Problem with Trottlestop and HWINFO64 on MSI Raider 18

[unclewebb]

Tell me, should the voltage reduction for cpu core and cpu p-cache be the same?

I have not tested this yet. I get great results when the CPU Core and P Cache offsets are set to the same value, -150 mV. Do some Cinebench testing to find out what works best. Watch for best performance and watch for lowest temperatures.

On older 8th and 9th Gen CPUs it used to be best to set the core offset to a bigger number compared to the cache. Doing that might not make any difference on a 14th Gen HX CPU. I like to use Cinebench R23 when testing.

Cinebench R23

Cinebench is the perfect tool to compare CPU and graphics performance across various systems and platforms

www.techspot.com

 

[KomAndr007]

I have not tested this yet. I get great results when the CPU Core and P Cache offsets are set to the same value, -150 mV. Do some Cinebench testing to find out what works best. Watch for best performance and watch for lowest temperatures.

On older 8th and 9th Gen CPUs it used to be best to set the core offset to a bigger number compared to the cache. Doing that might not make any difference on a 14th Gen HX CPU. I like to use Cinebench R23 when testing.

Cinebench R23

Cinebench is the perfect tool to compare CPU and graphics performance across various systems and platforms

www.techspot.com

Hi. I noticed that your processor microcode is not up-to-date. You did not install the 12B version? And also, what temperature will you have on your MSI with such parameters if you run a RAM test in Aida64? Can you make a screenshot of such a test?

P. S. With a turbo boost multiplier of 52, I set the undervoltage to -0.158 mV for the processor and the ring bus. It seems to work stably.

 

[unclewebb]

I noticed that your processor microcode is not up-to-date.

I am getting good results with the old microcode version. I am not using the 58 multiplier and I am undervolting. The voltage going to my CPU should be safe. I might update later when I have more time to do some testing. I did use microcode 0x12B for a little while. Performance seemed about the same.

if you run a RAM test in Aida64

I do not use Aida64.

For a long term stress test, done inside at a normal room temperature of 23°C, I would have to lower the turbo power limits significantly to prevent the CPU from reaching 100°C and thermal throttling. The settings I have been using so far are just so I can see a big number when Cinebench R23 testing. I am up over 39K. The P and E cores are overclocked +100 MHz. With a little more overclocking and some cool outside air, I just might make it to 40K.

I have been doing a little more ThrottleStop programming lately. The new Per Core feature is almost done. Users with a hot running core or two will be able to adjust the speed of individual cores. That might help prevent thermal throttling when the rest of the cores still have some temperature headroom left.

 

[KomAndr007]

I don't quite understand how this setting will help? For example, my 5th core heats up more and goes into throttling earlier than the others.

 

[JaimeT]

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.

View attachment 390745

Hi @unclewebb , I have a newbie question: If you have phochot to 1, does your max peak temp goes up more the 100? I'm asking because I have it set to 3 and my peaks temp reaches 100.

Thanks

 

[unclewebb]

my 5th core heats up

The ability to slow down your one hot running core should allow all of your other cores to run at full speed a little longer before thermal throttling begins.

phochot to 1

PROCHOT Offset controls the temperature when thermal throttling begins. Under full load, Intel CPUs can be changing their core temperature so rapidly that a couple of degrees of overshoot beyond the thermal throttling temperature is normal. Thermal shutdown does not happen until approximately 125C so there is nothing to worry about. The CPU is well engineered to look after itself.

 

[JaimeT]

The ability to slow down your one hot running core should allow all of your other cores to run at full speed a little longer before thermal throttling begins.


PROCHOT Offset controls the temperature when thermal throttling begins. Under full load, Intel CPUs can be changing their core temperature so rapidly that a couple of degrees of overshoot beyond the thermal throttling temperature is normal. Thermal shutdown does not happen until approximately 125C so there is nothing to worry about. The CPU is well engineered to look after itself.

Thanks, I'll try setting it to 1 and test.

 

[unclewebb]

@KomAndr007 - Boot Performance Mode controls whether Intel Turbo Boost is enabled or not when your computer is booting up. When turbo boost is enabled, a computer runs faster and should boot up faster. Often times, it does not make much difference. Turbo boost can create a lot of heat which some people do not like.

 

[KomAndr007]

@unclewebb , Thank you. This screenshot was also shown to show that those two cores, where the larger multiplier is shown, were the ones that were heating up more for me.

