The ThrottleStop Guide

Hi unclewebb,

A new discovery from me... in Intel XTU, I have both a "Processor Graphics Voltage Offset" and "Processor Graphics Media Voltage Offset".

I never really understood why undervolting the GPU in throttlestop would result in artifacts etc. appearing over my screen (mostly in states which didn't use a lot of GPU power). In Intel XTU these 2 settings are linked but changing the setting in throttlestop only changes the first without changing the second.

I've been using it with both undervolt settings set to the same and I can now achieve a similar undervolt to my CPU i.e. around -80mV. Without them linked, I could only undervolt around half of that amount without causing stability issues.

I can't explain what this setting is or what it does. All that I know is that in Intel XTU there is a separate clock speed read out for "Processor Graphics Media" which is seemingly in lockstep with "Processor Graphics Voltage".

Hope this might assist in some way.

 

I have found that when Set Multiplier does not work, this is usually because of the Non Turbo Ratio setting. In ThrottleStop 8.00, click on the TRL button and have a look for that setting. The default value for most CPUs should be zero (0). Some of the U CPUs used in Asus laptops seem to set this to something different. Sometimes it is set to the CPU's default multiplier or 1 less than the default multiplier. I think my daughter has the same laptop model but it has a Core i7-U processor instead. If she lets me use it, I will see what I can figure out. The bios might be changing the Non Turbo Ratio value after you resume from Sleep.

Does Power Saver save any power for you? I know it lowers the multiplier when it is working correctly but depending on what you are doing, a Core i CPU might not save very much power when using this feature. Core i CPUs save power by spending lots of time in one of the low power C States like C6 or C7. In this state, MHz and voltage are not important because the cores internal clock is zero and it is disconnected from the voltage rail and is getting zero volts. Intel uses Power Gating so large portions of the core go off line when not needed. Changing the reported CPU speed when in C6 or C7 does not save anything. The ThrottleStop - Power Saver feature was originally designed for the older first generation Core 2 Duo. A lot has changed since then. The power consumption that ThrottleStop reports and other software reports is a meaningless number when a CPU is lightly loaded. Do not use that number to judge how much power your CPU is using when it is lightly loaded or idle. Intel only designed their estimated power consumption data so they would have a way to control their Turbo Boost function. It is only a rough approximation based on VID voltage. A large change in VID voltage will change the reported power consumption data but if a core is in C6 or C7, the change in actual power consumption will be close to zero.

@illuzn - I am not sure where that other voltage is hiding. If I ever find it, I will add adjustment of that to ThrottleStop. Thanks for the info.

 

Thanks for your detailed answer!
The default of the Non Turbo Ratio setting is 16 for my CPU, that is 1 less than the default multiplier as you stated above.

I don't pursue energy saving, but this option allows me to reduce CPU temperature by 5-7 degrees when it is lightly loaded or idle. It is very pleasant, considering that the laptop lies on my knees in general. It seems that battery run time also increased.

 

@holyunholy - Thanks for the feedback. There goes my theory that the Power Saver feature is not very useful on a Core i CPU. I will look into this issue further and see what I can figure out. After you resume from Stand By, can you try setting the Non Turbo Ratio to zero to see if that allows the Set Multiplier feature to resume working for you.

 

Yes, I tried setting the Non Turbo Ratio to zero, but Set Multiplier does not work after sleep again.

 

I've seen drops of 5-7c as well when idle when using the power saver feature on my i7 4710 MQ.

 

Hello,

I had pretty high cpu temps, than i cleaned the laptop and reapplyed thermal paste. that did help but i wasnt happy enough.

So i started using Intel XTU to lower voltage, do decrese some temps. it somehow worked a bit, anyways if i just run AssassinsCreed i get around 65degress on cpu, but if i use the Throttlestop7 beta3, my temps go up for 10degress to around 75. it would be reasnoble if the TS boosted my cpu to max ghz and i would get those temps, but my cpu is performing at the same speed without TS, so i gain 0 performance but 10degress?. So my guestion is, Does TS7 counter the XTU voltage and sets it to normal and i gain 10degress or what? because its a bit weird to gain 10degress just by turning this program on. Anyone has any ideas?

 

What thermal paste did you get? If its crap like MX-4 it will pump out in no time and your temps will be bad... Also temps of 65 is good...

 

There is a single register in the CPU that contains all of the voltages. If you are using two different programs and they are both writing voltage information to the same register, the results are going to be unpredictable.

