1.1659 10.5
2012-09-24 04:56:13 24.32 6.0 100.0 100.0 0 84 1.2109 10.7
2012-09-24 04:56:14 22.91 5.8 100.0 100.0 0 83 1.1909 10.4
2012-09-24 04:56:15 23.13 6.0 100.0 100.0 0 86 1.1909 10.6
2012-09-24 04:56:16 26.64 5.8 100.0 100.0 0 84 1.1959 10.4
2012-09-24 04:56:17 26.07 6.0 100.0 100.0 0 83 1.2109 10.6
2012-09-24 04:56:18 26.71 6.7 100.0 100.0 0 83 1.1659 11.2
2012-09-24 04:56:19 26.56 5.7 100.0 100.0 0 83 1.1659 10.6
2012-09-24 04:56:20 22.53 6.2 100.0 100.0 0 82 1.1959 11.3
2012-09-24 04:56:21 22.24 8.3 100.0 100.0 0 84 1.1659 13.2
2012-09-24 04:56:22 26.92 6.9 100.0 100.0 0 83 1.1659 11.5
2012-09-24 04:56:23 21.53 6.6 100.0 100.0 0 82 1.1959 11.5
2012-09-24 04:56:24 22.50 6.0 100.0 100.0 0 82 1.1659 10.6
2012-09-24 04:56:25 23.19 5.7 100.0 100.0 0 81 1.1909 10.1
2012-09-24 04:56:26 23.96 5.9 100.0 100.0 0 83 1.1659 10.4
2012-09-24 04:56:27 26.03 5.8 100.0 100.0 0 83 1.1959 10.2
2012-09-24 04:56:28 23.97 6.1 100.0 100.0 0 82 1.1659 10.6
2012-09-24 04:56:29 26.73 6.0 100.0 100.0 0 82 1.1659 10.4
2012-09-24 04:56:30 23.70 6.2 100.0 100.0 0 83 1.1659 10.6
2012-09-24 04:56:31 26.13 5.9 100.0 100.0 0 83 1.1959 10.4

