• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Problem with Trottlestop and HWINFO64 on MSI Raider 18

KomAndr007

New Member
Joined
Mar 8, 2025
Messages
5 (0.36/day)
Hello everyone. I have an MSI Raider laptop with 14900HX. I started undervolting and changing frequencies through this program and noticed this problem. When I change something in the program, I then run HWInfo64 to monitor temperatures, frequencies, voltage and other indicators. And when I run a test in Aida64, or during a game, the laptop turns off. This happens only after changes using this program. And only with HWInfo64 running in the background. If you do not change anything and run Hwinfo64 in the background, then the shutdown does not occur. Maybe someone knows what the problem is and how to solve it?
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
Maybe someone knows what the problem is
Post some ThrottleStop screenshots so I can see what settings you are using. Include the main window, the FIVR and TPL windows. Are you using the V/F Point feature? Setting V/F Point 1 to 150 for both the core and the cache can help improve stability.

If you are using ThrottleStop to undervolt then perhaps your undervolt settings are not 100% stable. If you have increased the turbo power or current limits, perhaps your power adapter is not happy with that.

I use ThrottleStop everyday on my MSI Vector laptop. It also has a 14900HX. Performance and stability are fantastic. I think your settings just need a little tweaking.

 

KomAndr007

New Member
Joined
Mar 8, 2025
Messages
5 (0.36/day)
I'll take screenshots of the settings tomorrow, I don't have the opportunity now. I'd like to add that overall it seems to work stably, it only shuts down if HWInfo64 is running in the background. Most likely there is some problem with it. But I don't understand why there are no crashes if you don't change anything in the settings.

P. S. There are such photos on the phone. I didn't change anything in the TPL menu, since MSI Center will still give a maximum of 175 W on the GPU and 75 W on the CPU in the game. Stability in games and the absence of throttling are important to me. But it is problematic to track it when you can't launch hwinfo64.
That's why I settled on a multiplier of 50, since in higher games the temperature goes up to 90 or more.
 

Attachments

  • IMG_0690.jpeg
    IMG_0690.jpeg
    2.2 MB · Views: 46
  • IMG_0691.jpeg
    IMG_0691.jpeg
    1 MB · Views: 45
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
I am guessing that your -170 mV undervolt for the core and the P cache is too much. My 14900HX starts losing stability if I go beyond -150 mV. Try reducing your undervolt. Start conservatively. Maybe -125 mV and work your way towards -150 mV. Try to find a voltage where you can run AIDA64 without it crashing.

I can sometimes run Cinebench at -160 mV but I think this undervolt is too much for day to day use. Cinebench will sometimes crash at -160 mV so the CPU is not 100% stable at this level.

Consider setting Power Limit 4 in the TPL window to the max, 1023. I would also set IccMax to the max, 511.75, for the CPU Core, P Cache as well as the Intel GPU and iGPU Unslice. When living on the edge this can help keep things stable. I also use the Windows High Performance power plan to try and keep the CPU at a steady MHz.
 

KomAndr007

New Member
Joined
Mar 8, 2025
Messages
5 (0.36/day)
But your multiplier is 52, and mine is 50. Accordingly, more voltage is needed for stability. With a multiplier of 52, if I remember correctly, I also could not take a decrease higher than 150. Or am I wrong, and the multiplier does not play a role in the ardervolting?
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
The maximum multiplier does play a part in how much voltage is needed to be stable. The problem is, your computer is not stable. That is why I suggested adding some voltage. If AIDA64 is stable at -125 mV or -100 mV then you know that your problem is not enough voltage.
 

StarRacer

New Member
Joined
Dec 31, 2024
Messages
4 (0.05/day)
I use ThrottleStop everyday on my MSI Vector laptop. It also has a 14900HX. Performance and stability are fantastic. I think your settings just need a little tweaking.
Have you made any modifications to your MSI laptop besides software tweaks? Your temperatures are surprisingly low for such performance levels, both at idle and under load!