Why not start by trading in your cracked version of ThrottleStop 7 and upgrading to ThrottleStop 8. The new version has a monitoring table on the right hand side of the FIVR window and it will accurately show you the FIVR voltages. If you want better performance, completely exit XTU and learn how to use ThrottleStop. Put ThrottleStop fully in charge of your CPU and it will do a great job for you.

UPDATE

Speaking of FIVR, I was playing around with a Core i5-6400 Skylake yesterday and I discovered that Skylake continues to use the exact same FIVR register. I originally disabled the majority of the ThrottleStop - FIVR window for Skylake CPUs but it looks like there was no reason for me to do this. Adjustable voltage seems to be possible whether it is a 4th, 5th or 6th Gen Skylake mobile or desktop CPU. @Mr. Fox will be happy with this news.

 

Indeed... very glad to know that and look forward to helping with testing that with the 6700K for you.

 

I'm pretty sure lots of people will be happy with that news, that's great news.

 

i got the artic silver 5, you didnt fuly get my questions but its okay, yeah i know 65 is good, i had 90+ before haha

 

well the XTU has a simple voltage settings, i didnt find anything in TS that does that, ill upgrade to ts8 now to see how that one looks, but i willl need some guidence if someone can help out here. in XTU i have -80 on core,-60 on catche and -30 on graphic.


i found those stuff, when i downloaded ts8 the voltages were already set the same as in XTU. My question still proseeds, why do i get 10degress more when gaming and using this versus not using it?


I get the same performance with or without TS8... i just get 10 more degress if i run it.. so i guess i just wont use it..

 

Silver 5 is junk.. Get something like Gelid GC Extreme and IC Diamond and you'll see the difference..

Sent from my LG-H811 using Tapatalk

 

¸

well its the first time i hear abou gelid GC extreme so it cant be that good, i checked alot of sites and never saw that name..
well AS5 isnt junk if u know how to use it.. the only downsite is that it has a 200h sattle time.. i choose AS5 cuz it was the beast i can find in local stores. im not gona pay 100bucks for shipping for a 10buck item lol.

 

Actually, from the benchmarks I've seen AS5 compares within like 1-3C of Gelid GC Extreme and IC Diamond.

 

That´s a fallacious argument.

And there is also Grizzly Kryonaut to consider. But if you want proper cooling get CLU.

 

As5 is outdated. Maybe ok on desktop processors with great cooling but not on laptops. Maybe useful 6 - 8 years ago but not now. Also Noctua Nt-H1 is outdated if you didn't know that

 

Okay so it is outdated, my laptop temps are pretty good now, it hard to have low temps on laptop totally differant than deskop ones. anyways i have 32degress totally idle, 40 when browsing and waching stuff. max 70 when gaming (and i really mean gaming, 100%cpu 100%gpu and stuf cuz my laptop suxs haha) thats without an coolingpad cuz it in repairs. so yeah im happy under 70 so lets leave the thermal paste debate alone shall we?

i stll have no idea why i get 10 more degress when using TS8 vs not using it, +0 performance gain thats the thing we shauld be talking about

 

Have you tested if you gain higher scores in pure processor benchmark test like wprime 1024/Cinebench 11.5 ?

 

i did not do that true. but since the performance in BO3 and ACS is the same. i guess i just wont use it here

 

On the core m7, interestingly, TS allows the chip to sustain nearly 15W TDP. This is interesting. I wonder if that's normal? I have set it up like so (from defaults):

- Check Set Multiplier. Set the box to 31T (the max)
- In TPL box, Package Power Limits:
* Move Turbo Time Limit all the way to the right (3670016, orig = 28)
* Package Power Long = 15 (orig = 7)
* Package Power Short = 30 (orig = 15)

TS Bench gives me under 18.5 seconds for the 4-thread standard test.

However, I have noticed TS does not accurately report the processor speed/multipliers, and shows it constantly at 3.1GHz even though it locks at 2.9GHz with the 4 cores running.

[edit]
I spoke too soon. In the longer (1024M) test it throttles back down again to around 4W, which it does at some 1.6GHz. Any way around this? The temps don't even touch 70 while this is running...

 

This is probably a CPU package level setting. I would be highly surprised if Intel allowed its Core m7 CPUs to run at that turbo clock speed indefinitely.

For comparison, my Core i7-6500U running at 3.1Ghz generates a score of 17.797s (mighty close to your m7 score). If Intel allowed this to incur indefinitely, there would be little reason to opt for the i7 core. Yes, there are architecture differences which favour the i7 and a slight graphics performance improvement - particularly given that they are the same price.

It's the same reason why my i7-6500U is capable of 25W level performance but in the long term that isn't allowed.

Whether there is a physical reason for this beyond purely being in software - I can't say.

