Alienware m15r6: weird power limit throttle after 30 mins of gaming

[kakashidinho]

I am having a weird throttling issue with my m15r6. Basically the CPU downclocked to 3.2 Ghz after 30 mins of gaming, XTU reported that it was due to power limit throttling. However, the thing is that Alienware is supposed to disable power limits (PL1 and PL2) in their laptops. I checked with Throttlestop, both of these limits are disabled. It seems Dell hard coded a hidden power limit (50W or some sort) after 30 mins of gaming. See the screenshot.

Based on the XTU report, there was no thermal throttling (the temp was 87c which was fine). So the downclocking was not because of high temp.

FYI: I already unlocked undervolt on this laptop and undervolted its CPU by -50mV.

 

[unclewebb]

However, the thing is that Alienware is supposed to disable power limits (PL1 and PL2) in their laptops.

Where did you hear that? Can you post some information from Dell that confirms that? It does not mean anything if a random person on reddit posted that info.

Hold the Shift key down on your keyboard and select Restart in the Windows menu. This will reset the CPU. After you boot up, run ThrottleStop and post a picture of the TPL window. This will show if the primary and secondary turbo power limits are disabled or not and it will show what they are set to.

Many Dell laptops use a third set of turbo power limits which are managed by an embedded controller (EC). This power limit can randomly be set to 50W for any reason at any time and there is nothing you can do about it. Intel XTU and ThrottleStop do not have any access to control the EC power limit.

The 11800H has a 45W TDP rating. Dell is not likely going to fix any power limit throttling problems that are happening when the CPU is above its rated spec. It is throttling by design.

Product Specifications

quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.

ark.intel.com

 

[kakashidinho]

I already changed PL1 PL2 using throttlestop to 70 & 100 for a while (before that the "Disable turbo power limit" in TPL panel was checked by default hence I assumed Alienware disable it). However, after prolong gaming session (both GPU & CPU are full loaded), the laptop automatically forces power limit throttling (<=50W) for no reason (the temp is still fine).

Here is the screenshot of TPL panel where I configured PL1 & PL2.

In the log file, I saw these lines:

2022-03-25  04:11:24  36.30   79.2  100.0       0   94   1.1100   49.5   PL2
2022-03-25  04:11:25  36.08   78.1  100.0       0   93   1.1100   49.7
....
2022-03-25  04:55:22  37.56   42.0  100.0   18422   79   0.8999   37.8   PL2
2022-03-25  04:58:39  29.13   10.4  100.0   14410   60   1.0800   16.5
....
2022-03-25  05:05:39  28.64    3.6  100.0   14410   42   1.1399    8.7
2022-03-25  05:06:39  25.62    3.4  100.0   10716   43   0.6775    7.6
2022-03-25  05:07:39  38.16    6.5  100.0   10716   45   1.1299   13.3
2022-03-25  05:08:39  36.85   41.2  100.0   10716   82   1.0299   36.3   PL2

It seems the laptop force PL2 throttling even at 37W.

 

[unclewebb]

The MSR and MMIO power limits are set appropriately in ThrottleStop.

If you are going to log data, do not use the Stop Data feature. When Stop Data is enabled, your CPU is only being sampled once every minute. Attach a proper log file so I can have a look.

It looks like Dell is power limit throttling your computer and there is nothing you can do about it. That is why I have avoided this brand for the last 10+ years. When you pay extra money for an Alienware laptop, you should be able to run it however you like. Dell does not agree with that philosophy. You will need to shop elsewhere next time if you are interested in maximum performance.

 

[Dr. Dro]

That is quite bizarre. My G15 Ryzen Edition (uses an Alienware design as well) with a R5 5600H quite happily runs full tilt all day, and that is with pretty hefty overclock on the GPU, too. No throttling whatsoever. Did you download Alienware Command Center and enable the high-performance mode for your laptop? Mine has this G-Mode thing on AWCC that seems to increase the power budget and kick the cooling into high gear when its toggled on, and I pretty much get 4.2 GHz on the 5600H non stop with that. Does it also exhibit the same behavior if you intentionally max out the cooling speed? I did notice that even under this situation you are getting 94 C maximums, and that is no good.

Just brainstorming a bit, because that seems outright foreign after the amazing experience I had with my laptop.

