Clear "Limit Reasons" on launch

[Valour549]

When TS starts after my PC starts, a lot of the limit reasons are already present. Why this is I do not know, maybe TS reads everything even before TS starts?

Regardless, it means that if I want to check reasons later (eg. to see if I ever thermal throttle), I must manually clear the limit reasons every time my PC starts up.

Is it possible to add an option to clear all the limit reasons once when TS starts? Thanks

 

[unclewebb]

The Limit Reasons data is stored within the CPU. That means this information is being generated by the CPU whether ThrottleStop is running or not. Some of the yellow boxes can be triggered before the BIOS has had a chance to set up the various CPU power limit registers. It is completely normal to see some yellow boxes in Limit Reasons after first booting up.

Is it possible to add an option to clear all the limit reasons once when TS starts?

It would be possible to add an option like that but since you are the first person in 10+ years to ever ask for that option, I will probably never get around to adding this to ThrottleStop.

I must manually clear the limit reasons every time my PC starts up.

Yes.

When you are playing a game or doing something important, I would check the Log File box on the main screen. Boxes lighting up in Limit Reasons are good but a Log File does a more thorough job of telling the whole story.

 

[Valour549]

It would be possible to add an option like that but since you are the first person in 10+ years to ever ask for that option, I will probably never get around to adding this to ThrottleStop.

But come to think of it, this shouldn't even be an option: this should be the default behavior when TS launches.

After all, what is the point of reading Limit Reasons data that is already completely irrelevant, if such data is logged the moment the PC is turned on? Such data like the screenshot I posted has virtually no meaning at all.

Any meaningful monitoring, be it through looking at Limit Reasons or logging, only commences once TS itself starts.

 

[unclewebb]

this should be the default behavior when TS launches.

Maybe it should be but it is not.

Some users are interested in what throttling flags are being triggered during boot up. I know I am interested in that data which is why Limit Reasons works the way it does. I think you need to be concerned that your laptop appears to thermal throttle during start up and the voltage regulators are not able to supply enough current to the CPU.

 

[Valour549]

Maybe it should be but it is not.

Some users are interested in what throttling flags are being triggered during boot up. I know I am interested in that data which is why Limit Reasons works the way it does. I think you need to be concerned that your laptop appears to thermal throttle during start up and the voltage regulators are not able to supply enough current to the CPU.

Feel free to take a look at my settings and give any advice regarding the thermal throttle and VR current issues. I think overall it's decent because my laptop's i9-14900HX gets over 35k in R23 on a regular basis.

The only part I'm not very sure is the C state settings, regarding what exactly C1/C3/Package Demotion and Undemotion do. So I've just left it as the default, I think the OS will take care of it? I've deactivated Modern Standby and gone back to S3 if that makes any difference.

In TPL I have purposely left MMIO unlocked, for two reasons.

One, I have repeatedly found a strange bug where if I lock it, my FID multipliers will get locked to a lower value like 42 or even 38 upon the next startup, regardless of the fact my Turbo ratios in FIVR are set to be higher (of course the alternative solution is to set PL1 and PL2 to 175 and then check sync MMIO). Two, because TS currently does not offer TPL profiles, so when I go on battery I let Windows switch to a different power plan and that actually adjusts the MMIO (I can see the values changing), which actually serves nicely as a TPL profile.

BTW I want to double check Clamp and Lock under Power Limit Controls. So clamp forces the CPU to throttle if necessary so it stays strictly under PL1 and PL2 right? If we didn't clamp it then on occasions the CPU could actually draw values higher than what we've set? The Lock on the right hand side, that locks the MSR?

 

[unclewebb]

thermal throttle

If this is happening when you boot up before ThrottleStop starts then there is nothing you can do about it. The 14900HX is a powerful processor that puts out an unGodly amount of heat. I am not sure if there are any laptops on the market that allow these CPUs to run at full speed and full power without thermal throttling.

Your core and P cache undervolt look better than average. Are you 100% stable at -163 mV? If you have any light load instability issues, keep increasing the mV Boost value and see if that helps. Some users have mV Boost between 150 and 185. Extra voltage at low MHz is useful when pushing an undervolt as far as you are. This might also allow you to undervolt the E cache some more.

VR current

There is nothing you can do about this type of throttling. In a perfect world you should never see any VR Current throttling. This could be a sign that the manufacturer decided to save a buck by using voltage regulators that are barely adequate. More current is needed when you start increasing the turbo power limits. If you run a steady load like Cinebench, the voltage regulators might be OK at first but as the heat builds up within your laptop, the voltage regulators might start to trigger VR Current throttling. If you find that you have this problem then all you can do is lower your power limits to avoid ever triggering this type of throttling.

I have repeatedly found a strange bug where if I lock it, my FID multipliers will get locked to a lower value like 42

Try checking the Speed Shift box in the TPL window and set Speed Shift Max to the recommended value. That feature is designed to control the maximum multiplier.