I replaced the stock phase-change thermal pad with a PTM7950 I bought, which helped lower temperatures by 5°C. However, I kept the stock thermal putty. Hopefully, that doesn't affect performance too much.

I limited the Turbo Boost Ratio to 50 and undervolted the CPU and cache by -140 mV. On average, stress test temperatures reach 85°C, but if I set IccMax to the maximum, it immediately hits 95°C. Default IccMax 216.

The funny thing is that P-core P3 almost instantly reaches 95°C, while the others stay around 88-90°C. Even under a static all-core load (where voltage and frequency are the same in OCCT stress tests), P03 runs 10°C hotter than the rest. Without increasing IccMax, and Ratio 50 limit, all cores stay at 75°C, but P03 is at 85°C, if not for that, the laptop would be quieter.

Do you also have one or more cores that are significantly hotter than the rest in this CPU? You can check under a static load in AIDA, Intel XTU stress tests, or OCCT.
I even cleaned the copper plate and die again, reapplying PTM7950 to make sure I didn’t mess up the application...


Additionally, cores P02 and P03 are considered the "best" ones, they can reach 5.8 GHz and are marked with a star in Intel XTU. Unfortunately, P03 throttles very quickly. This is bad because Windows always prioritizes these best cores for single-threaded tasks. Even in multi-threaded workloads (not static loads like stress tests, but in games, for example), these two cores do most of the work before the rest kick in.

I found a way to "trick" Windows. The system doesn't primarily look at Intel's "best core" designation, it checks the Fused Max Core Ratio first.

1000030107.jpg


(1 Active core I use 50, like everywhere else. I just played with TS and TS reset the values and I changed it manually and didn't notice.)

Unfortunately, for some reason, BIOS doesn’t allow changing these values, only viewing them. But in Intel XTU, they can be modified, and the changes persist in BIOS.

So, I set the highest ratio (52) for my coldest cores, P00 and P02, while lowering P03 below the rest. Most cores got a ratio of 51, and P03 was set to 50. This didn’t affect performance since my Turbo Ratio is already capped at 50, even for single-threaded tasks.


Here’s an old screenshot with previous values, but the principle remains the same.
1000030104.jpg


1000030103.jpg


If you check core usage in HWinfo, you'll see that P03 is used less frequently than the others. Now, P00 and P02 are the primary cores for single-threaded tasks, even in games. The system normally prioritizes the best cores based on Fused Max Core Ratio, so adjusting these ratios on the coldest cores helps prevent throttling and performance drops, especially in 1–4 thread workloads. This also lowers overall package temperatures, making the fans quieter, as they no longer need to compensate for a single overheating core.

The only issue is ThrottleStop for some reason, it sets all E-cores to 85 by default when launched, even on a cold boot without a config file. Sometimes, it affects P-cores too, but I haven’t figured out exactly when.

1000030109.jpg


And if I change any value in Turbo Ratio, TS overwrites the Fused Max Core Ratio for all cores to the value set in Group 0 (the 1-core ratio). However, in Intel XTU, modifying any value except the first one doesn’t overwrite the Fused Max Core Ratio.

Is there any way to fix this or add functionality similar to Intel XTU, allowing per-core ratio adjustments?
 
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
Your temperatures are surprisingly low
Cinebench testing outside at 8°C is the trick I use. That is why the core temps are so low in the screenshot I posted.

Do you also have one or more cores that are significantly hotter than the rest in this CPU?
With the P cores numbered from 0 to 7, my screenshot shows that during Cinebench, core 0 peaked at 79°C and core 3 peaked at 90°C. The temperature sensors that Intel uses are not 100% accurate. Their only purpose is to trigger thermal throttling. They were never designed or intended to be used for 100% accurate core temperature reporting from idle to full load. The amount of error is somewhere in the +/- 5°C range. You could have two cores that are both at the exact same 90°C. The temperature sensor on one core might report a little low. It might show 85°C while the core beside it reads a little high and reports 95°C. This is a normal amount of variation even if two cores are physically at the exact same temperature.

