The ThrottleStop Guide

How responsive do you want the system to be? 0 to 30 feels just right.

 

Scroll to the top of the page and have a look at the picture I posted. A fast CPU or a slow CPU makes virtually no difference to idle power consumption or idle temperatures. When plugged in, why not go for maximum performance? You get this by setting Speed Shift EPP to 0.

In the old days, before C states like core C7 and package C8 existed, slowing a CPU down when idle was a necessity. Now, not so much. Whenever a modern CPU has nothing to do, it will enter one of the low power C states automatically.

When running on battery power, an EPP setting of somewhere between 80 and 128 should be OK. That will slow your CPU down when idle but still allow the CPU to ramp up to an efficient speed when it has something to do. That is the whole point of Speed Shift Technology (SST).

I have not done any real world EPP battery testing. Slowing a modern CPU down when running on battery power might not make a huge difference to power consumption or battery run time. A slow CPU can reduce peak power consumption. If a task takes 2 or 3 or 4 times longer to complete, total energy consumption might be very similar or worse with a slow and inefficient CPU. It depends on what type of software you are running. It is always best to do your own testing on your own laptop running whatever software that is important to you.

 

iv'e done some testing using 800mhz vs 2.5ghz my battery life estimates are lower when downclocking this is with battery bar so unclewebb is right slowing cpu does not help

 

i used the tweaks and lost my camera and cortana not cool....after un doing the tweaks I gained it back and tweaked shutup im getting 6 c state and thats not bad witha 2.4 w consumption i recommend tweaing the settings to get better performance while keeping everything

 

has anyone tested this?

 

I did not do exactly that. However, due to some issues I've been having with my keyboard I was desperate to try a BIOS update on my Hp Spectre 13". Someone else used TS after the update so I felt ok with the limited risk. Well a few hours later I never could get it to take, tried the windows method, tried the local file on USB, even tried using the horrible HP assistant tool which I hate and had never even installed. Nothing worked. So I enabled windows update, got a firmware version 15.0.0.0 which doesn't jive with the BIOS number but like I said I was desperate.

This didnt fix my issue either, but i noticed TS was stuck showing my same undervolt and I could no longer change it no matter what I tried. Could not change multipliers either. I thought I was screwed, but I simple rolled back the firmware driver in device manager and it went back to normal.

Looks like I'm stuck on my current BIOS. My luck it will magically start working and I'll get undervolting locked out.

 

Could program crashes be attributed to too excessive of an undervolt? No BSODs but Edge and Task Manager seem to be crashing more frequently than they used to.

 

@djdelarosa25, any crashes should make the undervolt suspect. The idea is not to trade heat/performance/battery life for an unstable system.

Undervolt as much as possible, but not too much.

 

2 more G3 users got back to me yesterday saying it worked.

 

exciting news!!

I disabled everything thats useless and kept cortana search ect and my c0% is below 1 often 0.6 and this just proves you don't need to make windows gimped in order to have stellar battery life and also disabling un-nessessary services with eso and disabling basic garbage with shutup10 my 3dmark score and cpu mark score bolted into 10 000 on cloudgate and 21fps for cpu with it locked at 2.5ghz vs the stock 3.4ghz so no thermal throtling, gains in gaming and incredible batterylife improvements for example my laptop is supposed to get 9 hours with minimum things running and now im getting 11-18 hours easily with moderate usage. so yes anyone can do this and if you need help just give me a jingle

 

@unclewebb I just wanted to say thanks again for everything you do. I have another success story from today:

I wanted to revive my girlfriends old laptop to use in our bedroom TV to watch shows/movies/youtube on, it hadn't been turned on in 2 years. GF didn't even realize that we still had it. It was filthy, covered in dust and some wax or something spilled on top. I cleaned it off, booted it up and updated windows 8.1 the best I could. I had it working decent but was pretty slow when doing more than one thing at a time.

Come to find out the CPU was stuck at 800 mhz. Tried fiddling with power settings, min/max processor states and nothing seemed to help. Until I put TS on there. I few clicks and I had an underclocked, full speed CPU from 2013 (i3-4005U). Plays movies and youtube great and cost me zero dollars, and an hour of my time in total! One less computer in the dump too which is a win.

The issue appears to be the PROCHOT, once I disabled that it worked fine. I did notice no Speedshift EPP was available so I assume it's because of the 4th gen CPU. I just put it in balanced mode so it can ramp up and down as needed.

Thanks!

 

And how are the temps? Might be needing a clean and repaste..