Someone else with a Lenovo HX laptop just recently reported the same issue. ThrottleStop has not changed in the last year so I am guessing that there was either an Intel microcode change, or a Windows 11 change or a change to the default power plans that Lenovo uses. No way for me to know for sure without owning an HX laptop for testing purposes. Interesting that you found a way to work around this issue. When looking at a log file it almost looked like Windows was constantly writing a low value to the Speed Shift register to control the maximum CPU speed. If Windows writes to this register more often than ThrottleStop does, Windows will be in control of the maximum CPU speed.

Demotion and Undemotion

The purpose of these is it allows the CPU to decide how long to wait before going into a deeper C state when a core is idle. I am not sure what would be best for your HX CPU. For my 10th Gen desktop CPU, I ended up checking the three Undemotion boxes and I left the three Demotion boxes clear. This slightly increases C state residency percentages which in theory should slightly decrease power consumption. This made more of a difference 15 years ago when these controls were first designed by Intel. Modern CPUs enter and exit various C states so rapidly that it might be difficult to notice any difference no matter what demotion and undemotion boxes are checked or clear. I have rarely if ever seen anyone do any sort of testing to try and prove what would be best.

Clamp

When Clamp is checked, the CPU speed is reduced as much as necessary so the CPU does not exceed the turbo power limits that are set. When Clamp is not checked, the CPU will only throttle down as low as the base frequency. This feature is more relevant for Intel's 15W low power CPUs. Clamp checked or not should make zero difference to a high power HX CPU.

Lock under Power Limit Controls

Checking this box will lock the MSR turbo power limits so they cannot be changed until the next reboot. The vast majority of laptops never change the MSR turbo power limits after you boot up. If you never see the MSR power limits randomly change in ThrottleStop then there is no reason to lock these. Leaving Disable Controls checked is a good idea if you are confident that the MSR power limits are never being changed. With that box checked, ThrottleStop will never have to monitor or adjust the MSR power limits. This might improve the efficiency of ThrottleStop some microscopic amount.

 

[Valour549]

I am not sure if there are any laptops on the market that allow these CPUs to run at full speed and full power without thermal throttling. Are you 100% stable at -163 mV? If you run a steady load like Cinebench, the voltage regulators might be OK at first but as the heat builds up within your laptop, the voltage regulators might start to trigger VR Current throttling.

Yeah the PL limit of 175W will over time cause it to hit 100C in longer stress tests, though the biggest detriment is the fact that it will BD Prochot completely throttling the CPU for a good 15 seconds, and it's locked in the BIOS; I really wish there was a way to unlock it. On the bright side, I have never seen VR Current throttling apart from that reasons list on start-up.

I found the the i9-14900HX gets the best performance by limiting the turbo limits lower than the default 56-56-52-52-52-52-52-52 since the max all-core turbo it can sustain with 175W is 48. If I don't set it to 48 it seems like the CPU cores aren't smart enough to distribute power among themselves equally (like some will have FIR of 52 and others 42), and that yields suboptimal result at least in Cinebench.

-163 mV has been unstable on some very niche occasions like when running GPU stress tests, where the CPU has a low C0% but high FID values. So yeah I've turned it down to -154 mV with 120 mV on the boost.

Someone else with a Lenovo HX laptop just recently reported the same issue.

I am using ASUS Scar 18. I have decided that it's best to leave my TPL settings as is, and not force the MMIO to always be at 175.
- On a heavy CPU work load I've observed that MMIO always stick at 175W anyways.
- On GPU heavy loads, MMIO fluidly decreases between 0 to 30W and fluctuates. I assume this is the GPU Dynamic Boost at play here, taking some wattage from the CPU so the RTX 4080 mobile itself can boost to 175W.
- As mentioned when I pull the plug MMIO decreases to 110 which works great on battery.

I ended up checking the three Undemotion boxes and I left the three Demotion boxes clear.

Doesn't ticking Demotion mean going into a deeper C-state which would help decrease power consumption? Or have I got it the other way round.

When Clamp is not checked, the CPU will only throttle down as low as the base frequency.

Interesting. So not ticking clamp allows the CPU to ignore the PL limits and instead use the base freq as the limit. Does that disrespect only apply to the values you set under Power Limit Controls, or can the CPU ignore the MSR/MMIO limits too (assuming they are set low)?

 

[unclewebb]

I do not know what Demotion Undemotion settings will work best on a 14900HX. If you have a computer that does not have a lot of stuff running in the background, you can make changes in the C States window, press Apply and watch the C State monitoring table to see if you can notice any significant differences. It might be difficult or impossible to notice anything different. I have done zero C state testing on a 14900HX so I cannot help you with this.

So not ticking clamp allows the CPU to ignore the PL limits and instead use the base freq as the limit.

That is how my 10th Gen desktop CPU works. You will need to do some hands on testing to see if your 14900HX works the same.

Here is a test with both PL1 and PL2 set to 40W. The 10850K base multiplier is 36. When Clamp is not checked, when the CPU needs to power limit throttle, it will only slow down to the base 36 multiplier and not any lower. CPU power consumption is 51.2W so the 40W power limits are not being strictly enforced.