My CPU is right on the edge of normal variation. Replacing the thermal paste might improve things but it also might not make any noticeable difference. Most users put too much faith in the temperature data being generated by Intel CPUs when Intel themselves admit that their temperature sensors are far from perfect. Some PTM 7950 is on my things to do list but I do not think it is going to make a significant difference to the core to core temperature variation that I am seeing.

if I set IccMax to the maximum, it immediately hits 95°C
If increasing IccMax causes your CPU temperatures to go up 10°C, setting IccMax to 216 must be causing your CPU to EDP throttle when running stress tests. I prefer setting IccMax to the max. I would rather adjust the turbo power limits or the turbo ratios to control the maximum CPU temperature.

ThrottleStop for some reason, it sets all E-cores to 85 by default
I think starting with ThrottleStop 9.7.2 (maybe 9.7.3), the Overclock box in the FIVR window now controls the per core values. If Overlock is not checked, ThrottleStop should not make any changes to the per core turbo ratios. The default turbo ratios for my 14900HX are 58 for the two preferred cores and 56 for the other six P cores. All of the E cores default to the 41 multiplier.

When Overclock is checked, ThrottleStop will set all of the per core values to the max, 85, so they do not interfere with maximum performance. Realistically, the default 58 and 56 maximum multipliers for the P cores are high enough so checking the Overclock box is not necessary.

per-core ratio adjustments?
I have thought about adding per core ratio adjustments for the P cores to ThrottleStop. I know it will happen someday. Just not sure when.

Thanks for the feedback.
 
Joined
May 6, 2023
Messages
137 (0.20/day)
System Name Precision 7540
Processor i9-9980HK
Memory 4x16GB HyperX Impact 2666MHz CL15
Video Card(s) RTX 4000 Max-Q (90W)
Storage 3x1TB XPG SX8200 Pro Gen3x4
Power Supply 240W
I use ThrottleStop everyday on my MSI Vector laptop. It also has a 14900HX. Performance and stability are fantastic. I think your settings just need a little tweaking.

Unclewebb, with this 230W consumption can your i9 reach and sustain its maximum clocks (P-cores = 5.2GHz and E-cores = 3.9GHz) throughout the Cinebench R23 test?
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
reach and sustain its maximum clocks
YES!! Lucky Cinebench R23 is a short test or something might go ka-boom. Running any laptop at 230W is kind of pushing things near the limit or perhaps a little beyond.

The default max multi is 52 when all 8 P cores are active and the default max for the E cores is the 41 multiplier.

I updated the log file recently so it now keeps track of the P cores and E cores separately. This will make it easier to see if both are maintaining full speed. Maybe tomorrow I will do another run to prove that both P & E cores are running at full speed for the entire test. Here is a log file when idle just to show this feature.

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU
2025-03-15  16:02:35    52.00    48     0.4    41.00    46     0.2   1.1627     2.2        0       0     0
2025-03-15  16:02:36    52.00    47     0.9    41.00    46     0.2   1.1622     2.7        0       0     0
2025-03-15  16:02:37    52.00    44     0.7    41.00    46     0.1   1.1633     2.5        0       0     0
2025-03-15  16:02:38    52.00    43     0.3    41.00    46     0.1   1.1639     2.1        0       0     0
2025-03-15  16:02:39    52.00    47     0.2    41.00    45     0.1   1.1776     2.0        0       0     0
2025-03-15  16:02:40    52.00    45     0.8    41.00    46     0.1   1.1749     2.6        0       0     0
2025-03-15  16:02:41    52.00    44     0.7    41.00    45     0.1   1.1647     2.8        0       0     0
2025-03-15  16:02:42    52.00    43     1.1    41.00    45     0.2   1.1626     3.0        0       0     0
2025-03-15  16:02:44    52.00    46     1.0    41.00    46     0.2   1.1880     2.7        0       0     0
2025-03-15  16:02:45    52.00    45     1.1    41.00    46     0.2   1.2007     3.1        0       0     0

I also added Per Core control of the 8 Performance cores. This might be a useful feature for some laptops if one hot running core is holding back the rest.

