The ThrottleStop Guide

Hi,

would it be possible to add the Limit Reasons to a log, or at least the information that throttling took place (as this info is already available in the checkmark in throttlestop)? I am trying to figure out at what exact times a system runs into thermal throttling over a longer period of time. I am looking for something like
[time1] Throttling started Reason: PL1
[time2] Throttling stopped Reason: PL1

Unfortunately there seems to be no tool that can log these information, so it would be cool to see this functionality in ThrottleStop.

 

Funny, I was experiencing a mistérios throttling, and couldn't pinpoint because I couldn't look at LimitReasons quickly enough.

And for some reason, I wasn't bright enough to come up with a logging suggestion! Ha, nice one

 

For thermal throttling, all you have to do is look at the temperature column in the log file.



The main screen of ThrottleStop shows you what temperature the laptop manufacturer has set the thermal throttling temperature to so if the log file shows the temps getting up to that level then it is thermal throttling.

I agree that logging of the Limit Reasons data could be a useful and unique feature. I am thinking about adding this as an optional feature that you can toggle on and off when needed. Something simple like just adding VR, PL1, PL2, EDP, etc. to the end of each line of the log file. When combined with the More Data option, this should give you a good idea of exactly when and why your CPU has gone into gimp mode. I will probably add this later today.

Edit - Just testing out the new Limit Reasons logging feature. I like it!

 

ThrottleStop 8.00 beta 5
https://www.sendspace.com/file/p1q40a

New Features:
- added Limit Reasons reporting to the log file.
- fixed Atom Braswell power consumption reporting.
- another Skylake tweak.

Limit Reasons Definitions
-------------------------
XPRO - External PROCHOT
TEMP - Thermal Event
GT - Intel GPU
VR - Voltage Regulator
EDP - Electrical Design Point (current)
PP0 - Core Power Limit
PL1 - Long Term Power Limit
PL2 - Short Term Power Limit

Here is an example of the new logging feature in action.

Code:

   DATE       TIME    MULTI   C0%   CKMOD  CHIPM   BAT_mW  TEMP  GPU     VID   POWER
2015-11-01  19:34:16  35.65    1.1  100.0  100.0        0   40     0   1.0458    2.9
2015-11-01  19:34:17  34.64    1.2  100.0  100.0        0   41     0   1.0457    2.4
2015-11-01  19:34:18  35.53    1.2  100.0  100.0        0   42     0   1.0457    2.7
2015-11-01  19:34:19  35.71    1.3  100.0  100.0        0   41     0   1.0457    2.7
2015-11-01  19:34:20  35.38    1.2  100.0  100.0        0   41     0   1.0458    2.9
2015-11-01  19:34:21  35.19    1.0  100.0  100.0        0   41     0   1.0458    2.4
2015-11-01  19:34:23  30.42   44.1  100.0  100.0        0   76     0   0.9879   23.2   PL2
2015-11-01  19:34:24  29.02   99.8  100.0  100.0        0   78     0   0.9639   56.0   PL2
2015-11-01  19:34:25  29.00   99.8  100.0  100.0        0   80     0   0.9639   56.6   PL2
2015-11-01  19:34:26  29.04   99.8  100.0  100.0        0   80     0   0.9639   55.5   PL2
2015-11-01  19:34:27  29.03   99.9  100.0  100.0        0   81     0   0.9639   56.0   PL2
2015-11-01  19:34:28  29.01   99.8  100.0  100.0        0   83     0   0.9591   56.0   PL2
2015-11-01  19:34:29  29.05   99.8  100.0  100.0        0   83     0   0.9591   56.2   PL2
2015-11-01  19:34:29  29.07   99.8  100.0  100.0        0   83     0   0.9591   55.7   PL2
2015-11-01  19:34:30  29.03   99.8  100.0  100.0        0   84     0   0.9832   56.6   PL2
2015-11-01  19:34:31  29.02   99.8  100.0  100.0        0   84     0   0.9591   56.0   PL2
2015-11-01  19:34:33  29.02   99.9  100.0  100.0        0   85     0   0.9591   56.0   PL2
2015-11-01  19:34:34  29.01   99.8  100.0  100.0        0   87     0   0.9591   56.0   PL2
2015-11-01  19:34:35  29.02   99.8  100.0  100.0        0   87     0   0.9832   55.7   PL2
2015-11-01  19:34:36  29.02   99.8  100.0  100.0        0   88     0   0.9591   56.0   PL2
2015-11-01  19:34:37  29.01   99.8  100.0  100.0        0   88     0   0.9591   56.0   PL2
2015-11-01  19:34:38  28.98   99.8  100.0  100.0        0   88     0   0.9634   56.3   PL2
2015-11-01  19:34:39  28.97   99.8  100.0  100.0        0   90     0   0.9591   55.7   PL2
2015-11-01  19:34:40  28.97   99.8  100.0  100.0        0   88     0   0.9591   55.7   PL2
2015-11-01  19:34:40  28.97   99.8  100.0  100.0        0   90     0   0.9591   56.0   PL2
2015-11-01  19:34:41  28.96   99.8  100.0  100.0        0   90     0   0.9591   56.0   PL2
2015-11-01  19:34:43  28.96   99.8  100.0  100.0        0   92     0   0.9591   56.0   PL2
2015-11-01  19:34:44  28.99   99.8  100.0  100.0        0   91     0   0.9591   56.0   PL2
2015-11-01  19:34:45  28.98   99.8  100.0  100.0        0   92     0   0.9591   55.8   PL2
2015-11-01  19:34:46  28.94   99.8  100.0  100.0        0   92     0   0.9591   56.0   PL2
2015-11-01  19:34:47  28.98   99.8  100.0  100.0        0   92     0   0.9542   56.0   PL2
2015-11-01  19:34:48  28.98   99.8  100.0  100.0        0   93     0   0.9832   56.0   PL2
2015-11-01  19:34:49  28.97   99.8  100.0  100.0        0   93     0   0.9542   56.0   PL2
2015-11-01  19:34:49  28.96   99.8  100.0  100.0        0   92     0   0.9585   56.6   PL2
2015-11-01  19:34:50  28.95   99.8  100.0  100.0        0   92     0   0.9542   56.0   PL2
2015-11-01  19:34:51  28.97   99.8  100.0  100.0        0   92     0   0.9591   56.0   PL2
2015-11-01  19:34:53  28.97   99.8  100.0  100.0        0   94     0   0.9542   56.0   PL2
2015-11-01  19:34:54  28.98   99.8  100.0  100.0        0   94     0   0.9542   56.3   PL2
2015-11-01  19:34:55  28.99   99.8  100.0  100.0        0   93     0   0.9591   55.5   PL2
2015-11-01  19:34:56  28.97   99.8  100.0  100.0        0   93     0   0.9341   56.0   PL2
2015-11-01  19:34:57  28.98   99.8  100.0  100.0        0   93     0   0.9542   56.0   PL2
2015-11-01  19:34:58  28.99   99.8  100.0  100.0        0   93     0   0.9542   56.0   PL2
2015-11-01  19:34:59  28.98   99.8  100.0  100.0        0   94     0   0.9585   56.0   PL2
2015-11-01  19:34:59  28.99   99.8  100.0  100.0        0   93     0   0.9542   55.7   PL2
2015-11-01  19:35:00  28.97   99.8  100.0  100.0        0   94     0   0.9542   56.0   PL2
2015-11-01  19:35:02  28.98   99.8  100.0  100.0        0   95     0   0.9542   56.0   PL2
2015-11-01  19:35:03  29.00   99.9  100.0  100.0        0   95     0   0.9542   56.0   PL2
2015-11-01  19:35:04  28.98   99.9  100.0  100.0        0   93     0   0.9542   56.0   PL2
2015-11-01  19:35:05  28.97   99.8  100.0  100.0        0   94     0   0.9542   56.6   PL2
2015-11-01  19:35:06  29.01   99.8  100.0  100.0        0   94     0   0.9542   56.0   PL2
2015-11-01  19:35:07  28.96   99.8  100.0  100.0        0   95     0   0.9542   56.0   PL2
2015-11-01  19:35:08  28.97   99.8  100.0  100.0        0   94     0   0.9542   56.0   PL2
2015-11-01  19:35:09  28.96   99.8  100.0  100.0        0   95     0   0.9542   56.0   PL2
2015-11-01  19:35:09  28.95   99.8  100.0  100.0        0   95     0   0.9542   55.8   PL2
2015-11-01  19:35:10  28.96   99.8  100.0  100.0        0   94     0   0.9542   56.0   PL2
2015-11-01  19:35:12  28.98   99.8  100.0  100.0        0   95     0   0.9542   56.0   PL2
2015-11-01  19:35:13  27.13   99.8  100.0  100.0        0   88     0   0.8870   48.8   PL1
2015-11-01  19:35:14  26.71   99.5  100.0  100.0        0   88     0   0.9110   47.2   PL1
2015-11-01  19:35:15  26.65   99.4  100.0  100.0        0   88     0   0.9110   46.6   PL1
2015-11-01  19:35:16  26.66   99.4  100.0  100.0        0   88     0   0.9110   46.9   PL1
2015-11-01  19:35:17  26.65   99.4  100.0  100.0        0   87     0   0.9110   46.8   PL1
2015-11-01  19:35:18  26.66   99.4  100.0  100.0        0   87     0   0.9110   46.9   PL1
2015-11-01  19:35:19  26.66   99.4  100.0  100.0        0   88     0   0.9110   47.1   PL1
2015-11-01  19:35:19  26.66   99.4  100.0  100.0        0   88     0   0.8870   47.1   PL1
2015-11-01  19:35:20  26.65   99.4  100.0  100.0        0   88     0   0.8870   46.9   PL1
2015-11-01  19:35:21  26.65   99.4  100.0  100.0        0   89     0   0.9110   46.9   PL1
2015-11-01  19:35:23  26.63   99.4  100.0  100.0        0   89     0   0.9110   46.8   PL1
2015-11-01  19:35:24  26.63   99.4  100.0  100.0        0   88     0   0.9110   46.9   PL1
2015-11-01  19:35:25  26.63   99.4  100.0  100.0        0   89     0   0.9146   46.6   PL1
2015-11-01  19:35:26  26.65   99.4  100.0  100.0        0   89     0   0.8901   46.9   PL1
2015-11-01  19:35:27  26.66   99.4  100.0  100.0        0   88     0   0.9110   46.9   PL1
2015-11-01  19:35:28  26.64   99.4  100.0  100.0        0   89     0   0.9146   46.8   PL1
2015-11-01  19:35:29  26.64   99.4  100.0  100.0        0   89     0   0.9110   46.9   PL1
2015-11-01  19:35:29  26.70   99.5  100.0  100.0        0   90     0   0.8870   46.7   PL1
2015-11-01  19:35:31  26.64   99.4  100.0  100.0        0   89     0   0.9110   47.1   PL1
2015-11-01  19:35:31  26.64   99.4  100.0  100.0        0   90     0   0.9110   46.9   PL1
2015-11-01  19:35:33  26.64   99.4  100.0  100.0        0   90     0   0.8870   46.9   PL1
2015-11-01  19:35:34  26.66   99.4  100.0  100.0        0   90     0   0.9110   46.4   PL1
2015-11-01  19:35:35  26.66   99.4  100.0  100.0        0   91     0   0.8870   47.4   PL1
2015-11-01  19:35:36  26.68   99.5  100.0  100.0        0   90     0   0.9110   46.8   PL1
2015-11-01  19:35:37  26.63   99.4  100.0  100.0        0   89     0   0.9110   46.9   PL1
2015-11-01  19:35:38  26.63   99.4  100.0  100.0        0   89     0   0.9110   46.9   PL1
2015-11-01  19:35:39  26.63   99.4  100.0  100.0        0   89     0   0.9110   46.9   PL1
2015-11-01  19:35:39  26.64   99.4  100.0  100.0        0   90     0   0.9110   46.6   PL1
2015-11-01  19:35:41  26.62   99.4  100.0  100.0        0   89     0   0.9110   46.8   PL1
2015-11-01  19:35:41  26.62   99.4  100.0  100.0        0   82     0   0.8870   46.9   PL1
2015-11-01  19:35:42  30.42   11.0  100.0  100.0        0   64     0   1.0408   12.7
2015-11-01  19:35:43  35.84    2.5  100.0  100.0        0   60     0   1.0408    4.6
2015-11-01  19:35:44  35.88    1.4  100.0  100.0        0   59     0   1.0414    2.6