 

I think you mean BD PROCHOT. On older computers, when a sensor goes bad, it will send a permanent throttling signal to your CPU via the BD PROCHOT signal path. Disabling this box in ThrottleStop allows your CPU to ignore external throttling requests so it can run at its full rated speed.

PROCHOT is different. That signal happens within the CPU when the CPU actually gets too hot. Your CPU will still be able to thermal throttle regardless of whether the BD PROCHOT box is checked or not. BD PROCHOT just blocks the external throttling signals which seem to be false about 99 times out of 100 or maybe more.

Always great to hear of another ThrottleStop success story!

 

From my limited testing it looked like it was running nice and cool, below 60c after a 3 minute TS bench on all cores.

Yes @unclewebb you are correct, it was BD PROCHOT. Great explanation of what it is it and does.

 

Hi everyone !

I still couldn't fix "VR CURRENT" issue in Limit Reasons.

Before I was using 4.8Ghz with lower voltage and it was stable in Cinebench, no VR CURRENT error. But it was not stable in TS Bench, was giving error. Even some games were giving errors.

Now it's stable in Ts Bench but not in Cinebench and with 4.7Ghz.

don't know what else to try

 

lower your undervolt? not rocket science

im going 40 over the speed limit everytime i drive, why do I always get speeding tickets? same logic

 

@onrblbl - You have your CPU living on the edge of stability. It is not unusual for a CPU to need slightly more voltage as it wears in. When you live on the edge, tiny changes in your room temperature or which way the wind is blowing can change stability. There is no guarantee that an undervolt setting in ThrottleStop is going to have your CPU getting the exact same actual voltage all of the time. Maybe that voltage regulator that you are pushing to the limit is not providing 100% clean and consistent voltage to your CPU.

PL1 lighting up in Limit Reasons is a little unusual. Try checking the FIVR Disable and Lock Turbo Power Limits feature. I would also check the TPL - Turbo Boost Short Power Max box. When this box is not checked, the power limit information you have set in ThrottleStop is not being sent to the CPU. That means the actual value in the CPU register could be pretty much anything.

 

can you tell me which version or microcode they stopped at?

 

@unclewebb - You're right. I've done the same test after fresh start like every part cooler it's past the test without VR CURRENT error.


@unclewebb - I assume "VR CURRENT" error comes when temps hit 88C. I've searched in this forum about TS Bench and i am still not sure if i would consider being %100. Cause like now if i'm %100 stable in TS Bench then i can't decrease voltaj like i desire.

 

@onrblbl - Stability is temperature related. As the CPU cores or the voltage regulator heats up, stability decreases. High temps will limit your overclock. That is why extreme overclockers reach for the liquid nitrogen. They know if they can keep their CPU cool, they can get a few more hundred MHz out it.

The TS Bench test is not widely used but I think it has proven to be a good test of stability when setting voltages. Some tests like Prime95 hammer the CPU with AVX instructions. The TS Bench test hammers the other side of the CPU. It does not use any of the AVX instruction set. I remember one user in this thread told me that he decided to ignore a couple of errors reported by the TS Bench. No big deal. In a short amount of time he had to reinstall Windows because it had become corrupted.

Would you like me to tell you to go ahead and lower the voltage? A couple of TS Bench errors is no big deal? I cannot do that. I think any errors are a big deal. The test is literally doing the exact same calculation twice. If sometimes your computer thinks that 2 + 2 = 4 and sometimes it thinks that 2 + 2 = 5, that's a problem. It might not cause a BSOD error today or next week but at some point it probably will. To me, 100% stability is more important than living on the edge.

Edit - Here are a couple of posts that I remember. Maybe the user did not have to reinstall Windows but he does say that he lost a lot of his work after deciding to ignore a couple of TS Bench errors.

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-1020#post-10919625

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-1021#post-10919803

 

@unclewebb - I have just tried multiplier 48 with TS Bench to catch %100(-90v). It was too hot compare to 47.(-127v)
For 100Mhz i'll not take risk so stop it.

Thank you a lot for your help as always you're number one.

 

unclewebb do you know how to improve igpu performance, I was getting 11000 in cloudgate now its down to 9800 and literally nothing has changed

 

Whenever I read that "nothing has changed", I always read that as, "something has changed". It is your laptop so I do not know what you changed or when you changed it.

Low power U series CPUs are designed to throttle either the CPU or the iGPU or a little of both depending on your settings. When testing, turn the Log File option on in ThrottleStop so you can see if there are any CPU throttling problems. You can also try logging the iGPU speed with GPU-Z.