 

[unclewebb]

G15 Ryzen Edition

Big difference. The throttling methods that Dell uses in their Intel based laptops are unique.

 

[Dr. Dro]

Big difference. The throttling methods that Dell uses in their Intel based laptops are unique.

You're making me feel like I dodged a bullet here, man.

I just read your earlier post about the EC being responsible for the power management issuing a specific set of turbo limits and that kind of explained everything... unless someone finds a way to get into that black box, and understands how it works, I agree there is little that can be done. In OP's situation, I would then opt to disable the turbo mechanism or perhaps SMT to keep power consumption under the rated TDP. It sure sucks to do that, though... Cheers, man, and thanks for everything... have faithfully used ThrottleStop for years before I finally decided to stop lurking and join, you made my old Ivy Bridge HP a lot more bearable

 

[kakashidinho]

That is quite bizarre. My G15 Ryzen Edition (uses an Alienware design as well) with a R5 5600H quite happily runs full tilt all day, and that is with pretty hefty overclock on the GPU, too. No throttling whatsoever. Did you download Alienware Command Center and enable the high-performance mode for your laptop? Mine has this G-Mode thing on AWCC that seems to increase the power budget and kick the cooling into high gear when its toggled on, and I pretty much get 4.2 GHz on the 5600H non stop with that. Does it also exhibit the same behavior if you intentionally max out the cooling speed? I did notice that even under this situation you are getting 94 C maximums, and that is no good.

Just brainstorming a bit, because that seems outright foreign after the amazing experience I had with my laptop.

Damn, I should have bought a Ryzen edition. I know Ryzen run cooler but I chose Intel becoz I heard good things about Tiger Lake and I wanted to use its Thunderbolt feature (eGPU some day). However eGPU didn't work on this laptop, and tiger lake CPU still reached 100w when being fully turbo boosted (whereas I saw Ryzen only reached 45w in this case). What's a disappointment. It's too late to return this laptop, I only noticed this power limit recently.
The thing is that when the power limit kicked in, MSI afterburner didn't detect the correct cpu clock anymore (it still showed max boost clock on my gaming screen), hence I didn't notice this before. Only opening AWCC or throttlestop I could see the correct clock which was throttled down. Be careful, when you said your 5600h ran full all day, it could be because you saw wrong numbers on MSI afterburner.

The MSR and MMIO power limits are set appropriately in ThrottleStop.

If you are going to log data, do not use the Stop Data feature. When Stop Data is enabled, your CPU is only being sampled once every minute. Attach a proper log file so I can have a look.

It looks like Dell is power limit throttling your computer and there is nothing you can do about it. That is why I have avoided this brand for the last 10+ years. When you pay extra money for an Alienware laptop, you should be able to run it however you like. Dell does not agree with that philosophy. You will need to shop elsewhere next time if you are interested in maximum performance.

Here is the log file after I used Start Data. The power limit kicked in after around 1hr of playing heavy game such as Cyberpunk. FYI: I set TCC on this laptop to 97c, hence there are a lot of thermal throttling when 97c was reached. This laptop's CPU ran very hot when GPU was fully loaded, it seems GPU's temp caused CPU's high temp as well due to them sharing the same heat pipe. When playing Cyberpunk, GPU reached 74c and CPU reached 97c max when fully boosted (4.2 Ghz).

One thing I noticed is that when forced power limit kicked in, MSI afterburner and HWinfo didn't detect the correct CPU clock anymore. Wondering could it be because of undervolting making CPU run incorrectly?
The other thing is that normally when my CPU run below 3.8Ghz, its voltage was oftens less than 1v (0.9 for e.g). But when power limit kicked in, its voltage was very high (1.05v - 1.1v when the clock was only 3.2 Ghz). Making total power consumption higher than expected. Is there something wrong here?

 

[Dr. Dro]

Damn, I should have bought a Ryzen edition. I know Ryzen run cooler but I chose Intel becoz I heard good things about Tiger Lake and I wanted to use its Thunderbolt feature (eGPU some day). However eGPU didn't work on this laptop, and tiger lake CPU still reached 100w when being fully turbo boosted (whereas I saw Ryzen only reached 45w in this case). What's a disappointment. It's too late to return this laptop, I only noticed this power limit recently.
The thing is that when the power limit kicked in, MSI afterburner didn't detect the correct cpu clock anymore (it still showed max boost clock on my gaming screen), hence I didn't notice this before. Only opening AWCC or throttlestop I could see the correct clock which was throttled down. Be careful, when you said your 5600h ran full all day, it could be because you saw wrong numbers on MSI afterburner.