 

I don't get why people aren't manually setting the multipliers and voltage offsets... That way, you know EXACTLY what results you will get...

 

Good points. I've gotten TS to fix the throttling a little bit -- now it at least throttles near 2GHz rather than a measly 1.6GHz.

Also, interestingly, when on battery, the processor rigidly respects the 4.5W, never letting it get a hair above that level. But, when plugged in, the device is thermally throttled in an interesting way as described previously. What if it's throttling based on actual temperature in that case rather than TDP? The speed goes down very, very slowly over a 10 minute period.

How? Remember this is a Core M.

Also, I just found out changing adaptive voltage is broken in Skylake. Static works, and adaptive auto with offset works, but not adaptive specified. What a shame!

 

Oh, I didn't pick up on that. Sorry. That's a good point. With that said, I have seen people in the past who could set multipliers and voltage offsets and yet they messed with TDPs and guessed instead.

 

The 4.5W on battery may have to do with an electrical limit i.e. the battery isn't designed to allow that much current to be drawn (overdrawn current = battery explosion).

It may be throttling on temperature, there have been some reports of the Surface Pro 4 throttling based upon a temperature sensor on the skin of the device (to avoid you from burning yourself when picking up the device). A simple test is to fill a bag with ice water (make sure the outside is completely dry and leak proof) then place it on various places on your device - that may or may not work. You will need to move it around to find out where the temperature sensor is - if any.

Probably something to do with the fact that the Voltage Regulation is no longer fully integrated to the die but on a separate chip now (although addressed in the same way). This means that throttlestop and XTU still mostly work but it functions slightly differently.

 

Excellent idea, brilliant. Hm, even simpler to start out might be to put the laptop in a ziplock in the freezer for awhile (off initially of course) then take it out and see if it throttles any later. This is assuming it won't warm up super fast (which it might...hence the need for sustained physical contact with the bag of ice water?). Or just take it outdoors (it's just above freezing temp here), but then I worry about condensation bringing it back in. Oh this is exciting! Physical experimentation is fun.

Makes sense. Although interesting since XTU is made by Intel -- wouldn't they have fixed their own app's voltage access prior to release?

 

Results!

No difference with the freezer or cold weather test even though the little device remained quite chilly all over -- seems to be a TDP thing still, just a bigger limit and longer.

But good (?) news -- XTU can disable the TDP control completely. Scored me a 581 on the 1024M test, yeah! And I verified it was a temp throttle if anything at this point because the cores all hit 90 and the device got hot as blazes (I was afraid it would melt). To do this I raised both short and long term limits to 4095.8 (the max) on XTU and the turbo time to 100 (the max). Interestingly, this trick wouldn't work with the short-term limit disabled -- gotta keep it on but absurdly high.

This same trick doesn't work at all in throttlestop -- it lets you set a lower TDP, and also lets you change the TDP level from 0-2 just fine, but the turbo time and short/long-term power limit boxes don't seem to work right.

I wish there was something in between disabling TDP and being forced into 4.5W.

It also looks like XTU's voltage control is largely the same as TS's -- probably a little worse. I verified the voltages set in TS are respected by XTU and reported working, and XTU does not let you specify an Adaptive voltage target at all -- the moment you try, it forces you into Static. So TS is not glitchy here -- the adaptive option should just not offer the user the ability to enter in a voltage because it simply does nothing.

http://hwbot.org/submission/3054450_jordache_xtu_core_m3_6y30_496_marks

 

Hey been inactive for a while since my power brick is dying. I ordered a new one from sager and waiting it to get here.
It gets so hot it burnt me and also makes a huge buzzing noise under load and just sometimes disconnected even though connected to power.
even during my OC i notice that for benchmarks my time keeps fluctuating under idle.
once i get it ill be back on the horse.

on another note got some sick ram
g skill ripjaw 2x8gb ddr3 running at 2133mhZ.
is it worth it to overclock it to 2400mhZ? from timing 11-11-11-31 to 11-13-13-31 through the bios?
if it works out maybe it'll help my 4.0ghz instead of 3.9

 

starting to get worried it might be my dc jack on my motherboard. gonna need a volt meter to test out my psu

 

Honestly, overclocking RAM doesn't do a whole lot of anything.

 

I would stop using that PSU... It's about die and it might take out your notebook as it does... Wait for the new one..

 

Hi, Would anyone please explain what c7 interrupt response limit do?

My laptop has I7-4712hq with 37W TDP

c7 interrupt response limit was originally 18 i increased it to 20 and i notice big difference to Geekbench 3 score, specially the multi-core score

i notice the less power usage too and it doesn't TDP throttle while testing.

