Intel i7-8665U Throttling issue? Suspected cores stuck in sleep after laptop sleep

[ShadowXVII]

Hi all,

Long time lurker -- first time poster. I decided to post here after seeing the insightful blog post around Defender & Kernel Counters from @unclewebb... Alas, those are all reporting "normal".

I have a Dell Latitude 7400 that seems to get all but 1 of it's CPU cores stuck in sleep after waking up from sleep, but only on occasion. The only remedy I've found so far is to unplug the power -- this causes whatever firmware to realise the cores are asleep and wakes them up. Plugging the power back in, and it stays awake.

The CPU is throttled to ~0.5ghz, 1 core. And, everything as expected runs like a dog until I perform my power unplug/plug workaround.

I've tried raising with Dell, but I changed my SSD and now they won't have a bar of it as it's not OEM. So, I'd like to go to their engineering team with something more concrete, or find a way to bump it out of this sleep state.

• The power adapter is OEM and within spec, 65W.
• Happens on both Windows 10 and Windows 11, making me suspect it's a firmware level issue.
• Windows Performance Analyzer can run, but I don't know what to pick out of the details.
• This only happens when coming out of sleep. Sleep could be as short as 10 minutes, or overnight.
• No memory pressure. No Disk pressure.

Here's a before and after the workaround. I managed to run HWInfo after about 10 minutes of starting up
Aside from the obvious increase in Core CPU and temps being unleashed from 50c to 100c, what else can I use to prove the core states are in sleep?

Before​

After​

I'll let you play "pin the tail on the cpu graph" where I unplugged the power

Regards,
Jake

 

[Ferd]

I had similar “unexplainable “ experience with a relatively old Dell (4th gen -4800mq) , I couldn’t figure out the weird behavior, in my case the cpu was running at base clock by default, and would only boost to the max frequency after a “sleep , wake “ cycle , everytime I used that laptop I had to put it into sleep and wake it again in order to see it boost up to intel’s cpu specs .... probably a bios limitation imo .

 

[unclewebb]

Run ThrottleStop 9.5.

ThrottleStop 9.6 Download

ThrottleStop is a small application designed to monitor for and correct the three main types of CPU throttling that are being used on many lapto

www.techpowerup.com

Exit HWiNFO and the Task Manager when testing. Post a screenshot of the main ThrottleStop window and the FIVR and TPL windows so I can see your settings. Also post a screenshot of the Limit Reasons window while your CPU is throttling.

All active cores in your Intel CPU are locked to the same speed. You can have active cores and inactive cores but you cannot have two cores side by side running at two different speeds. I am not sure what HWiNFO is reporting so it will be interesting to see what ThrottleStop shows.

With Dell laptops, it is typically a BD PROCHOT throttling problem. If you see BD PROCHOT lighting up red in Limit Reasons, try clearing the BD PROCHOT box on the main screen of ThrottleStop.

Edit - The C States window might also show something interesting. Include a picture of that window while idle.

 

[Count von Schwalbe]

Hm. I had a similar laptop briefly. I cannot remember the model precisely - maybe an Inspiron 15? - using a 7700U processor. The power brick was dodgy, and plugging it in caused the BD PROCHOT warning to have a fit and throttle me to ~400 MHz.

The only solution I found, short of replacing the power brick, was to use throttlestop to disable PROCHOT throttling. Thanks @unclewebb for your handy (and free) utility!

 

[ShadowXVII]

ThrottleStop was actually my first point of call, and I think I replicated on a fresh install. I sent this info to Dell tech team from a fresh install of W10.
My understanding of C states is new, so forgive any misinterpretations.

Here was my email:

I had to try several times, but putting the device to sleep, then waking it up, caused it to get stuck in C7 state.

I’m guessing that’s what I’m seeing from the Core C States (40% time in C7) despite being resumed from sleep for a minute+.

I’m going to keep trying, but it seems like something is causing the CPU to get stuck in a low-power state after standby and doesn’t resume C0 power levels.

After power plug change, C0% started increasing normally (I accidentally cleared, so c7 dropped back to 0). You can see where I changed power ¼ through the graph.

Also, a reply from the creator himself . Thanks for all your hard work and low-level utilities. Appreciated!

Regards,
Jake

 

[unclewebb]

caused it to get stuck in C7 state

Here is some C States background info.

When a CPU core has a task to perform, it enters the active C0 state. When it finishes a task and if there are no more tasks that need to be performed, it will enter a low power C state such as core C7. A core spending time in the low power C7 C state saves more power than a core spending its idle time in the core C3 C state. This saves more power compared to the same core in the active C0 state or the inactive C1 state.