Second session
2012-09-24 11:36:35 23.01 7.3 100.0 100.0 0 74 1.1909 9.1
2012-09-24 11:36:36 16.79 7.2 100.0 100.0 0 75 1.0458 9.7
2012-09-24 11:36:37 16.44 8.2 100.0 100.0 0 74 0.9857 9.2
2012-09-24 11:36:38 19.26 9.9 100.0 100.0 0 75 1.0057 8.8
2012-09-24 11:36:39 17.68 12.6 100.0 100.0 0 75 0.9557 9.5
2012-09-24 11:36:40 14.36 17.1 100.0 100.0 0 75 0.9006 9.3
2012-09-24 11:36:41 12.46 16.9 100.0 100.0 0 75 0.8706 8.2
2012-09-24 11:36:42 12.48 15.2 100.0 100.0 0 74 0.8806 7.7
2012-09-24 11:36:43 13.86 17.1 100.0 100.0 0 74 0.8706 8.9
2012-09-24 11:36:44 10.87 29.8 100.0 100.0 0 74 0.7755 9.3
2012-09-24 11:36:45 8.00 40.4 100.0 100.0 0 74 0.7755 8.1
2012-09-24 11:36:46 8.00 30.1 100.0 100.0 0 74 0.7655 7.5
2012-09-24 11:36:47 8.00 22.3 100.0 100.0 0 74 0.7705 6.7
2012-09-24 11:36:48 8.00 22.8 100.0 100.0 0 74 0.7605 6.8
2012-09-24 11:36:49 8.00 24.7 100.0 100.0 0 74 0.7755 7.0
2012-09-24 11:36:50 8.00 22.9 100.0 100.0 0 73 0.7655 6.7
2012-09-24 11:36:51 8.00 21.4 100.0 100.0 0 74 0.7655 6.7
2012-09-24 11:36:52 8.00 22.7 100.0 100.0 0 74 0.7755 6.8
2012-09-24 11:36:53 8.00 26.2 100.0 100.0 0 74 0.7755 7.0
2012-09-24 11:36:54 8.00 27.3 100.0 100.0 0 73 0.7755 7.1
2012-09-24 11:36:55 8.00 28.4 100.0 100.0 0 73 0.7705 7.2
2012-09-24 11:36:56 8.00 21.6 100.0 100.0 0 74 0.7655 6.6
2012-09-24 11:36:57 8.00 21.1 100.0 100.0 0 73 0.7655 6.6
2012-09-24 11:36:58 8.00 24.0 100.0 100.0 0 74 0.7655 6.8
2012-09-24 11:36:59 8.00 29.7 100.0 100.0 0 74 0.7755 7.5
2012-09-24 11:37:00 8.00 27.0 100.0 100.0 0 75 0.7705 7.0
2012-09-24 11:37:01 8.00 36.8 100.0 100.0 0 75 0.7755 8.0
2012-09-24 11:37:02 8.00 29.5 100.0 100.0 0 74 0.7755 7.2
2012-09-24 11:37:03 8.00 24.6 100.0 100.0 0 74 0.7705 6.9
2012-09-24 11:37:04 8.00 21.5 100.0 100.0 0 73 0.7705 6.6
2012-09-24 11:37:05 8.00 23.1 100.0 100.0 0 73 0.7755 6.7
2012-09-24 11:37:06 10.09 25.4 100.0 100.0 0 75 0.9106 7.8
2012-09-24 11:37:07 18.58 17.1 100.0 100.0 0 76 1.1008 11.3
2012-09-24 11:37:08 27.09 10.0 100.0 100.0 0 79 1.1909 14.5
2012-09-24 11:37:09 15.36 31.1 100.0 100.0 0 75 0.7805 14.2
2012-09-24 11:37:10 8.11 22.7 100.0 100.0 0 74 0.7655 6.8
2012-09-24 11:37:12 8.00 23.5 100.0 100.0 0 74 0.7655 6.7
2012-09-24 11:37:13 8.00 23.6 100.0 100.0 0 75 0.7755 6.8
2012-09-24 11:37:14 8.00 22.5 100.0 100.0 0 74 0.7605 6.8
2012-09-24 11:37:15 8.00 20.6 100.0 100.0 0 73 0.7605 6.5
2012-09-24 11:37:16 8.00 20.9 100.0 100.0 0 74 0.7755 6.5
2012-09-24 11:37:17 8.00 22.3 100.0 100.0 0 74 0.7705 6.7
2012-09-24 11:37:18 8.00 21.4 100.0 100.0 0 74 0.7755 6.6
2012-09-24 11:37:19 8.00 23.0 100.0 100.0 0 74 0.7605 6.8
2012-09-24 11:37:20 8.00 21.4 100.0 100.0 0 74 0.7755 6.7
2012-09-24 11:37:21 8.00 23.6 100.0 100.0 0 75 0.7705 6.8
2012-09-24 11:37:22 8.00 20.0 100.0 100.0 0 74 0.7755 6.6
2012-09-24 11:37:23 8.00 20.5 100.0 100.0 0 73 0.7755 6.5
2012-09-24 11:37:24 8.00 20.6 100.0 100.0 0 74 0.7755 6.5
2012-09-24 11:37:25 8.00 20.3 100.0 100.0 0 74 0.7655 6.6
2012-09-24 11:37:26 8.00 20.8 100.0 100.0 0 74 0.7755 6.6
2012-09-24 11:37:27 8.00 21.1 100.0 100.0 0 74 0.7755 6.6
2012-09-24 11:37:28 8.00 19.6 100.0 100.0 0 74 0.7655 6.5
2012-09-24 11:37:29 8.00 21.1 100.0 100.0 0 74 0.7605 6.7
2012-09-24 11:37:30 8.00 22.6 100.0 100.0 0 75 0.7605 6.8