All Ryzen processors offer a measurement called "Effective Clock", which shows on the spot if there is anything shady going on, for example, the behavior you are observing of the nominal clock reaching X, but the processor is actually operating at Y. They should pretty much match throughout all of the workload, which indicates the processor is operating as intended. If this measurement deviates too much from the Active Clock domain, then your processor is throttling back.

Even running multicore Cinebench R23, my CPU still runs around 4 GHz, which is ~700MHz above its nominal clock speed. BenchMate even caught the 4.25GHz from when the workload started and there was no heavy load going on yet

9230 points for what is effectively a 45-watt Ryzen 5 5600G is pretty good I would say, and sorry about the driver number being censored, I am not allowed to publish that

 

[kakashidinho]

All Ryzen processors offer a measurement called "Effective Clock", which shows on the spot if there is anything shady going on, for example, the behavior you are observing of the nominal clock reaching X, but the processor is actually operating at Y. They should pretty much match throughout all of the workload, which indicates the processor is operating as intended. If this measurement deviates too much from the Active Clock domain, then your processor is throttling back.

Even running multicore Cinebench R23, my CPU still runs around 4 GHz, which is ~700MHz above its nominal clock speed. BenchMate even caught the 4.25GHz from when the workload started and there was no heavy load going on yet

View attachment 241393


9230 points for what is effectively a 45-watt Ryzen 5 5600G is pretty good I would say, and sorry about the driver number being censored, I am not allowed to publish that

I can run cinebench fine, my CPU scored 12k plus points. The CPU wasn't throttled down running alone. This only happened when I played heavy games and both GPU & CPU are fully loaded. Not saying that your Ryzen would face the same issue, just FYI.

 

[Dr. Dro]

I see, in that case, the EC might be kicking in to prevent overdraw... I wonder if configured as is, your system is exceeding the AC adapter's maximum wattage or current capacity somehow. If that is the case, it is possible that ordering a larger AC adapter from a higher end laptop that may have the same plug (for example, a 330W brick) could fix your problem. But that is a potentially expensive gamble.

My laptop came with a 180W adapter, and no matter how far I push the RTX 3050 on it the absolute maximum power draw is about 65W (specced for 80W), CPU also having a maximum of 54W allowance... it seems to be enough for my model's needs and then some. I wonder how does the RTX 3060 version do here.

 

[unclewebb]

MSI afterburner didn't detect the correct cpu clock anymore

The Afterburner on screen MHz data displayed while gaming has never been 100% accurate. The Intel recommended monitoring method that ThrottleStop uses has worked great on all Core i CPUs produced since 2008. It is a more complicated monitoring method compared to the method used by some of the competition but it is extremely accurate. It can detect even slight changes in the CPU speed. The different results are obvious when a CPU starts changing its speed hundreds of times per second because it is either power limit or thermal throttling. The ThrottleStop log file data can be fully trusted.

Your log file shows lots and lots of thermal throttling. In Dell laptops with Intel processors, high average CPU temperatures seem to trigger the EC to reduce the internal turbo power limits. The thermal throttling then changes to power limit throttling with the PL2 power limit reduced to somewhere around 50W to 55W. This drops the CPU speed from over 4100 MHz down to about 3300 MHz. Inadequate cooling is costing you a lot of CPU performance.

I can run Cinebench fine

That is typical. It is the extra heat from the GPU that pushes your laptop over the thermal cliff. Bad monitoring data and short tests are why many review sites never report these issues. These problems surface when trying to play a game long term.

I have been a long time Intel supporter but for laptops, the cooler running Ryzen seems like the way to go.

To avoid triggering this power limit drop, you have to improve your cooling or use ThrottleStop to force your CPU to run at a reduced speed. A constant 3800 MHz will reduce maximum performance but it might help you avoid the CPU power limit throttling down to 3300 MHz.