1742432266564.png


Edit - It was good benchmarking weather outside tonight. Cinebench R23 at full speed at the default 52 and 41 multipliers was no problem. It was so easy and there was still some temperature headroom so I bumped the P cores up to 53X. Another full speed run was my reward. It was +1 for the P cores so next up, I did the same for the E cores. 53 and 42 got me into the Cinebench 39K club. Still zero throttling. Power consumption was getting kind of crazy so I decided to call it quits for tonight. I still need this laptop in one piece so I can continue working on ThrottleStop. :D

I will attach the unedited log file and a screenshot of Cinebench. It looks like the E cores still have lots of thermal headroom. Good to know!

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU 
2025-03-19  19:49:44    53.00    94   100.0    42.00    74   100.0   1.2793   253.0        0       0     0
 

Attachments

  • 2025-03-19.txt
    51.4 KB · Views: 5
  • R23 - 39314.png
    R23 - 39314.png
    279 KB · Views: 10
Last edited:
Joined
May 6, 2023
Messages
137 (0.20/day)
System Name Precision 7540
Processor i9-9980HK
Memory 4x16GB HyperX Impact 2666MHz CL15
Video Card(s) RTX 4000 Max-Q (90W)
Storage 3x1TB XPG SX8200 Pro Gen3x4
Power Supply 240W
YES!! Lucky Cinebench R23 is a short test or something might go ka-boom. Running any laptop at 230W is kind of pushing things near the limit or perhaps a little beyond.

The default max multi is 52 when all 8 P cores are active and the default max for the E cores is the 41 multiplier.

I updated the log file recently so it now keeps track of the P cores and E cores separately. This will make it easier to see if both are maintaining full speed. Maybe tomorrow I will do another run to prove that both P & E cores are running at full speed for the entire test. Here is a log file when idle just to show this feature.

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU
2025-03-15  16:02:35    52.00    48     0.4    41.00    46     0.2   1.1627     2.2        0       0     0
2025-03-15  16:02:36    52.00    47     0.9    41.00    46     0.2   1.1622     2.7        0       0     0
2025-03-15  16:02:37    52.00    44     0.7    41.00    46     0.1   1.1633     2.5        0       0     0
2025-03-15  16:02:38    52.00    43     0.3    41.00    46     0.1   1.1639     2.1        0       0     0
2025-03-15  16:02:39    52.00    47     0.2    41.00    45     0.1   1.1776     2.0        0       0     0
2025-03-15  16:02:40    52.00    45     0.8    41.00    46     0.1   1.1749     2.6        0       0     0
2025-03-15  16:02:41    52.00    44     0.7    41.00    45     0.1   1.1647     2.8        0       0     0
2025-03-15  16:02:42    52.00    43     1.1    41.00    45     0.2   1.1626     3.0        0       0     0
2025-03-15  16:02:44    52.00    46     1.0    41.00    46     0.2   1.1880     2.7        0       0     0
2025-03-15  16:02:45    52.00    45     1.1    41.00    46     0.2   1.2007     3.1        0       0     0

I also added Per Core control of the 8 Performance cores. This might be a useful feature for some laptops if one hot running core is holding back the rest.

View attachment 390651

Edit - It was good benchmarking weather outside tonight. Cinebench R23 at full speed at the default 52 and 41 multipliers was no problem. It was so easy and there was still some temperature headroom so I bumped the P cores up to 53X. Another full speed run was my reward. It was +1 for the P cores so next up, I did the same for the E cores. 53 and 42 got me into the Cinebench 39K club. Still zero throttling. Power consumption was getting kind of crazy so I decided to call it quits for tonight. I still need this laptop in one piece so I can continue working on ThrottleStop. :D

I will attach the unedited log file and a screenshot of Cinebench. It looks like the E cores still have lots of thermal headroom. Good to know!