2012-09-24 11:37:31 8.00 20.2 100.0 100.0 0 75 0.7655 6.6
2012-09-24 11:37:32 8.00 19.8 100.0 100.0 0 74 0.7655 6.5
2012-09-24 11:37:33 8.00 19.6 100.0 100.0 0 75 0.7705 6.5
2012-09-24 11:37:34 8.00 21.7 100.0 100.0 0 74 0.7655 6.7
2012-09-24 11:37:35 8.00 21.4 100.0 100.0 0 74 0.7605 6.6
2012-09-24 11:37:36 8.00 22.1 100.0 100.0 0 74 0.7755 6.7
2012-09-24 11:37:37 8.00 23.4 100.0 100.0 0 74 0.7755 6.8
2012-09-24 11:37:38 8.00 21.7 100.0 100.0 0 74 0.7755 6.7
2012-09-24 11:37:39 8.00 20.1 100.0 100.0 0 73 0.7605 6.5
2012-09-24 11:37:40 8.00 19.7 100.0 100.0 0 74 0.7655 6.4
2012-09-24 11:37:41 8.00 22.0 100.0 100.0 0 74 0.7605 6.7
2012-09-24 11:37:42 8.00 21.7 100.0 100.0 0 74 0.7605 6.7
2012-09-24 11:37:43 8.00 24.6 100.0 100.0 0 74 0.7705 6.9
2012-09-24 11:37:44 8.00 23.8 100.0 100.0 0 75 0.7705 6.9
2012-09-24 11:37:45 8.00 25.8 100.0 100.0 0 75 0.7705 7.0
2012-09-24 11:37:46 8.00 21.5 100.0 100.0 0 75 0.7655 6.7
2012-09-24 11:37:47 8.00 19.3 100.0 100.0 0 73 0.7605 6.5
2012-09-24 11:37:48 8.00 20.1 100.0 100.0 0 74 0.7705 6.5
2012-09-24 11:37:49 8.00 20.4 100.0 100.0 0 75 0.7655 6.6
2012-09-24 11:37:50 8.00 21.0 100.0 100.0 0 74 0.7655 6.6
2012-09-24 11:37:51 8.00 18.7 100.0 100.0 0 75 0.7605 6.3
2012-09-24 11:37:52 8.00 23.2 100.0 100.0 0 75 0.7755 6.8
2012-09-24 11:37:53 8.00 26.6 100.0 100.0 0 75 0.7755 6.9
2012-09-24 11:37:54 8.00 23.9 100.0 100.0 0 74 0.7605 6.8
2012-09-24 11:37:55 8.00 12.2 100.0 100.0 0 73 0.7655 5.6
2012-09-24 11:37:56 8.00 11.6 100.0 100.0 0 73 0.7605 5.5
2012-09-24 11:37:57 8.00 14.3 100.0 100.0 0 74 0.7705 5.8
2012-09-24 11:37:58 8.00 21.1 100.0 100.0 0 74 0.7655 6.5
2012-09-24 11:37:59 8.00 14.9 100.0 100.0 0 74 0.7605 5.8
2012-09-24 11:38:00 8.00 19.0 100.0 100.0 0 74 0.7605 6.3
2012-09-24 11:38:01 8.00 15.9 100.0 100.0 0 74 0.7655 5.9
2012-09-24 11:38:02 8.00 18.8 100.0 100.0 0 74 0.7755 6.3
2012-09-24 11:38:03 8.00 15.9 100.0 100.0 0 73 0.7605 6.0
2012-09-24 11:38:04 8.00 12.5 100.0 100.0 0 72 0.7605 5.6
2012-09-24 11:38:05 8.00 12.8 100.0 100.0 0 73 0.7605 5.6
2012-09-24 11:38:06 8.00 12.8 100.0 100.0 0 72 0.7605 5.6
2012-09-24 11:38:07 8.00 12.8 100.0 100.0 0 73 0.7755 5.6
2012-09-24 11:38:08 8.00 12.9 100.0 100.0 0 73 0.7605 5.7
2012-09-24 11:38:09 8.00 11.8 100.0 100.0 0 73 0.7705 5.5
2012-09-24 11:38:10 8.00 13.1 100.0 100.0 0 73 0.7605 5.7
2012-09-24 11:38:11 8.00 12.7 100.0 100.0 0 73 0.7605 5.6
2012-09-24 11:38:12 8.00 11.8 100.0 100.0 0 72 0.7605 5.5
2012-09-24 11:38:13 8.00 12.8 100.0 100.0 0 72 0.7705 5.6
2012-09-24 11:38:14 8.00 12.9 100.0 100.0 0 72 0.7605 5.6
2012-09-24 11:38:15 8.00 12.1 100.0 100.0 0 73 0.7655 5.6
2012-09-24 11:38:16 8.00 11.8 100.0 100.0 0 72 0.7605 5.5
2012-09-24 11:38:17 8.00 11.9 100.0 100.0 0 72 0.7605 5.5
2012-09-24 11:38:18 8.00 11.5 100.0 100.0 0 72 0.7605 5.5
2012-09-24 11:38:19 8.13 11.8 100.0 100.0 0 72 0.7855 5.6
2012-09-24 11:38:20 10.16 10.8 100.0 100.0 0 72 0.8656 6.0
2012-09-24 11:38:21 11.71 8.6 100.0 100.0 0 74 0.9457 6.5
2012-09-24 11:38:22 11.92 7.3 100.0 100.0 0 73 1.0258 7.2
2012-09-24 11:38:23 13.53 6.8 100.0 100.0 0 73 1.1659 7.9
2012-09-24 11:38:24 12.40 6.1 100.0 100.0 0 73 1.1659 8.7
2012-09-24 11:38:26 12.26 6.1 100.0 100.0 0 74 1.1659 8.9