Here after increasing the c7 interrupt response limit

http://browser.primatelabs.com/geekbench3/4517875

it's normally 3200/11500

when i increase the limit to 24 i see too many zeros

 

done! got it from sager. it's smaller than the original one i got from mythlogic.

here is what i achieved with my new psu. so far so good. i wonder if it's possible to get the to flat 7 haha, but the heat is what kills me honestly.
feedback and suggestions accepted thanks for all the help, this is really fun.

 

The C7 IRL slider controls the amount of time the CPU will wait before it decides to enter one of the low power C States. This adjustment will not work for everyone but it seems to be working very well on your 4712HQ. Your screenshot that shows all 0.0 means that you have disabled all of the low power C States. This can reduce off idle latency. You might see improvements in hard drive benchmarks, especially if you have a SSD.

Actual power consumption is probably the same during your Geekbench tests but I think this adjustment might be tricking the CPU so it reports that it is using less power. If the CPU thinks it is consuming less power, it will use more Turbo Boost, run faster, and will give you a real increase in performance and your benchmark scores will improve.

The register in the CPU that controls the C7 IRL has a few more adjustments possible. I just added one simple slider so users could play around with this. Your results are very interesting. There is also a register that can be adjusted to control the C3 IRL and C6 IRL but I haven't added those to ThrottleStop yet.

This adjustment can also allow the processor to go into the low power C States sooner as well as spend more time in a low power C State. This could be a way to decrease power consumption when lightly loaded and improve battery run time.

Thanks for your positive feedback.

 

So descriptive. thank you for the awesome utility. waiting the next release!

 

Should be possible with a firmware mod if your hardware can handle it.

 

@Dufus - Always fun to see what a retail 4700MQ is really capable of. Very nice core temps, especially considering the MHz and 95 Watt power number that ThrottleStop is reporting.

 

I'm certain he did some impressive cooling feats like cooling the heatsinks with an A/C or something prior to turning it on and running the benchmark xD. Fully modded Clevos and Unlocked Alienwares can't handle haswell at that speed with such temps.

 

No special A/C, ambient 27C and a simple water block + rad with 1 slow fan. Not so different from it's DT counterpart. External VRM needs some work though but lost some enthusiasm since Skylake bclk has been unlocked.

Haven't played with Skylake myself but if the iGD problem can be gotten around might be interesting on some laptops where the power limiting clamp function is not used as turbo could be disabled while allowing for a much higher base frequency without power throttling.

 

WATERCOOLING.

I KNEW IT

 

So for some reason most of the time my E4310s i5 560m speed varies from ~1.2GHz-2GHz when I tell it to run 1.2GHz only. Sometimes it actually does run the 1.2GHz constantly, but seemingly without me doing anything. Also sometimes my package state limit gets reset to C0.

And recently I tried gaming on my big laptop (M4500) on battery. The battery overheated, wasn't even running full speed. Not a real problem, just alarming.

 

Hi, I don't know if this is a bug but TS does not minimize to the notification area on Windows 10. It stays in the task bar. Needless to say that I do not have "No Notification Area Icon" checked.

Also, what I do not quite understand is whether I have to keep TS running once it is configured and turned on. Will the throttling be disabled even after I close the program?

 

I do not have that issue on Windows 10. You do have to keep TS running if you want its features to continue working.

 

@cwazywazy - Some of the first generation Core i CPUs seem to ignore the Set Multiplier requests that ThrottleStop makes. I think the problem is that these processors handled the C States a little differently compared to newer Core i processors. They also have limited control of Turbo Boost. It is either full Turbo Boost or no Turbo Boost and nothing in between whereas most of the newer Core i processors give you full control of how much Turbo Boost you want. When testing, make sure there is a load on the CPU. Run something simple like a single thread of the TS Bench. While this test is running, try changing the Set Multiplier value and see what it shows. If you see anything that doesn't look right then post a picture of it so I can have a look. There is probably nothing that I am going to be able to do about it.

I think the package C State randomly changing might be a legitimate bug. I have had a hard time trying to figure this out because most of the computers I have owned have this register locked. It is difficult to properly debug features when you do not have the necessary hardware to test with. I will have a look at this issue again but it will probably be a while before I have the time. If anyone can come up with an exact procedure that causes this bug or any bug then let me know.

Some laptops can consume a lot of power and battery technology is no match for this kind of power draw. It seems like some manufacturers assume that when running on battery power, you should be able to check your email and that's about it.