19:34:23 - As soon as Prime95 starts, the multiplier drops, the peak core temperature increases 35°C and the CPU is already reporting that it has started throttling due to it hitting the PL2 short term turbo power limit. Power consumption has just started going up but the CPU is already crying for mercy.

19:35:13 - After 30 seconds at the 56 Watt limit, throttling now switches to the lower PL1 long term limit. Multiplier drops, VID voltage drops, core temperature drops and power consumption is heading down towards the PL1 - 47 Watt TDP limit.

19:35:42 - The Prime95 test ends. The core temperature immediately drops and the log file already shows that throttling has stopped.

This new Limit Reasons logging feature earns its first big

 

You guys probably already knew or expected this, but undervolting doesn't work on the i5-6300U. I tried to use the adaptive voltage offset in 8.00 beta 5.

Intel XTU's undervolting does work, but it's incompatible with InstantGo

 

Thanks, @unclewebb. The logging changes should definitely come in handy!

 

The 6th Gen CPUs got rid of the integrated voltage regulator (FIVR) so the under volting that ThrottleStop can do with the 4th and 5th Gen CPUs no longer works for the newer ones. That is disappointing but nothing I can do about it. I will try to block that section of ThrottleStop off in the next beta since it no longer works. I think it is already blocked off for the 6th Gen desktop CPUs so I will have to add the 6th Gen U CPUs to the list. Thanks for the feedback.

I have to admit that I wasn't sure at first if Limit Reason logging was necessary but after testing the latest version, I like it. This new feature helps show exactly what is going on when one of these CPUs starts hitting the wall. Thanks to @async for this suggestion.

Edit - I just heard from error-id10t on Overclock.net. He was been helping me with improving support for the new Skylake processors.

He helped me locate a minor bug with the monitoring timers not starting so on some computers, the previous betas were reporting 0.0 for the multi. Everything looks OK now for Skylake starting with beta 5.

 

Really? I've got an undervolt running on my laptop i7-6500U and I can definitely see a performance difference with FIVR adaptive voltage on.
I get around 200mhz more CPU clocks and 100mhz more GPU clocks when running a Prime95 and GPU burn-in test. Obviously, the system also runs turboboost mode far longer than otherwise as well.

In fact, I know it works on my laptop because changing the values to extremes instantly crashes my computer. Maybe it differs depending on the implementation by the manufacturer.

Images of VID with offsets on and off

I am a complete noob though - you guys may be talking about something else, in which case ignore me.

 

Hi,

i know ThrottleStop is primary for mobile CPUs.
I am using it a while now for my desktop CPUs and just switched to Skylake.
It would be very nice if you could get it work also for this kind of CPUs.

I appendend you several screenshots and hope it is helping you.

 

No, that's exactly what we're talking about. Weird! Have you changed anything else in Throttlestop besides the adaptive voltage?

 

Hi unclewebb, i need your help, i have here a laptop with a celeron 1007u and i need to increase the tdp limit but i can't with your software, theres any possible way that you can unlock that feature for my cpu in your software in the future?

I il post some screenshots





My pc is an asus x551cap, and whenever i play a game it starts throttle because of the tdp limit, can you help me? theres any way to increase the tdp without a bios mod?

Have a nice day!

 

I think I got the adaptive voltage working on my i5-6300u after uninstalling XTU!

 

Thanks for your work. @unclewebb. I have a question and maybe can you help. We all know that the new line AW 2015It has different limitations of power to the CPU. In my case, an AW 17R2, have serious problems with the Power limit throttling.
In these cuts screen you can see:
1. Stock


2. With undervolt of -70mV cpu core / 50mV cpu cache:



The configuration on ThrottleStop:






The question I have is whether the problem of Power limit throttling might be solved? the limitations are in the bios and I suspect that little can be done.
Thank you very much for any advice.