When trying to maximize performance, you have to learn to change one thing at a time and do lots of testing and closely keep track of everything you do. That way you will be less likely to see a sudden 12% change in your benchmark scores with no idea why it happened.

 

thats graphics score of 9800..

cpu score is up a bit not sure, but I suspect undervolting and turning off unnessessary programs/services...I am out of options and changing power limits seems to do nothing as well as turbo time limit...with my hd 615 it did wonders but with the uhd 620 it does nothing perhaps I need a newer version of throttle stop im currently using 8.7

 

Cloudgate is highly CPU speed dependent, it's likely a speed shift/other power configuration that's causing your drop in performance. If you think it's a throttle stop setting, the burden is on you to figure that out. Don't run TS on startup and restart your system, see what happens. Change ONE setting at a time and see what happens...

 

oh ok I'll do that, do you know how to reset throttle stop? and thanks again

 

You can rename throttlestop.ini to backup.ini or something to keep a backup.

 

The last working BIOS for the Dell G3 was 1.8.0.
It should be uCode 0xB4.
ThrottleStop works on these.

 

What's a good free program to test how stable my iGPU undervolt is?

Is it even worth it to undervolt the iGPU?

 

Just found that throttlestop can be used to fix my NUC6i5syk that was experiencing a voltage regulator fault causing non stop BSOD "whea uncorrectable error".

Details of this error are here:
https://nucblog.net/2016/04/skylake-i3-and-i5-nuc-whea-errors/

Updating the bios did not help (which is expected as the bios update would only ensure the fault didn't occur, once the fault kicked in then the bios update was useless).

The fix was to increase the cache voltage using the FIVR menu in throttlestop (in my case I increased by 125mv).

I find it interesting that Intel engineers couldn't figure out that this was all you needed, more cache voltage. They probably could have created a new bios with this option if they knew...

In order to boot into windows without a BSOD I had to change the current limit option in the bios to the lowest setting (4 amps). This would let the Nuc boot but the cpu would be stuck at 800mhz.

After increasing cache voltage by 125mv to ensure stability and no BSOD, I was then able to bump CPU core current up to the max which gave me back full turbo on the cpu.

I have also used UBU to roll back the microcode to a 2015 dated one and unlocked options in the bios via the setup IFX extraction / modification method.
The method is found here:
https://puissanceled.com/vrac/Bios_modding/EN.html

In my case the PE32 image section containing the setup forms was called "dpsdsetup".

Now I'll work on modding the ME section to enable the turbo overclocking option in the FIVR menu hopefully. As this is skylake on an old microcode I want to use the multiplier bug to see how high the frequency can go past stock.

 

@moral hazard - That is an interesting use of those ThrottleStop features. Good work.

I found an interesting problem over on TechPowerUp.

https://www.techpowerup.com/forums/threads/throttlestop-not-working.265055/

It looks like some user got hit with a BIOS update or perhaps a Windows Update that set all of the turbo ratio limits on his 7700HQ to 0. Worse than that, it also set the lock bit so he has lost Intel Turbo Boost as well as voltage control. These supposed engineers with all of these great ideas need to be taken to the wood shed.

 

Hi @moral hazard,
could you please reveal how you want to modify ME so you can overclock.
Because want to do the same thing.
I can't see any section ME regarding to OC.
Just was able rewrote few of registers in order to enable OC BLCK at ICC profile section, but even this could not be achieved on H61 board. Just up to100 MHz.
When I experimentally disabled ME (partly) I lost my 100 MHz BCLK for 99,75 so was reverting back to ME..
Sorry for my bad English

 

I was hoping to experiment using this method;
https://www.bios-mods.com/forum/Thread-request-b85-g43-i5-4460-OC-Unlock?pid=146601#pid146601

 

My attempts to undervolt the iGPU have so far been futile. I've pushed it passed -300mV and nothing happens, even while running a stress test on it.

 

When under volting the iGPU, you have to under volt the iGPU and the iGPU Unslice equally or your under volt will be ignored. That is likely the problem that you are having.

 

Thank you for this. I'll give it another go.

Managed -100mV. This is surprising. I've read people only reaching about 50mV.

 

Hi Sir,
My system is Lenovo Legion Y740 with i7-9750H (Undervolt vcore and cache -150mv)
When I set the EPP value under 128 but set the TPL value high (Long = 60W, Short = 90W) to avoid the Power Limit Throttling issue, I get bad temp (90+ degrees C) and the result is Thermal Throttling issue.
Right now I have 2 options after intensive test to choose:
1. Set EPP 128 and TPL 70-90, No Power Limit and Thermal Throttling (Max temp 75c). But Benchmark scores are below standart and when playing games the clock speed is limited 3Ghz only.
2. Set EPP 75 and TPL 40-50, Power Limit Throttling occur but Thermal is safe (Max temp 88c). Bench scores are Normal and CPU Clock Speed can run above 3Ghz, Around 4Ghz to minimum 3,5Ghz when PL Limit active.