When all cores enter a low power C state at the same time, then the entire CPU package can also enter a low power C state. Think of this as the CPU is smartly turning off various parts of itself that are not being used. This reduces power consumption when idle and also when the CPU is working on something. Many tasks are single threaded so you do not need to have all 4 cores and 8 threads actively doing something if there are no tasks in the queue that need to be completed.

When a computer is idle at the desktop, the individual cores should be spending the vast majority of their time in the core C7 state. This saves power and allows the CPU to run cool. Less power consumption means less heat is being produced. When a computer is idle, high C state percentages are desirable. This improves battery run time and allows a person to use a laptop on their lap without it burning through their lap.

Here is an example of my desktop computer which has a 10 core 10850K.

The vast majority of cores and threads are spending virtually zero time in the C0 state processing background tasks. I never thought a 10 core CPU could run this cool and efficiently when idle. A cool CPU is a quiet CPU. It is very powerful when loaded but virtually silent when idle or lightly loaded.

Here is what the C states window reports for my computer.

The individual cores are spending more than 99% of the time in the low power core C7 state. That is why my computer runs so cool when idle. The cores are not bogged down running background tasks, most of which are completely unnecessary. You might need an advanced course on how to hunt down useless crap that is running on your computer. I have not yet written that course but it looks like I really need to.

To do some proper testing, exit the Task Manager and exit that stress testing program I see on your desktop. Exit everything except ThrottleStop. What do you see for average C0% when your computer is idle? A typical computer with a 4 core - 8 thread CPU should not need to average more than 1% of its time in the C0 state. If your C0 number is way higher than that when your computer is idle then open the Task Manager, go to the Details tab and find out what is running on your computer. Click on the CPU heading in the table to organize the running tasks by CPU usage. Post a screenshot of that if you need help understanding what you see. System Idle Process at 99% looks good. With only Google Chrome running, minimized to the task bar, my computer truly is idle.

The data in the ThrottleStop FID column shows that none of your cores are stuck running at a low speed. I am not sure what HWiNFO is reporting. It does not seem to accurately report the speed your individual cores are running at. ThrottleStop closely follows the Intel recommended Core i CPU monitoring method. You can trust what ThrottleStop is showing you. All I see is a completely normal CPU bogged down by too much stuff running in the background.

ThrottleStop also shows a computer with barely adequate cooling which is typical for many Dell laptops, especially the Latitude line. The first screenshot shows the CPU is reaching 100°C when peak power consumption is only 19.3W. The 8665U has a 15W TDP rating. It looks like the cooler Dell is using in your Latitude is right at the bare minimum. Have you cleaned out your laptop recently? The heatsinks might be full of dust and dirt.

With proper cooling and unlocked power limits, Intel's low power U series CPUs are capable of so much more. Here is an example of a Lenovo laptop with a similar 8th Gen U series CPU. It can pull almost 40W at full load before it reaches the thermal throttling temperature.

Edit - After you clean up the background tasks, try showing some ThrottleStop screenshots before and after you unplug the power. Include the Limit Reasons window in case it shows a reason for any throttling. Make sure HWiNFO is not running on your computer when gathering data with ThrottleStop.

Edit - Your first screenshot shows the CPU is thermal throttling. I guess HWiNFO reports that as individual cores being stuck at 382 MHz. That is not what really happens during a thermal throttling episode but at least it makes sense now.

During normal thermal throttling, the CPU cores only slow down the bare minimum. All active cores will continue to run as fast as possible without the temperature exceeding the thermal throttling temperature which your laptop has set to 98°C. Fix your cooling. That is the main problem. Getting rid of background tasks will reduce the stress on your under designed cooling system.

 

[ShadowXVII]

Normally, I'd agree with the temperature being the issue here, however the laptop isn't hot, nor is the fan running.
There is ample airflow under the laptop, and I've even tried a cooling pad.

Here's a screen capture of ThrottleStop when this happened recently while being used. It suddenly dropped to this "1 core, 0.5ghz state"
Note the temp is only reported at 55C and usually sits around 65-70C for this machine.

I haven't found a behaviour pattern that causes this state.

 

[unclewebb]

If you believe BD PROCHOT throttling should not be happening, clear the BD PROCHOT box on the main screen of ThrottleStop.

Sensors can fail and can send throttling signals to the CPU. Turning off BD PROCHOT tells the CPU to ignore any external sensors. The CPU will still thermal throttle if it ever gets too hot whether BD PROCHOT is enabled or disabled.