I still don't find the correlation between temp and when cpu is being throttled. In session 1, CPU temp was even higher at 85C but CPU doesn't seem be throttled. But in session 2 CPU is throttled even at 75C and it is on and off at random times it seems.

 

Can you post another screen shot of ThrottleStop while it is throttling.

Make sure BD PROCHOT is not checked. BD PROCHOT stands for bi-directional processor hot. Basically this is a signal path to the CPU that lets other items on the motherboard like the voltage regulator circuit tell the CPU that it is too hot which forces it down to the 8 multiplier. This can override the CPU so even if the CPU is really not too hot, it will throttle like this anyhow.

If BD PROCHOT is being used to protect something else on your motherboard then turning this off might damage something. Playing with this is at your own risk. Seeing a steady 8 multiplier while the CPU is barely loaded is usually a good sign that BD PROCHOT is the problem. It's usually a bug and is likely being triggered when running Mac OS too. If you ever notice variable performance in Mac OS, this is likely why.

Edit: One more thing. If your CPU is over 70C when it is barely above an idle then Apple needs to learn how to design and include a proper cooling solution when installing a hyper threaded 4 core CPU into their laptops. That temperature is ridiculous. There is no way you will ever be able to run your CPU fully loaded for any length of time with that crappy of a cooling solution.

When posting log files, try to use Pastebin.com and then post a link here or use the code HTML tags before and after your data.

[code*]
log data
[/code*]

Remove the * in each code tag.

 

Isn't this how passive cooling (quiet) policy should be implemented? Perhaps there is such a setting that was set on that laptop that will keep the fans at bay for as long as possible. i.e. the idle temp maybe @ 70°C, which is fine if the fans are not spinning...

 

That's not Apple implementing passive cooling - it is how it implements aggresive replacement of MBP's.

 

When a Core i7-2720QM is properly cooled, it should be able to run fully loaded at close to 3000 MHz, without ever thermal throttling or slowing down.

I am all for a quiet, passively cooled laptop as long as it can run at its rated speed. I don't think chasely's laptop is ever going to be able to do that even if he is able to solve his current problem.