What should I do?

 

I read the threat you posted and also little bit about your machine.
If I remember the multiplier bug is also on Skylake but dont know if U - cpu are included at, so let me know how was it.
I myself am afraid to boot machine less uCode, because it will take me about 2 hours to dissassembly and flashing backup in case I was wrong
About ME, I am not sure you could change ratio by working on different version ME and you probably can not simply skip main version, lets say from version ME11 to 12.
But I believe you can find about BCLK OC Skylake to 103 MHz exactly on the same site as you posted.
Here is thread about pwne your NUC6i5syk machine code http://blog.cr4.sh/2016/
Good luck and sorry for my bad english
Dan

 

Your Lenovo Legion Y740 has inadequate cooling. You will not be able to run all 6 cores of your 9750H at their full rated speed indefinitely. This unfortunately is normal for laptops with this CPU. Some users with this problem will take their laptops apart and will apply new and improved thermal paste. Thermal Grizzly Conductonaut is a popular way to improve cooling.

https://www.thermal-grizzly.com/en/products/26-conductonaut-en

Keep in mind that any CPU temperature under 100°C is safe according to Intel. That is why Intel sets the thermal throttling temperature to 100°C. Your CPU will automatically and continuously adjust its speed just enough to always keep your CPU from exceeding 100°C. I have tested Intel thermal throttling and it works great.

https://ark.intel.com/content/www/u...9750h-processor-12m-cache-up-to-4-50-ghz.html

Look at the main screen of ThrottleStop. Does it show PROCHOT 100°C or did Lenovo reduce the thermal throttling temperature for your laptop? Any number less than 100°C will force your CPU to start thermal throttling even before the Intel specification. In other words, it will be extra safe.

Edit - I saw one screenshot of a similar Y540. Lenovo dropped the throttling temperature to 94°C. In my opinion, that is way too conservative but they can do whatever they want. As a consumer, I would shop elsewhere.

I do not think it matters whether your CPU thermal throttles or power limit throttles. Set ThrottleStop up however you like so your laptop provides the best overall gaming experience and stop worrying about your CPU temperature.

 

I am curious about this myself, might take some time to get back to you with an answer as I am still learning.

That information is gold, Thanks Dan!

 

Yes, lenovo set the BDPROCHOT value at 94c.
Thank you for the enlightment

 

Lenovo set the PROCHOT value to 94°C. PROCHOT and BD PROCHOT are two different things.

Setting the PROCHOT temperature to 94°C reduces maximum performance. This is not something a company should be doing on their performance oriented laptops. That is kind of like putting a rev limiter on a car and setting it 1000 rpm lower than what the engine is capable of.

 

Thanks for the correction, yes I mean prochot... Sorry for the typo, i just woke up when i write that down

 

I have an XMG PRO 15 (Early 2019) / Clevo PB51EF-G and I'm trying to enable Hardware Duty Cycling (HDC) but it keeps auto disabling itself when I click okay. Why is this happening?

 

Not at all, It is ok I am also learning, for example I am stuck at very first step trying to port Chipsec somehow

 

It is possible that I never got around to finishing this feature.
I think HDC has been around for a year or so and you are the first person to ever mention anything about it.

My 4th Gen laptop that I use for programming does not support HDC so I might have accidentally forgot about finishing this feature. I have access to a 7th Gen desktop computer so I should be able to properly test and get HDC working in the next day or two.

Maybe this feature will improve laptop battery run time a little.

 

I get -160 easy on my Dell 9550 6300HQ. If you look on the XPS forums here you can see averages for that laptop which very stable probably range -130 to -160. Nobody was stable past -170 and older CPUs might require less aggressive undervolts.

Also repaste as your CPU is old now.

 

Do you guys stick core and cache values?

 

Okay so I'm confused. Some say to UV the core more than the cache, some say the opposite, while others say they should be set equally. Even @unclewebb himself provides contradicting statements regarding this.

 

Some CPUs will ignore undervolt if core and cache don't match. You can test that pretty easily with a benchmark. I tested my 6300hq but don't remember the results lol. I run core & cache at -160mv.

Also, I do not undervolt the Intel GPU. It was causing some DPC latency issues which are important to my live audio work.