 

[ShadowXVII]

I don't believe it shouldn't happen at all, but in this instance, dropping suddenly to 0.5ghz and being stuck at 55c until AC power unplug, seems strange.
For my understanding, BD PROCHOT is external sensors where as just PROCHOT and HOT are CPU based limits (next to the LIMITS button) -- I only see THERMAL in Limits Reasons.

I'm pretty across what's running in userspace, so I'm confident it's not a rogue process.
Though I wouldn't say something kernel-level isn't interfering. Hard to say with tools available.

I'll continue to log/monitor with and without BD PROCHOT.

Here's idle.

Having a look with my issue and BD PROCHOT specifically, seems there might be a common issue with the sensors going bad and reporting BD PROCHOT when it's in-fact fine.

i.e. https://www.dell.com/community/XPS/XPS-9560-Stuck-at-0-798-GHz-BD-PROCHOT/td-p/7350884 and https://www.reddit.com/r/Dell/comments/adv2m3
Interesting that a full motherboard replacement didn't fix it. Or, the Dell thermal sensors are crap...

 

[Count von Schwalbe]

Have you tried a new power adapter? If it only happens when the AC adapter is connected, I bet something in that is causing a fit.

 

[unclewebb]

Interesting that a full motherboard replacement didn't fix it.

Often times a "motherboard replacement" is not a new motherboard. It is just recycled crap that was previously returned. If a tech plugs it in and it starts up, it would not be properly tested for BD PROCHOT throttling issues. It would then be put back in their inventory as a good board. I am not sure if Intel has any readily available tools to even test for BD PROCHOT. I have heard that they do not. I do know that BD PROCHOT throttling is not well documented and it is not well understood because of this. When this happens, Intel XTU simply reports the problem as thermal throttling which makes no sense when it can happen at very low idle temperatures.

If I had a board with this issue, I would clear the BD PROCHOT box in ThrottleStop and then I would move on. No one has ever contacted me to tell me that disabling BD PROCHOT destroyed their motherboard. Quite the opposite. Many, many people have let me know that their computer when stuck at 798 MHz was completely unusable. Disabling BD PROCHOT saved them from throwing their computer in the garbage.

On MSI desktop motherboards, it is as simple as a little bit of dust build up between two pins or two solder points on the board. When there is enough dust in there, the switch shorts out and constant BD PROCHOT throttling is the result.

a new power adapter

Some Dell power adapters use a switch that can be used to trigger BD PROCHOT throttling. The purpose of this switch is so the computer can determine if you are using a genuine Dell OEM power adapter. If a generic power adapter is detected, BD PROCHOT throttling begins and never ends. When this switch in a Dell power adapter goes bad, it can cause the same throttling problem. You can waste your money on a new power adapter but why bother? I would rather use ThrottleStop to disable the BD PROCHOT signal path. That costs nothing. Users can send me half the money they just saved by not buying a new Dell power adapter.

 

[Count von Schwalbe]

Some Dell power adapters use a switch that can be used to trigger BD PROCHOT throttling. The purpose of this switch is so the computer can determine if you are using a genuine Dell OEM power adapter. If a generic power adapter is detected, BD PROCHOT throttling begins and never ends. When this switch in a Dell power adapter goes bad, it can cause the same throttling problem. You can waste your money on a new power adapter but why bother? I would rather use ThrottleStop to disable the BD PROCHOT signal path. That costs nothing. Users can send me half the money they just saved by not buying a new Dell power adapter.

Yet another reason to stay away from OEM stuff... Pity there is no standard for laptops for DIY. Anyways, Throttlestop did certainly fix it when I had that issue. I wonder if a bad switch issue would be fixed by unplugging and re-inserting the power adapter? In my case, that did not help.

 

[unclewebb]

Some defective switches that are in the process of failing can trigger BD PROCHOT randomly. Things might work 100% correctly one day and your computer might be stuck at 798 MHz the next day. If you ever see BD PROCHOT lighting up red in Limit Reasons, I would immediately clear the BD PROCHOT box on the main screen of ThrottleStop. This is an overused throttling method that is rarely if ever needed for any legitimate reason.

Many 12th Gen motherboards are now locking the BD PROCHOT setting so it can no longer be disabled. When you get BD PROCHOT throttling on one of these boards, if the warranty is up, you can take your motherboard and throw it in the garbage. Planned obsolescence. These issues have already been reported on 12th Gen desktop motherboards.

 

[ShadowXVII]

Interesting discussion. I've swapped out adapter just for testing purposes.

Also, incase anyone is wondering the Dell Power Manager does have Thermal Profiles which I usually operate in Optimized, but I might set to Ultra Performance. (Behind the scenes SMBIOS config)

Short of that, using DellFanCmd to force the fan to 100% is good for thermals if you don't mind the noise