 

Out of curiosity, going by Intel's specs, my i7-2820QM should do severe throttling at 100C, yet it doesn't seem to do so. Due to that behavior, performance is acceptable but both load temps are always between 80-90+ and when both the CPU and GPU are under load, temps are between 95-100. I know that having a single cooling solution for both the CPU and GPU isn't good at all, but higher end models were out of my budget. What bothers me most about these temps is that the laptop was supposed to have come with a custom cooling solution from the shop I bought it from, yet temps seem to be above average for high single and combined loads, which leads me to suspect that maybe the cooling system isn't properly secured or thermal paste wasn't applied properly.

In any case, I attached a screenshot of the temps on the main ThrottleStop window from my current Windows session that had already fully loaded the CPU alone as well as CPU and GPU combined.

 

Should I bring over the frying pan, eggs and bacon? Holy cow..............

 

Basically, anything over 95C means both CPU and GPU are under full load. CPU alone (during summer) idles at 59-65 and under load is at 75-90+.

Never got a forced shutdown so far, only a couple system freezes due to too high a GPU OC.

EDIT: The CPU runs happily at 2.4-3.1GHz, it usually only goes below stock 2.3GHz when it's doing very little, and when it isn't doing anything at all it sits at 0.8-1.6GHz.

 

As long as your heatsink hasn't fallen off, thermal throttling should not be severe. When an Intel 2nd Generation mobile CPU reaches 100C, it will only throttle just enough to keep the temperature from increasing further. If you see severe throttling, that's usually because some guy writing the bios had a great idea to create their own throttling scheme that technically speaking, isn't in the same league compared to the sophistication of Intel's thermal throttling.

I disconnected the heatsink fan on my desktop E8400 and ran Prime 95 Small FFTs for 3 hours. The core temperature basically flat lined, just under the thermal throttling temperature. I think the Core i CPUs are even better designed and can run another hair or two closer to the 100C spec.

http://img11.imageshack.us/img11/276/hote8400fw5.png

When throttling, the multiplier is adjusted hundreds of times a second to maximize performance without going over 100C so a user will barely notice any slow down. If the temperature can not be controlled, before it becomes completely out of hand, the CPU will slow down further and it will likely become more noticeable then.

While your CPU is throttling, the ThrottleStop reported multiplier will show you exactly how much performance you are losing. For mild throttling, you don't lose much performance at all. If you do any more full load testing, post a log file from ThrottleStop.

 

You misread me, I said that it didn't do any sort of severe throttling, much to my surprise

In any case, I've ran a 10min full load test and while not the most stressing test, it's a usual scenario, but this time the CPU didn't reach that high temps as usual.

Here is the log: ThrottleStop 5.00a 25_09_2012 Log - Pastebin.com

EDIT: Here is how ThrottleStop's window reported temps: http://i.imgur.com/LSlnh.png

 

Sorry about that. That's not the first time I've misread a post in a forum.

Your log file is an excellent example of how an Intel CPU is supposed to work. Even with the core temperature hitting 97C to 98C, it is still running at full speed while continuing to use full Turbo Boost.

Your log is a lot better looking than the log at the top of page 133.

http://forum.notebookreview.com/har...t-upgrades/531329-throttlestop-guide-133.html

Locking a CPU to 800 MHz when the core temperature is nowhere near the Intel specified throttling temperature is either a failure on the motherboard or a failure in design.

 

I have the i7 3610QM. I set the multiplier to turbo but it constantly locks it at the turbo clock speed (3.3GHz). Is there a way to make it so it varies instead of constantly being at 3.3GHz?

 

zb1234: The Core i7-3610QM only uses the 33 multiplier when a single core is active. Post a ThrottleStop log file so I can have a look at your multiplier and also post a screen shot of ThrottleStop so I can see how you have the program setup.

The Intel Turbo Boost feature is designed so cores are either asleep and consuming very little power or running as fast as possible. The reason for this is that a fast CPU is also an efficient CPU. Here's a good paper that explains this.