@Artaios - Read the included ReadMe file. It mentions what to try if you are having any problems with the system tray icons. I use Windows 10 and I have not had any issues. There are 1001 different laptops and throttling issues. Some issues can be corrected with ThrottleStop and you can then exit ThrottleStop after that. Your computer will run at its full rated speed until you reboot or use Stand By mode. Other laptops need ThrottleStop to be running all the time to make sure that throttling does not resume. Tell me what laptop you have, what type of throttling problem you are trying to correct and maybe I can tell you some more info. You can also exit ThrottleStop and do your own testing. If throttling returns then I guess you need to leave it running. ThrottleStop is very efficient when minimized to the system tray.

 

Awesome, thanks a lot. I did not bother with the instructions regarding the notification area because it seemed to me it was all about the problem of removing the icon. My problem was that I could not make the icon appear. Anyway, that fixed it.

I have an old Dell Studio XPS 1640 from 2009 with a Core 2 Duo P8700 @ 2.53GHz and a Mobility Radeon HD 4670. It's the well-known problem with the heat and the power consumption, I think. I followed these instructions here and bought a 130W adapter.

 

I had to do some Google searching to jog my memory.

http://forum.notebookreview.com/threads/1640-throttling-post-here.458668/page-22

The XPS 1640 uses Chipset Clock Modulation throttling. ThrottleStop needs to be running all the time to prevent this. There is no need to disable SpeedStep (EIST) in the bios. ThrottleStop may not work properly when this is disabled. There is also no need to check off the regular type of Clock Modulation.

Check off Chipset Clock Modulation and set it to 100.0%, check off the Set Multiplier box and set that to the maximum, use an appropriate voltage, click on Turn On, minimize ThrottleStop to the system tray and you are done. It really is amazing how ThrottleStop plus a few appropriate mouse clicks can make a huge difference to these laptops. From unusable garbage to not that bad. The XPS 1640 and 1645 could have been great laptops but Dell really dropped the ball when they decided to save a few dollars and go with a 90 Watt adapter instead of something more appropriate. Damn bean counters.

Great to hear that the system tray icon is working. I think this is properly programmed now but when I originally wrote ThrottleStop, my icon code was not compatible with Windows 7 or later. Running ThrottleStop 6.00 will cause various icon issues but if you force Windows to rebuild the icon cache and stick with ThrottleStop 8.00 after that, everything should be OK.

The one time when disabling SpeedStep might be useful is when using Dual IDA mode.

http://forum.notebookreview.com/thr...ion-ida-on-both-cores-of-a-core-2-duo.477704/

This trick lets you lock both cores of your CPU to the maximum multiplier for a small increase in performance. Let me know if you need some help with this trick.

 

Hi, thanks so far. Could you repost those pictures in the instructions, please? The first time I set this up I got a WHEA Uncorrectable Error and the PC crashed (after about 30 seconds). Not sure if I should enable or disable Power Saver, C States, BD PROCHOT, SLFM and Turbo. Also, the maximum I seem to be able to set the multiplier to is only 10, default is 9.5. Is this normal? Thanks!

 

That thread is more than 5 and a half years old and the pictures are with a very old version of ThrottleStop. Unfortunately, the pictures in that forum are long gone and I do not have access to a Core 2 Duo laptop.

If you are trying to use Dual IDA mode, I would disable Power Saver, disable C States, disable BD PROCHOT, disable SLFM. Disable Turbo should NOT be checked.

9.5 is the default multiplier and 10.0 is the multiplier used during Intel Dynamic Acceleration. Intel renamed IDA mode and called it Turbo Boost when the Core i CPUs were released. The way IDA works is that when a single core is active, the CPU uses the 10.0 multiplier and when both cores are active, the multiplier for both cores decreases to the default 9.5 multiplier.

ThrottleStop lets you trick the processor so that instead of one core using IDA, you can have both cores using the 10.0 multiplier at the same time. Thanks to Dufus for uncovering this trick.

If you get a BSOD or a WHEA error, this usually means that you are trying to use a high multiplier but you are not using enough voltage. You have to adjust the voltage first before trying to use the 9.5 or 10.0 multiplier. I start by setting the voltage to the maximum possible value. This voltage is the appropriate voltage so your CPU can run the 10.0 multiplier reliably. Once this starts to make sense, you can try reducing the voltage one notch at a time.

For Dual IDA, you need to initially boot up with SpeedStep (EIST) disabled in the bios. After that, maximize the voltage, maximize the multiplier, click on Turn ON and then enable EIST. Next, manually disable EIST and see what multiplier your processor gets locked to. There is also an automatic feature. Right mouse click on ThrottleStop and have a look for a start Dual IDA menu option. The CPU has to be lightly loaded for this to work. Post some pics. It has been a while since I played with Dual IDA.