 

[dnm_TX]

Are there any EC options in BIOS that one could thinker with? I'm asking anyone who would read that post and is familiar with(basically having unlocked BIOS and know their way around it)?

 

[unclewebb]

Are there any EC options in BIOS

There are typically no options in the BIOS that give you access to the EC power limit. Even an unlocked BIOS will not have this option available.

 

[dnm_TX]

There are typically no options in the BIOS that give you access to the EC power limit. Even an unlocked BIOS will not have this option available.

Thanks. Worth the try i guess. Would've been nice to get rid of one more of the so so many throttling layers.

 

[SS374]

In TS 9.5, try turning on the sync MMIO function on. I've had a similar, though not identical problem, which was caused by the MMIO PL1 value shifting to values nearing 37 W, causing an insane CPU bottleneck. Worth a shot.

 

[NoxanTG]

Hey!
I had experienced similar power throttling with Alienware M16 13900HX and used the sync and lock function to enforce 90W/90W power limits.
The CPU stopped throttling and W11 started to generate error reports of Intel DPTF device timing out. I believe that the device controls the dynamic power limits.
I am trying to understand the underlying reason for dynamic power limiting (decreases to 30W in my case). Could this be the VRM heating issue? Could lifting power limits damage the PCB? Is there a way to interpolate VRM temperature? I am a bit worried about this case.

 

[unclewebb]

In Windows 10, deleting the DPTF driver and preventing it form being reinstalled was a solution for DPTF problems. Do some Google searching to learn more.

 

[NoxanTG]

In Windows 10, deleting the DPTF driver and preventing it form being reinstalled was a solution for DPTF problems. Do some Google searching to learn more.

I have done that already, deleted the DPTF driver, and blocked it in regedit.
I would like to know the reason for automatic power limiting, though. Although my temps never go above 90 degC (Prochot 0 set) yet something is getting triggered and my laptop wants to decrease the fed power.
I think it might be high VRM temps?

 

[unclewebb]

I think it might be high VRM temps?

That should show up as VR TEMP in Limit Reasons. There could be a separate IR temperature sensor pointing at the backside of the keyboard or something like that which triggers the EC to drop the power limits. Does TS show power limit throttling?

 

[NoxanTG]

That should show up as VR TEMP in Limit Reasons. There could be a separate IR temperature sensor pointing at the backside of the keyboard or something like that which triggers the EC to drop the power limits. Does TS show power limit throttling?

After detailed testing, I have encountered no VR TEMP error in the limits section in TS 9.5.1. However, I have seen occasional EDP OTHER warning with yellow label and realized that this automatic PL decrease occurs right after the warning. I believe the energy consumption of the 13900HX is getting lowered after activation of the dynamic boost of RTX4080 and this results as EDP OTHER warning.
When benchmarking with CB23 or OCCT, there are no errors or warnings at all.

 

[unclewebb]

It is likely that CPU current consumption is reduced by the EC whenever the RTX dynamic boost feature is enabled. That is an easy way to reduce power consumption. Less power going to the CPU leaves more power available for the Nvidia GPU. This helps keep the power adapter within spec.

Dell has been doing stuff like this for years. A power adapter that is big enough and powerful enough to fully power both the CPU and GPU costs more to produce and it costs more to ship. Using some throttling to reduce performance a little helps keep costs down. If performance is horrible, you should return your laptop.

 

[SS374]

It is likely that CPU current consumption is reduced by the EC whenever the RTX dynamic boost feature is enabled. That is an easy way to reduce power consumption. Less power going to the CPU leaves more power available for the Nvidia GPU. This helps keep the power adapter within spec.

Dell has been doing stuff like this for years. A power adapter that is big enough and powerful enough to fully power both the CPU and GPU costs more to produce and it costs more to ship. Using some throttling to reduce performance a little helps keep costs down. If performance is horrible, you should return your laptop.

Yeah, I'd have to say it's something Dell enforces rather than a throttling limit; though I'm disappointed that Dell lets my i7 get limited to 30w~ when the GPU is at 115w; it creates a horrible CPU bottleneck (45w CPU/100w GPU would be so much better). Alienware's control of power profiles is extremely strict, so using the sync MMIO option is pretty much the only way to enforce user limits.