Power Optimization – a Reality Check
http://www.cs.berkeley.edu/~krioukov/realityCheck.pdf

A fast CPU gets its work done quickly which allows it to spend a bigger percentage of time in one of the low power sleep states like C3/C6. Doing this reduces overall power consumption. Have a look in the ThrottleStop C States window and make sure your CPU is using the deeper sleep states when it is idle. Eliminate any background tasks that are preventing your CPU from going to sleep. At idle, if a core is spending over 90% of its time in the C6 sleep state, whatever multiplier it is at is not that important.

You can use ThrottleStop to disable Turbo Boost or to run your CPU at a reduced multiplier. A slower CPU will decrease peak power consumption but it probably won't decrease overall power consumption.

 

Hi unclewebb,

If you remember me, we had pleasant conversation over e-mail some time ago. Anyway, I've been using ThrottleStop on every computer I can install it and there are no problems. It undervolts beautifully on C2D P7350 and Atom N450.

I even use it on desktop (Asus P8Z68V/Gen3 & I7-2600K) for the following reasons:
- I think it gives most accurate thermal monitoring (at least it should be on par with RealTemp, right?). And it's always there in the tray for a quick check, and I can unminimize it any time for a quick inspection of CPU usage.
- While most of the settings don't make much difference, I can achieve steady 16X multiplier on idle only if I check Power Saver checkmark. Then it reduces the VID also. I am absolutely certain this makes a difference, the temps are couple of degrees lower and it says the Wattage is lower. But the downside is that the computer is a little more sluggish (I can sense it).
- I think (maybe I'm wrong) that ThrottleStop's multiplier management is superior to the one built in Windows or the processor itself - it appears to me that it is faster and reacts to change of load with greater agility. When overclocking, this might be the straw that breaks the camel's back - adding more stress to VRM and/or exposing it's weaknesses. Again, I'm not certain about this and I haven't made any tests. But I'm also nearly certain that TS's light activity can also contribute to the overal activity when testing with, i.e. Prime95 (it adds diferent kind of stress).

So, I have many more questions about fringe situations like this, but can't think of a concrete example right now (and also, they pertain more to overclocking desktop boards than TS usage). OK, how about:

- Does "more data" checkmark also influence the rate of multiplier management or it's the rate of monitoring only? Also, how would you quantify it's CPU usage ("more data" vs no "more data")? I ask this because of the case I mentioned, overclocked to the max and don't want to put any light load more than necessary (of course, disabled most of the Windows background processes). But I'm always fascinated with the beauty of the numbers dancing in front of me :smile: so I cant force myself to uncheck it.
- Would "Turbo Ratio Limits" do anything on a desktop? I overclock it "by all cores" so I think turbo isn't active. But I admit, I generally I don't understand this feature.

Again, thank you for the wonderful program!

GL

 

Hi unclewebb

First excuse my english. I installed your Throttlestop in my Clevo W150ERM. The computer has a Nvidia 650M and I wanted to use nvidiainspector too. I will tell you what happened to me because I can't understand it.


Without TS and nvidia inspector 3dmark 11 goes to 2260-2080 points that is a normal value,

With TS and ONLY UNCHECK BD prochot (i7 working only at 2,3ghz) and nvidiainspector with 650M at 1150core/2800memory goes to 2987-3063).

This is what I don't understand. With TS with BD prochot unchecked, multiplier Turbo checked (3,3ghz full load) and 650M overclocked I receive the same 3dmark11 points around 3000 (not more than 3030-3060).

I don't understand it. Also I have seen that without overclocking the 650M I receive the same points in 3dmark11 unchecking BD prochot, checking it and/or cheching multiplier turbo (all only around 2280 points).....

any idea?

thanks in advance

 

With the new system XTU has everything greyed out. Running 5.0 TS unfortunately gets me nothing more than the stock setting, but this was expected. on 3dmark06 I get 6900 + for the cpu mark and turbo keeps all cores at x33 or above from what I can see. I rarely ever see anything doing the x37 but usually bounce between x35 and x36. Highest sustained usually is around 43watts. the power just kicks my old Q9200's butt all over the place, even without the +400.