 

On the Skylake based desktop CPUs, Intel removed the FIVR so I assumed that FIVR voltage control was not going to work on anything Skylake related.

I found an interesting picture on Laptop Media.

http://i.imgur.com/gGi8OQr.png

The CPUID for the Skylake U CPU on the left starts with a 4. The 6700HQ on the right starts with a 5 so that is usually a sign that these cores are coming from different stock.

6500U - 0x406E3
6700HQ - 0x506E3

And over here is a Broadwell desktop CPU.

http://www.guru3d.com/articles_pages/core_i7_5775c_processor_review_desktop_broadwell,7.html

5775C - 0x40671

Long story. Broadwell used FIVR, the first 3 CPUID digits of the Skylake 6500U is the same as Broadwell so I guess I shouldn't be surprised that the Skylake U uses FIVR. Skylake U might be sort of a hybrid chip with some Skylake features as well as some features from the previous Broadwell Generation. Can you post a screenshot of the ThrottleStop - FIVR window for your 6500U? I do not own any recent hardware so all of these screenshots are the only way I can learn about Skylake. Can you also download the latest TS8 beta from my signature. I made a few fixes to the C0% reporting and I disabled the Chipset Clock Modulation which doesn't seem to work on Skylake. I might need to tweak this some more to correctly report Skylake U.

@00riddler - As mentioned, the 6600K and 6700K desktop Skylake CPUs do not use the integrated voltage regulator (FIVR) so I do not think that voltage adjustment is going to work on those CPUs. That is why FIVR is blocked off. Can you be specific about what other things work or do not work in ThrottleStop?

@TheEvolution - Do you see in your screenshot where it says Package Power Limits and then the box that says Lock has a check mark? This means the register that contains the power limit information is locked by the bios. Once a register has been locked, it cannot be unlocked by any software. The only way to unlock it is to use a bios that does not lock this register.

That's a win - win for your laptop. Those U CPUs have to work hard enough as it is without having to run XTU too.

Your 3rd screenshot that includes Limit Reasons shows that PL1 is the reason your CPU is throttling. The 4980HQ has a TDP rating of 47 Watts so long term it is designed to throttle so that it remains at a maximum of 47 Watts. That is exactly what your CPU is doing. It is throttling for a reason. I do not know of any way to get a BGA HQ series CPU to go beyond this limit. Short term maybe but not long term. Gimped to 47 Watts is how it is.

Finishing the Power Limit 3 section of ThrottleStop is on the things to do list but I do not think it is going to help with your problem. Can you run the Dump tool that Dufus wrote and then copy and paste the data in the Dump.txt file that it creates to www.pastebin.com and then post a link here. There might be something in there.

@00riddler - I noticed in your screenshot above that your BCLK was at 105.98 MHz. Did you boot up at this frequency or did you use software while you were in Windows to overclock to that speed? The reason I ask is because Windows used to have a bug where changing the BCLK while in Windows would screw up one of the system timers. If you are overclocking in Windows, can you try running my testing tool, WinTimerTester. Let it run for about 30 seconds and then stop it and post a screenshot. It compares two different system timers and reports if they are running in sync or not.

WinTimerTester 1.1
https://www.sendspace.com/file/xadvhe

 

Here's a screenshot of the FIVR window for my i5-6300U. http://i.imgur.com/WR092P7.png

 

Your picture shows that Digital I/O reporting and adjustment is not supported by your Core i5-6300U but everything else looks like it should work. Did XTU have any voltage options available? I think XTU blocks or prevents adjustment of some voltages on some CPU models.

Now that I know some more about the Skylake U, I can make a few changes to ThrottleStop for improved support. Thanks for sharing.

 

Here's what options I had in XTU for the i5-6300U:
Core
Cache
Graphics
And the setting in the Other section is System Agent IccMax

 

Progress!

To address a few specific things first:
1. I was using the latest TS8.0b5. Chipset clock modulation is available in TS but changing it does nothing (as expected).
2. I have exactly the same screenshots as darkcond0, so I won't upload mine.
3. Congrats darkcond0, you seem to have won the silicon lottery. A 120mV underclock is madness - mine instantly crashes on that setting under 0 load.

Now, here is precisely what I did:
1. Completely disabled throttlestop on boot.
2. Download a fresh copy of throttlestop into a new folder.
3. Complete shutdown and power off of laptop.
4. Boot laptop.
5. Start HWInfo for logging.
6. Start Prime95 v 28.7 Small FFT Torture Test.
7. While running... try to set Core Offset to -75mV. This seemingly did nothing while in any states (PL2, PL1, etc) and I was confused (thought to myself maybe this was a fluke previously).
8. I remembered that in Intel XTU, my core voltage was linked to my cache voltage. I set the cache undervolt also to -75mV. Instant performance spike of around 300Mhz. I play with this by switching it on and off again and also resetting the cache offset to 0mV. It is clear that somehow core voltage and cache voltage are linked and the undervolt only works when both are set.
9. Run 2. This time switching core and cache undervolt on during PL2 and PL1 state. Core won't enter PL2 state (woops). I probably didn't allow enough time for the system to cool down. In PL1 state switching on undervolt results in a 300Mhz improvement in performance but within the 15W TDP window.

Here is a graph of the core Mhz in the first run - clearly showing undervolting working.