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU
2025-03-19  19:49:44    53.00    94   100.0    42.00    74   100.0   1.2793   253.0        0       0     0
I thought that on the 14900HX the E-cores would only reach 4.1GHz at lighter loads (below 50%) and 3.9GHz above 50% or full load.

Fantastic result!

To run the CBR23 at these clocks without undervolting you should need something close to 300W...
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
I thought that on the 14900HX the E-cores would only reach 4.1GHz at lighter loads
I have not seen any Intel documentation that confirms what speed the E cores are supposed to run at. The main Intel website shows that the 14900HX Efficient cores Max Turbo Frequency is 4.1 GHz. There is no further explanation.


I fixed the Defaults button in the FIVR window. It now sets all of the E cores to the 41 multiplier no matter how many of them are active. Reading info from the CPU confirms that 41 is the default max E core turbo multiplier. Some manufacturers might decide to program both Turbo Group 0 and Turbo Group 1. They could program Group 0 to the 41 multiplier when up to 8 cores are active and then program Group 1 to the 39 multiplier when all 16 cores are active. The E cores run cool. Slowing them down does not seem necessary.

1742448981231.png


Fantastic result!
I totally agree! It is fun to just be a ThrottleStop user for a change. I like seeing the results that are possible without having to make too many changes. This CPU has been very stable even when pushing the temperature and power to the limits.

I once told someone on TPU that it was going to take about 260W to get max performance out of my 14900HX. I think they thought that I was exaggerating. Today I reached 253W! I am likely nearing the limits of the power adapter or there might be some electrical component on the motherboard that was never designed for this kind of abuse. I was originally going to buy a Legion but I do not think they can be pushed this hard without triggering some type of throttling. I think the Legion voltage regulators will start to trigger VR Current throttling when trying to explore the outer limits.
 
Last edited:
Joined
May 6, 2023
Messages
137 (0.20/day)
System Name Precision 7540
Processor i9-9980HK
Memory 4x16GB HyperX Impact 2666MHz CL15
Video Card(s) RTX 4000 Max-Q (90W)
Storage 3x1TB XPG SX8200 Pro Gen3x4
Power Supply 240W
I have not seen any Intel documentation that confirms what speed the E cores are supposed to run at. The main Intel website shows that the 14900HX Efficient cores Max Turbo Frequency is 4.1 GHz. There is no further explanation.


I fixed the Defaults button in the FIVR window. It now sets all of the E cores to the 41 multiplier no matter how many of them are active. Reading info from the CPU confirms that 41 is the default max E core turbo multiplier. Some manufacturers might decide to program both Turbo Group 0 and Turbo Group 1. They could program Group 0 to the 41 multiplier when up to 8 cores are active and then program Group 1 to the 39 multiplier when all 16 cores are active. The E cores run cool. Slowing them down does not seem necessary.

View attachment 390676


I totally agree! It is fun to just be a ThrottleStop user for a change. I like seeing the results that are possible without having to make too many changes. This CPU has been very stable even when pushing the temperature and power to the limits.

I once told someone on TPU that it was going to take about 260W to get max performance out of my 14900HX. I think they thought that I was exaggerating. Today I reached 253W! I am likely nearing the limits of the power adapter or there might be some electrical component on the motherboard that was never designed for this kind of abuse. I was originally going to buy a Legion but I do not think they can be pushed this hard without triggering some type of throttling. I think the Legion voltage regulators will start to trigger VR Current throttling when trying to explore the outer limits.
I believe you can break 40k by simply increasing just the E-cores as there is still plenty of thermal headroom for the efficiency cores.

Maybe keeping the P-cores at 5.2GHz and boosting the E-cores to 4.5GHz or 4,6GHz (or higher if temperatures allow).
 