 

Unfortunately, XTU also has adjustments locked on my CPU, only monitoring is available.

 

aptec: If your laptop does not use the BD PROCHOT function or if your laptop is not throttling; having BD PROCHOT checked or not checked does not matter. It won't fix a problem if you don't have a problem.

The Clevo W150ERM turns off Intel Turbo Boost when the GPU is being used. Using ThrottleStop to turn Turbo Boost on will improve performance for any CPU dependent game. I don't play 3DMark11. Run a ThrottleStop log file which this benchmark is running. If Turbo Boost is fully working then ThrottleStop is doing its job. Most Clevo owners are happy with the increase in performance when playing any CPU demanding games.

Intel multiplier management is far superior to anything that ThrottleStop can do. If you use the Windows High Performance profile and the CPU's C3 and C6 sleep states, you will get a good balance between performance and low power consumption.

Here's an example of a 3570K, idle at the desktop.

http://img26.imageshack.us/img26/8351/powertb.png

Individual cores of this CPU are spending 98% to 99% of their time in the C6 sleep state. Whether the multiplier is high or low doesn't matter. Whether the VID is high or low doesn't matter because in C6, the CPU drops down to a lower VID anyhow. Power consumption at the wall is the same 35 watts whether I go high or low or use a Windows Balanced profile or use the Windows High Performance profile. The ThrottleStop - Power Saver setting changes the Multiplier and VID but on this Core i CPU, it doesn't save a single watt at idle.

 

For Uc, don't have my web publishing stuff installed and it was too big for a PM...........

 

Thanks TANWare. Your data is most appreciated. I have some time off this weekend so hopefully I can get something figured out.

 

I thought TS did some multiplier management because it makes a difference here (and I think on my C2D laptop, but I can't check now).

Before:


After:


The difference is only a ticked "Power saver" checkmark. The PC is set on a High Performance plan (100%/100%) and all power saving features are enabled in BIOS. Please note the VID. And the difference in Watts is usually somewhat higher, I caught a bad moment here.

Edit: Now that I think about it, maybe "power saver" acts as setting the power plan to balanced (lowers the minimum power state)?

Edit2: And the temperature dropped, now I saw it. It usually does.

GL

 

GrofLuigi: Thanks for the info.

Intel includes a register in their Core i 2nd and 3rd Generation CPUs which keeps track of power consumption. ThrottleStop and other monitoring applications report this data but it is not the same as the actual power consumption of the CPU. Its purpose within the CPU is to control the Turbo Boost feature. It is only an approximation based on the VID voltage and during some situations, a change in this value might not have anything to do with a change in the actual CPU power consumption.

Using ThrottleStop to force the multiplier to 16 reduces the VID voltage and this is going to reduce the reported power consumption. This might not make any significant difference to actual power consumption though because the individual cores are already spending almost all of their time in the low power C6 sleep state. When a core drops down to C6, the voltage is reduced internally. Monitoring programs like CPU-Z can report reasonably accurate core voltage on a desktop motherboard but they can not report the actual voltage of a core that is in C6.

To find out if ThrottleStop is really making a difference, I would trust core temperature data a lot more than power consumption data. The only problem with the core temperature data is that these sensors were only designed and calibrated by Intel to control thermal throttling and thermal shut down at approximately 100ºC to 125ºC. The data coming from these sensors is not 100% accurate, especially when a CPU is lightly loaded and is a long ways away from the calibration temperature.

I don't know of any software that can 100% accurately measure CPU power consumption and my $15 Kill-a-Watt meter is not accurate enough for this kind of testing either. The only way to do this correctly is to tap into the power line that feeds the CPU socket and measure the power consumption directly with some expensive test equipment.