For the same test with PL1 = PL2 = 40W with both Clamp boxes checked, now the 40W power limit is being strictly enforced. The CPU throttles below the 36 multiplier to make sure the CPU does not exceed the 40W power limit that has been requested. I have always assumed that Clamp works the same on any Intel CPU, new or old. If you do some 14900HX Clamp testing, let me know if anything has changed.

unstable on some very niche occasions like when running GPU stress tests, where the CPU has a low C0% but high FID values

If I had this problem I would increase the mV Boost value up into the 175 mV to 200 mV range. The purpose of mV Boost is to boost the low end voltage to help avoid any light load crashes when trying to run a big undervolt.

Have you tried using different offset values for the core and the P cache? On older CPUs these two voltages did not need to be set equal to each other. Some users got better Cinebench results or better temperatures when setting the core to a larger offset value compared to the cache.

 

[Valour549]

If I had this problem I would increase the mV Boost value up into the 175 mV to 200 mV range. The purpose of mV Boost is to boost the low end voltage to help avoid any light load crashes when trying to run a big undervolt.

Yeah I have since applied a 160 mV boost and it has stabilized my UV for core and cache at -158 mV.

Have you tried using different offset values for the core and the P cache? On older CPUs these two voltages did not need to be set equal to each other. Some users got better Cinebench results or better temperatures when setting the core to a larger offset value compared to the cache.

I have tried, but I couldn't find any observable improvements without introducing instability, so I've just left both at the same.

In options, what does the Lock Prochot Offset do? If it's not ticked, does my Prochot Offset of 0 get dynamically adjusted or something?

 

[unclewebb]

Lock PROCHOT Offset

If you do not see a yellow lock icon near this setting, any software can vary the thermal throttling temperature whenever it wants to. Many Lenovo laptops do this. You only need to lock this setting if your computer has this problem.

 

[Valour549]

No wonder... I didn't have it ticked before, and I would observe on the main interface that the PROCHOT text would turn red, even though none of the CPU cores or package reached my PROCHOT of 100C. Ever since ticking it I've never seen PROCHOT turn red again.

Anyways, looking forward to different TPL profiles and possibly a clear reasons on launch option in the next version

 

[Valour549]

If you do not see a yellow lock icon near this setting, any software can vary the thermal throttling temperature whenever it wants to. Many Lenovo laptops do this. You only need to lock this setting if your computer has this problem.

What exactly causes the PROCHOT text to turn red? I cannot figure it out. I haven't reached the value, nor is there a thermal reason.

 

[unclewebb]

What exactly causes the PROCHOT text to turn red?

The package temperature only has to reach 100°C for a millisecond to trigger PROCHOT to light up red. Your screenshot shows that ThrottleStop reported a maximum package temperature of 97°C. ThrottleStop is only sampling this temperature data once per second. If you check the More Data box, ThrottleStop will sample this data more often and you might get another degree or two closer to the actual maximum temperature.

Core and package temperatures can rapidly change. It is not unusual that ThrottleStop or any other monitoring software will not be 100% accurate when reporting the maximum package temperature.

I would trust PROCHOT lighting up red more than anything else. PROCHOT turning red confirms that a temperature sensor somewhere on the CPU package reached 100°C.

 

[Emmnot]

Feel free to take a look at my settings and give any advice regarding the thermal throttle and VR current issues. I think overall it's decent because my laptop's i9-14900HX gets over 35k in R23 on a regular basis.

The only part I'm not very sure is the C state settings, regarding what exactly C1/C3/Package Demotion and Undemotion do. So I've just left it as the default, I think the OS will take care of it? I've deactivated Modern Standby and gone back to S3 if that makes any difference.

In TPL I have purposely left MMIO unlocked, for two reasons.

One, I have repeatedly found a strange bug where if I lock it, my FID multipliers will get locked to a lower value like 42 or even 38 upon the next startup, regardless of the fact my Turbo ratios in FIVR are set to be higher (of course the alternative solution is to set PL1 and PL2 to 175 and then check sync MMIO). Two, because TS currently does not offer TPL profiles, so when I go on battery I let Windows switch to a different power plan and that actually adjusts the MMIO (I can see the values changing), which actually serves nicely as a TPL profile.

BTW I want to double check Clamp and Lock under Power Limit Controls. So clamp forces the CPU to throttle if necessary so it stays strictly under PL1 and PL2 right? If we didn't clamp it then on occasions the CPU could actually draw values higher than what we've set? The Lock on the right hand side, that locks the MSR?

Can you send screenshots of your current settings? (my Razer Blade 16 can work without overheating if TPL is set to 130, if set higher, the CPU temperature is constantly 100°C)

 

[Valour549]

Can you send screenshots of your current settings? (my Razer Blade 16 can work without overheating if TPL is set to 130, if set higher, the CPU temperature is constantly 100°C)

It's in the description of the YouTube video