Here is a link to my log files: https://www.sendspace.com/filegroup/OSfF6T5q5ZcnUrqhpO/a7g

In my experience, I can say a couple of other things:
1. Turbo Power Level controls (in both Intel XTU and TS) both seem to do absolutely nothing - both for PL1 and PL2 states (with and without undervolt enabled).
2. Clock Modulation Works - although why you would want to stop your CPU from working at full power is beyond me.
3. Chipset Clock Modulation doesn't work - this is expected.
4. Set Multiplier does not seem to do anything.
5. I'm not entirely sure why my CPU powers down to a 9W TDP on extended load. Mind you, I never run it through torture loads like Prime95 so this is not a practical issue. I can audibly hear the fans spin down when it hits this power state though. I think it might be a BIOS level block from HP designed to limit chassis heat. Running at full load for 5 minutes does make the device uncomfortably hot.

Future research:
1. Does undervolting the graphics do anything? This is enabled in Intel XTU so presumably yes. I also experience graphics corruption with too high undervolting settings - however, I can't confirm whether there is any performance increase.
2. Does undervolting system agent and Analog I/O do anything? These options are not enabled in Intel XTU. HWInfo doesn't seem to track these voltages so I can't confirm.

 

http://ark.intel.com/products/88194/Intel-Core-i7-6500U-Processor-4M-Cache-up-to-3_10-GHz

The 6500U has a configurable TDP. Intel says the low TDP value is 7.5 Watts so your CPU dropping down to 9 Watts is likely because it is dropping down to the low power TDP Level. Have a look in the ThrottleStop TPL window to see the specs for your CPU. They should be right.

IccMax is the latest way to cripple an Intel processor. You can see that listed in your XTU screenshots for various parts of the CPU. I have a general idea of where some of this stuff is hiding but I am not yet sure if it will be possible to change these values. Intel CPUs and chipsets use multiple methods for throttling. It is beyond ThrottleStop to find everything that could be causing throttling, hope it has been left unlocked by the bios and then change it. Intel is winning this game.

What is the Non Turbo Ratio set to? On a 3rd Gen U CPU, setting this to zero allowed Set Multiplier to control the CPU speed.

Undervolting the System Agent and Analog I/O does not reduce power consumption any meaningful amount so will probably not improve performance. Some voltage adjustments are more useful on the desktop processors. You might be able to bump up the VCCIN and get a couple of extra watts of head room and maybe a few more MHz using that. Playing with VCCIN might also help you avoid the low TDP mode.

Not sure why core and cache voltage seem to be linked. I am sure someone is going to find that info useful when they are trying to max out the latest generation of U CPUs.

Can you try running ThrottleStop for logging purposes while running Prime95? I am interested to see what shows up in the ThrottleStop log file for Limit Reasons when the CPU drops down to 9 Watts.

Can you also run the Dump tool by Dufus so I can see what is hiding in your CPU's registers. I need to find a laptop that uses the PL3 register for throttling purposes so I have some motivation to complete this new feature.

Dump
https://www.sendspace.com/file/iumw0k

 

Hi, I done a stress test and Dump is:
http://pastebin.com/m7L8YiCQ

Thank you very much for your help

 

I've got a Skylake 6700HQ on my Alienware 15r2 if you need anything.... I might need some guidance though as I've never really messed with changing CPU parameters.

 

Voltage control would have been nice.
Button BCLK is not working.

Multiplier setting is working fine.

So far the first tests.

Overclocking is done in BIOS.

 

Your Dump file shows that the PL3 register is not being used to limit your CPU. As Limit Reasons shows, it is the 47 Watt PL1 limit that is forcing your CPU to throttle when fully loaded.

Here's the problem. There are multiple ways for Intel or a laptop manufacturer to set a 47 Watt TDP limit. ThrottleStop and Intel XTU give you access to the power limit register in the CPU. You can adjust PL1 in this register beyond 47 Watts but this value is going to be ignored if another duplicate PL1 register is still set to 47 Watts. The U CPUs work the same way. ThrottleStop does not have access to these other registers. I am not sure if Intel is doing this with the HQ processors but the 47 Watt number can also be hard coded into the CPU at the factory which would make it impossible for any software to get the CPU to go beyond 47 Watts. This brick wall is one of the reasons why enthusiasts are not very happy with HQ processors.

I noticed in the Alienware forum where you were comparing to another user who had a 4710HQ. Based on MHz, a 4980HQ should be faster than a 4710HQ but both processors have the exact same 47 Watt TDP rating. This means that when all 4 cores are fully loaded, both CPUs are going to run at almost the exact same speed because both processors are being limited to the same 47 Watt power consumption level.

When lightly loaded, the 4980HQ will be faster. What happens during Cinebench is that benchmark requires right around 47 Watts. A 4710HQ when not overclocked should be able to complete that test with little to no power limit throttling being needed. Your 4980HQ is trying to run faster but that would require it to go over the 47 Watt limit. This forces the 4980HQ to use a lot more power limit throttling compared to the 4710HQ to keep it under the 47 Watt limit. That is why your XTU graph looks like a sea of red. Spending extra money for the 4980HQ provides no advantage when fully loaded compared to the cheaper 4710HQ because both have the same 47 Watt rating.

@00riddler - Thanks for the info. It looks like the way that ThrottleStop calculates the BCLK MHz that worked well during previous generations is broken for Skylake when overclocking in the bios. It might be a while before this gets fixed. Good to hear that at least the multiplier works.