The 1ºC drop in core temperature that your screen shots show makes me want to believe that ThrottleStop has reduced the actual power consumption some tiny amount but the difference is probably a lot less than the 1.2W difference that ThrottleStop is reporting.

One big negative of the Power Saver function is that it prevents the CPU from getting back up to full speed quickly enough. Try doing some 1M Super Pi mod testing and it will probably show less performance without much, if any, savings in actual power. This feature was designed for the early 65nm Core 2 processors. For Core i processors, I think you are better off to let the CPU manage itself.

Download Super PI Mod v1.5 | techPowerUp

 

unclewebb, is there any reason why there is very little C6 percentage, the bulk being C3 and C2? Windows power plan has minimum and maximum CPU usage set to 100%.

Here is a screenshot of ThrottleStop's main window and C State window:

 

@unclewebb:

Yes, power saver probably doesn't make a real difference, and the computer is sluggish.







But for testing single threads, I think Prime 95 is better:



In this screenshot, I've set it to 4 workers. Again I've caught a bad moment, but 99% of the time, the load was 100% on every other core. Prime95 is aware of hyperthreading, it says so in the main window.

With both programs, on single thread load, the highest c0% core rotates nicely among the cores.

If you're wondering why the multiplier is sometimes above 44, I've changed overclocking in BIOS to "by per core" and it gives me a nice boost in light loads. It's set up like this:



ThrottleStop can't do it, I just show it as reference. I couldn't get it to 44-45-46-47 stable without upping the voltage. And the boost is really nice at light loads (which is most of the time that everyone spends on the desktop), while the 4-core multiplier has been proven stable with lots of testing before. The rotating by the windows scheduler (or the processor's) gives me assurance that nothing bad could happen. I wonder why the overclocking guides don't recommend this method more. Maybe because of n-times more testing?

GL

 

tribaljet: Your CPU is processing a lot of background junk. If that is a screen shot of when you think your computer is idle, it is not. That's why it is being prevented from using the C6 sleep state. Open up the task manager and find out what programs are running on your computer and keeping your CPU active. After a clean Windows install, individual cores should be able to spend over 95% of their time in the C6 sleep state when idle. Every time you install a new program to your computer you should have a look to make sure that it is not loading your CPU when idle. Keeping your system clean can make a big difference to battery run time.

GrofLuigi: If you are using the ThrottleStop - Power Saver feature, you need to adjust the Power Saver C0% number in the Options window. By default it is set to 35 but for a Quad Core CPU with hyper threading, you should probably set that lower. This should reduce the sluggishness when adjusted correctly. I never use the Power Saver feature so I haven't played with it very much. I think modern CPUs are quite capable of taking care of themselves.

 

I'm very careful about what tasks and services are running and I simply cannot dismiss any of them. Can apps like CoreTemp be causing that usage due to active monitoring?

 

Wow, I totally forgot about that. I experimented with it a lot (but really a lot) on my C2D P7350 laptop at the time and concluded that 35 is perfect. I've tried 15 and 10 now and it's much more responsive. I hope the VRM circuitry will not complain about the increased stress at the workplace.

Aaand another question, if I may... I appologize to everyone and especially to Unclewebb if I'm becoming boring, but I hope everyone (or at least some of us) can learn something new here...

What about TDC/TDP? I have a growing suspicion that they are the thing that starts to limit my overclock (besides the mediocre Hyper212+). This is how it looks like in TS:



But the Intel specs says it's 95W. Or that's only with the stock cooler?

I'm not trying to do anything through TS, I just don't know what to set it to in BIOS. I've seen recommendations to pump it up to around 200 for I7-2600K, but that seems like very dangerous to me. And I found even less about TDC.

* Edit: I found something more about TDP/TDC, but I still can't find the default values from authoritative source. And then more questions rise, like: what are the default long duration limits etc; does CPU current capability (100%-140%) limit in BIOS apply to the default value or to the modified value. Etc. etc. I appologize that these are not notebook questions and if necessary, I will take them elsewhere.

GL

 

*** delete ***