KomAndr007

New Member
Joined
Mar 8, 2025
Messages
5 (0.36/day)
YES!! Lucky Cinebench R23 is a short test or something might go ka-boom. Running any laptop at 230W is kind of pushing things near the limit or perhaps a little beyond.

The default max multi is 52 when all 8 P cores are active and the default max for the E cores is the 41 multiplier.

I updated the log file recently so it now keeps track of the P cores and E cores separately. This will make it easier to see if both are maintaining full speed. Maybe tomorrow I will do another run to prove that both P & E cores are running at full speed for the entire test. Here is a log file when idle just to show this feature.

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU
2025-03-15  16:02:35    52.00    48     0.4    41.00    46     0.2   1.1627     2.2        0       0     0
2025-03-15  16:02:36    52.00    47     0.9    41.00    46     0.2   1.1622     2.7        0       0     0
2025-03-15  16:02:37    52.00    44     0.7    41.00    46     0.1   1.1633     2.5        0       0     0
2025-03-15  16:02:38    52.00    43     0.3    41.00    46     0.1   1.1639     2.1        0       0     0
2025-03-15  16:02:39    52.00    47     0.2    41.00    45     0.1   1.1776     2.0        0       0     0
2025-03-15  16:02:40    52.00    45     0.8    41.00    46     0.1   1.1749     2.6        0       0     0
2025-03-15  16:02:41    52.00    44     0.7    41.00    45     0.1   1.1647     2.8        0       0     0
2025-03-15  16:02:42    52.00    43     1.1    41.00    45     0.2   1.1626     3.0        0       0     0
2025-03-15  16:02:44    52.00    46     1.0    41.00    46     0.2   1.1880     2.7        0       0     0
2025-03-15  16:02:45    52.00    45     1.1    41.00    46     0.2   1.2007     3.1        0       0     0

I also added Per Core control of the 8 Performance cores. This might be a useful feature for some laptops if one hot running core is holding back the rest.

View attachment 390651

Edit - It was good benchmarking weather outside tonight. Cinebench R23 at full speed at the default 52 and 41 multipliers was no problem. It was so easy and there was still some temperature headroom so I bumped the P cores up to 53X. Another full speed run was my reward. It was +1 for the P cores so next up, I did the same for the E cores. 53 and 42 got me into the Cinebench 39K club. Still zero throttling. Power consumption was getting kind of crazy so I decided to call it quits for tonight. I still need this laptop in one piece so I can continue working on ThrottleStop. :D

I will attach the unedited log file and a screenshot of Cinebench. It looks like the E cores still have lots of thermal headroom. Good to know!

Code:
   Date       Time     P_Cores   °C    C0%    E_Cores   °C    C0%     VID     Power   Bat_mW   NVIDIA  GPU
2025-03-19  19:49:44    53.00    94   100.0    42.00    74   100.0   1.2793   253.0        0       0     0
Hello. I see that your maximum temperature is 95 degrees. Mine seems to start throttling at 91. Is 95 a normal temperature?
 

KomAndr007

New Member
Joined
Mar 8, 2025
Messages
5 (0.36/day)
I just noticed that in hwinfo64 the "thermal throttling" indicator becomes Yes and red already somewhere around 91 degrees, if I remember correctly.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,228 (1.34/day)
somewhere around 91 degrees
Setting the PROCHOT Offset variable allows individual manufacturers to reduce the thermal throttling temperature. When PROCHOT Offset is set to 9, the CPU will begin thermal throttling at 91°C instead of the Intel recommended 100°C. I have set PROCHOT Offset to 1 so thermal throttling begins at 99°C.

If a manufacturer locks this setting, you will see a yellow lock icon on the left hand side of PROCHOT Offset in ThrottleStop. Many Acer laptops have locked PROCHOT Offset. You cannot use ThrottleStop to make any changes to any control register if it has been locked.

1742500476827.png
 
Top