@epic_ninja420 - If you run CPU-Z and look in the Ext. Model box, I think you will find that value starts with a 5. That means there is no FIVR so ThrottleStop cannot be used to adjust the voltages anymore. You can post some screenshots of ThrottleStop so I can see if it is recognizing your processor correctly but there might not be much ThrottleStop can do to improve the performance of your 6700HQ. Thanks for offering to help.

 

http://www.intel.com/content/www/us...ktop-6th-gen-core-family-datasheet-vol-1.html
http://www.intel.com/content/www/us...ktop-6th-gen-core-family-datasheet-vol-2.html
maybe some relevant info regarding the throttling for skylake.

 

I appreciate the real explanation, very sad but true. Suspected that the problem was the Bios and its limitations ... but it's really all restrictions on Intel - Dell. Unfortunately I lost with the replacement of my old AW 17R1and now I have this garbage. But all is not lost I have researched a little and despite having few options. I made some adjustments to TS and the results confirm me that the CPU can run stable at 3.60 GHz (with no throttling). And with fluctuations in 3.70 Ghz. Therefore, I have to settle for that.

I know that every cpu is different but maybe these settings can help others. The first is to uninstall "Intel Platform and Dynamic Thermal Framework" with this application you will never use the 57W (seconds or minutes). The benefit is that you have to throttle the cpu It is coolest.
An image is worth a thousand words:

1. Stock configuration, throttle the power limit fully :-( but core frecuency is running to 3.40-3.46


2. With TS and multiplier block to 38 (UV cpu core -70 / -60 cpu cache, power long 57, power short 68, turbo time max and current limit 256A).....throttle the power limit fully. With difficulty hit the 3.60 GHz.


3. With TS and multiplier block to 37. Exists throttle in the power limit but continuously fluctuates and the rate is stable at 3.63 Ghz.


4. With TS and multiplier block to 36. There is nothing of throttle in the power limit and and speed is stable 3.60 Ghz.


The configuration that I use for this test is:





In conclusion, may be with a undervolt more (very little because my system fails if I passage of -76mV) could run stable at 3.70 Ghz. However, it is amazing that consumers are deceived with lies, in reality my CPU can not run under load to more than 3.45 Ghz with stock values almost like a 4710HQ. It is a shame.

@unclewebb thanks for your time, explanations and this great application.

 

It shows the two power levels as 7W and 25W. Neither of these seems correct as I've never hit 25W and my CPU has never been throttled to 7W (yet). Screenshot below:

Yes, I can see those too. Changing them in Intel XTU does absolutely nothing.

Sadly this still does nothing for me. I suspect a BIOS level lock on changing this.

Noted. I've disabled the undervolting on that. I'm not sure how to even begin setting VCCIN because I don't have a baseline measurement for what it should be.

Again this might be a bios level thing for me. I can definitely confirm that changing either of these independently does nothing for me.

Dump and logs here: https://www.sendspace.com/filegroup/yIynIZVz6VisJV5LHIhx+om2H6XkAUfL

Limit Reasons has never seemed to work for me. When I activate it I get a yellow solid EDP light but nothing else seems to happen. If I reset by clicking the titles, then after a few seconds I get a yellow solid EDP light but nothing else.

 

Yellow light doesn't actually mean anything. I don't know why some things light up and other things don't, but I've always had yellow PL1, PL2 and EDP lights on my limit reasons, and when they flash or stay red is when that part is limiting things.

I can't talk about Limit Reasons on Skylake CPUs though, unfortunately. But they work for Haswell, I am sure of that.

 

Thanks for all the data. I appreciate it.

@illuzn - A VCCIN of 1.80 V seems pretty safe. Setting VCCIN higher tricks the 4th Gen CPUs so they report less power consumption which allows the CPU to use more Turbo Boost and clock higher.

The TDP Level is another one of those items that is stored in multiple locations. The CPU supports a higher 25W TDP Level but a laptop manufacturer can decide to block that. Have you tried setting this to Level 2 in ThrottleStop? If those other locations are set to Level 2 then jackpot. If any of those other locations are set to Level 0 then using ThrottleStop to change this register in the CPU will not get you anything. When your CPU really drops down during a stress test, it might be going into TDP Level 1 and there is probably no way to prevent this. On previous U CPUS, I found that the reported power consumption is not always exactly the same as the TDP Level rated power consumption.

I just read the updated Intel docs. The register that contains the Limit Reasons data for 4th Gen Haswell and 5th Gen Broadwell no longer exists in 6th Gen Skylake. MSR 0x690 has been reassigned to something completely different. I will have to disable the Limit Reasons button in ThrottleStop when running on a Skylake.

For each limit reason, the register contains one bit which indicates if throttling is in progress and a second bit that contains a log of any throttling that has happened since you booted up. Clicking on the headings in Limit Reasons lets you clear the log bits that are saved in this register.

If a log of throttling is recorded by the CPU, this is the data that Limit Reasons reports as a Yellow box. Throttling in progress will show up as a Red box. Things happen pretty fast inside a CPU. If it throttles for only a millisecond, you may never see the box turn red. That's why Intel included logging for each limit reason. The slightest amount of throttling will be logged within the CPU whether Limit Reasons is running or not and it will be reported as Yellow boxes the moment you start Limit Reasons.

In a perfect world or perhaps on a desktop CPU where all of the power and current limits are set appropriately high enough in the bios, you should never see any of these boxes turn red or yellow.

 

That makes sense... I always wondered why my Limit Reasons started with yellow EDP, PL1 and PL2 from the first time I ever launched it, even though I've never seen it hit red on PL1 or 2 at all. If it's a permanent record, that would make sense.

 

Thanks for your support and for making this awesome tool.

Setting TDP Level does nothing on my laptop. Presumably this is blocked somewhere else.

Setting VCCIN has no effect for me. TS still thinks it is sett to default according to the indicators in the side panel.

Looks like the end of the road for me and my laptop (unless icc Max can somehow be set - it doesn't work in Intel XTU). That said, the undervolt definitely nets me more time in turboboost - which does have real life use impacts. Other than benchmarking I don't think I ever hit the throttling window (I don't game with the device). Conceivably, if I was gaming I could see how this could be a more significant issue.

Actually, one thing I noticed while doing a graphics benchmark, the PL1 (UnCore), PL2 (UnCore) and EDP (Ring) indicators do work under a iGPU stress load. So Limit Reasons is not completely useless.

I have also found that PL4 icc Max supposedly triggers according to HWInfo under a graphics load. So it seems that unlocking the iccMax might be of some benefit for gamers - I haven't tried this in Intel XTU under these conditions.

 

Probably because your laptop has a HQ CPU. If you have PL4 however that probably means skylake, which automagically means HQ, so yeah =D.

 

I am running Skylake but it's Skylake-U (more specifically i7-6500U).

Besides some smaller niche manufacturers, the only manufacturer I know that uses the HQ processor is Apple (in their MacBook Pro) and the Skylake model for that hasn't been announced yet.

 

Ah. Well yeah, your CPU is a ULV chip. It's 15W limited. If you're lucky, some boards allow 20W, but that's generally the highest it'll go.

Dell, Alienware, Clevo, MSI, ASUS, Razer, Gigabyte, HP, Toshiba, Lenovo all use HQ chips. I don't know how you haven't seen them, unless you're been looking at very low powered machines.

 

You learn something new every day... I always thought the HQ chips were used in "gamer" laptops like Alienware, Clevo, MSI, ASUS and Razer. Didn't think the more mainstream brands used them too - especially given the 45W TDP window - that would absolutely obliterate battery life I imagine.

I guess I was always looking for an ultrabook class device (hence why I never even bothered to look at these higher power devices).

 

If you want to know more, the mobile CPU guide in my signature is a good read.

 

Okay, this was a huge user error from me, never mind

 

Do you think a PL3 register could hide in a HP broadwell i5-5200u? Just say and i'll run the tests you require

A VCCIN of 1.80 V instantly hangs my computer and then i have to hardreset it for 20sec, HP x360 Spectre, Broadwell i5-5200U. Do you have any other recommendation for us Broadwells?

Thanks a lot for a very good program!

 

My 4th Gen 4700MQ has a PL3 register so it is likely that a 5th Gen Broadwell has this register and maybe Skylake too. I do not have documentation for this one so it is all trial and error. My 4th Gen has it but by default it is set to zero so it is not being used. Intel now offers manufacturers multiple methods to throttle their CPUs. There are so many methods available that manufacturers do not need to use all of them.
If you want to know if your laptop uses PL3 then you can run Dump by Dufus. Unzip it, run Dump and then Copy and Paste the Dump.txt file that it creates to www.pastebin.com and post a link here.

https://www.sendspace.com/file/iumw0k

For VCCIN you could try 1.90 V or 2.00 V. Hard crashes always scare me so you might want to think twice before doing this. I got stuck in a BSOD boot loop because I had ThrottleStop booting up with Windows and I was using the OK - Save voltages immediately option. I created a useful safety feature and then forgot to use it.

In the name of science, I have maxed VCCIN out to 2.30 V. The world and my CPU did not come to an end. For most users, this setting is only going to be useful if you are trying to get the last little bit out of your CPU.

You are welcome for ThrottleStop and thanks for your offer to help.

 

Hello unclewebb,

Here is a link to the "dump" of an i7-5600U (broadwell) from my Lenovo T450s laptop, if it helps you with the PL3 etc. stuff... I don't know, if this laptop uses PL3.

http://pastebin.com/7jxzRN6v

If you need something else, please let me know! Thanks for this great program. It helps me run my laptop both cooler (saving energy) and faster, if I wish

Best regards,

Jukka

 

Hey guys, been inactive for a while cause well repasted my cpu since it was dry and than stripped my cpu intel screw, than stripped my gpu screw... all fixed up now.

been attacking 4.0ghz. the last/max out my processor. been going really good so far. it's freaking fast.
but sometimes i get bs0d 124. did some reading and it says voltage is too low? i had it undervolted, but at -20 vs -9 for example, i could play longer at the -20 at times.. i can manage to play bf4 for an hour and temps are under 78 degrees celcius.
as always im open to tips haha. this is addicting lol. just need a point in the right direction pretty much.

 

It sounds like you are right on the edge of stability. When living on the edge, random BSODs can happen so this makes it difficult to troubleshoot.

If your temps are fine and they are, then give your CPU core and cache some more voltage. I would go with zero offset to start with. Who knows, with that multiplier, maybe you will need some positive offset for 100% stability. When trying to find out what multi your CPU is capable of, it would be a good idea to back off the cache multiplier while testing. 4 or 5 less than the CPU multiplier is usually good so about 35 or 36 for the cache. When trying to get the last little bit out of your CPU, it is easier if you know what each individual component is capable of.

You have Clock Modulation and Chipset Clock Modulation checked. Does your laptop use either of these throttling methods? If these are not used by your laptop then there is no reason to check those boxes. Run a Log File with neither of these checked. If the entire log file shows 100.0 for both of these then there is no reason to check either of them. Why waste CPU cycles fighting against something that is not happening?

Another trick I found is that setting the Non Turbo Ratio to 1 allows my CPU to run at full speed without needing to use the Set Multiplier option anymore. This allows me to leave ThrottleStop in Monitoring mode without any multiplier drops. The Non Turbo Ratio setting takes care of the throttling scheme that Lenovo uses on many of their laptops.

Keep up the good work. It is always fun to see someone go OCD while trying to max out their CPU, using ThrottleStop of course!

Thanks for the data Jukka. Your laptop is not using the PL3 register. I might get rid of that new feature unless I can find a laptop or tablet that is actually using PL3 throttling.

 

Yup give more voltage... I used to have that issue with the 3940XM, lots of BSOD's due to lack of voltage at 4.4GHz... Gave more and problem went away...

 

Hey, so i took your advice and ran the log and yea i don't need it checked. I also changed the cache to 37. so far no bsod and even got a better result in the benchmark! I kept voltage at 0, better not raise it unless i really need to!

Only thing is for uncore clock, when i'm playing bf4 is it running at the max of 3.7ghz of cause of the uncore clock or 4.0ghz?
so basically i could get better results of having a lower cache than my multiplier is what you guys are trying to say?
is my turbo time limit okay?
so basically since my multiplier sliders are maxed out to 40, by adjusting cache and voltage should get me where i want.


Here again is what i did, and in red is what has been tweaked or stayed the same. I appreciate the help you guys are providing me to get to my max clock!

 

Most newer laptops throttle based on power consumption. Clock modulation throttling is not used much on recent laptops and that is a good thing because clock modulation throttling was horrible.

Your Cache Ratio is set to a multiplier of 37 so it is running at 3691 MHz (37 X 99.77 MHz). The other monitoring program you are using calls this the Uncore Ratio but it refers to the same thing. The reason most people with highly overclocked desktop processors reduce the cache ratio is to improve stability. Many 4th Gen Haswell desktop CPUs top out at about 4500 MHz to 4700 MHz so they are using a CPU multiplier of 45 to 47. This is usually only stable if the cache ratio is in the low 40 range. It is very difficult to get one of these CPUs maxed out with the CPU and cache multipliers set equally. If dropping your cache ratio improves stability, that might allow you to increase your CPU multiplier up to 41 when 1 or 2 cores are active for a little bit more performance when the CPU is lightly loaded. A faster CPU core is more important than a faster CPU cache.

Did you try using a Non Turbo Ratio of 1 and turning off the Set Multiplier feature? If the log file shows the multiplier steady at 40.00 then maybe this might be the setup for you.

Edit - Just had a second look at your screenshot. It looks like you did play with this because it shows that your CPU Non Turbo Ratio is Locked to 1. You will have to reboot to unlock this.

As long as your CPU is running at full speed in the log then your power limits are set appropriately.

The C State Demotion stuff does not seem to make a big difference one way or the other.

 

I couldn't get it stable. i played with it for hours. Only time i was able to play bf4 was max 1.5 hour without a bsod.
is there a huge difference of 39 vs 40?
also i changed it back to 39 multiplyer and max/min cache 39x39 and cache ratio to 39 and runs smooth.
im aiming to get my bf4 fps decently constant. right now it fluctuates when i scope in with a snuper like 175fps and unscope like 99fps lol. or like when im playing 80-90-100 and when i die 145 lol. is that normal? or just means i need to fine tune the oc?
also is my ts benchmark for the seconds okay or can i really get down to low 7/high 6 seconds. i'm assuming no since i have a laptop lol

i tried using the non turbo ratio and didnt really notice a difference tbh

 

Just use vsync... Also what details are you running at? Increase them a bit.. Unless you have a 120Hz display, more than 60fps is useless...

 

Hi, only 2 questions...
1) why dont i have FIVR button? (i5-430m)
2) Why cant i minimize throttlestop to tray? It only minimizes to task bar and there is no option "task bar" to uncheck like in 6 version.

 

Your CPU predates the integrated voltage regulator.


Sent from my iPhone using Tapatalk

 

Only Intel's 4th and 5th Generation CPUs have an integrated voltage regulator. Your CPU is a 1st Generation Core i so IVR does not exist.

The Task Bar option is on the main ThrottleStop 8.00 screen on the left side towards the bottom. No one could find this option in ThrottleStop 6.00 so I tried to made it easier to find.

If you are still having problems, read the included ReadMe file that comes with the ThrottleStop 8.00 download.

 

Good afternoon! The program is really wonderful, but there is a problem on a laptop Asus K56CM.
When laptop wakes up from sleep, feature "Power Saver" doesn't work, FID resets to default value and doesn't reduce even manually. Problem can be solved by rebooting the system or using of hibernation, but appears again after sleep. I used Windows 7 (x64), ThrottleStop 6 and ThrottleStop 8 beta4, PowerSaver C0%=35 and selected High peformance Profile in Power Options.
Screenshots before/after sleep are below. Sorry if i used an inadmissible picture hosting, I don't understand how to upload pictures here.
Before sleep:

After sleep: Nothing happens after I change multiplier's value.