 

[unclewebb]

the larger multiplier

The two cores that are fused to the 58 multiplier are the two favored cores. When a CPU is only partially loaded, Windows should schedule tasks more often on the two favored cores compared to on any of the other available cores. It makes sense that these two cores would tend to run hotter compared to the rest of the cores.

What Is an AMD CCX? A Basic Definition

What is an AMD CPU’s CCX? AMD core complexes explained.

www.tomshardware.com

Even when a CPU is fully loaded, there is always going to be some temperature variation from core to core. Intel temperature sensor accuracy is far from perfect.

 

[StarRacer]

This screenshot was also shown to show that those two cores, where the larger multiplier is shown, were the ones that were heating up more for me.

The ability to slow down your one hot running core

It is not necessary to limit the performance of the core that is heating up (if we are talking about the best core marked by intel, that what comes with the biggest ratio). As I already mentioned, you can simply set a higher Ratio for the cooler cores, and then Windows will consider them the best cores. For example, run a CPU stress test and find the two cores with the lowest temperature, then set their Ratio to 58 and set the Ratio for the other cores to 56.

Although I’m not sure if the other cores can reach 5.8 GHz, I think they might. If not, you'll have to set the Ratio to 56 for the cooler cores and 55 for the rest. But in any case, this will not affect multi-threaded performance at all. Even single-threaded performance won’t be affected much since reaching 5.8 GHz is almost impossible due to temperature throttling.



@unclewebb Today, I reinstalled Windows, and naturally, the first thing I installed was TS and MSI Center. The first thing that caught my eye was that MSI Center no longer changed the MMIO PL1 and PL2 values they remained the same as the MSR values. At first, I was happy and went ahead with testing, but to my surprise, I was limited to 55W TDP in Balanced mode and 75W TDP in Performance mode (such a limit when nvidia gpu works at full).
Previously, I could reach 140W TDP even in Discrete Graphics mode, and exceed 140W in MSHybrid mode when the NVIDIA GPU wasn’t consuming much power. Now only the Integrated Graphics mode allowed me to go beyond 75W. At first, I reset the BIOS, thinking I had messed something up, but that wasn’t the case...

Long story short, it turned out that this was controlled not by MSI Center but by the Intel IPF Driver and Intel Dynamic Tuning Driver. After installing them from MSI’s website on my laptop’s support page, I was once again able to exceed 75W TDP in Discrete Graphics and MSHybrid modes.


I even came across this page: https://www.intel.com/content/www/us/en/support/articles/000090464/graphics.html


It seems that for some manufacturers, disabling Dynamic Tuning removes the limit, which is probably why this page exists on Intel’s website. but I didn't have it turned off, I just had the standard drivers from Microsoft. I might try this before my next windows reinstall: https://www.reddit.com/r/Dell/s/xKoSs4JbcW

 

[KomAndr007]

Hello. Can you tell me if it is possible to set different voltage offset values for CPU Core and CPU P Cache?

+
Is it better to leave IccMax for CPU Core as is or to set more? So that the performance is maintained and it doesn't heat up too much.

 

[unclewebb]

it is possible to set different voltage offset values for CPU Core and CPU P Cache?

These two voltages are usually coming from the exact same voltage rail. You can use ThrottleStop to request two different values but the CPU will likely decide what value to use and will use the same offset value for both voltages. I use the same -150 mV request values for both the Core and the P Cache on my 14900HX. If you get better results setting these two offset voltages differently then go ahead and do it.

I set IccMax to the max, 511.75, for the CPU Core, P Cache, Intel GPU and iGPU Unslice. Setting all of these values to the max is the best way to avoid any EDP throttling.

it doesn't heat up too much

If your computer gets too hot, lower the turbo power limits. This is the best way to control how much heat a CPU puts out.

 

[KomAndr007]

Hi! Could it be that the voltage reduction is simply ignored? I don't understand. I set the multiplier to 52 maximum for all cores, I set the voltage offset for the processor and bus to 165 mV (there are no crashes in games, tests in Aida pass without a blue screen). But the temperature in games still rises above 80 degrees, and sometimes, like yesterday, jumped to 91, and of course, thermal throttling of the processor immediately. At the same time, during games, MSI does not give more than 75 W to the processor. I don't understand. At the same time, the laptop cooling tries and makes a loud noise (I set the maximum to 95%, and it is very noisy). Do I really have to go to a multiplier of 51 or even 50 for a comfortable game? I am starting to be very disappointed in this laptop with such a processor. Maybe I configured something wrong.

 

[unclewebb]

It looks like you have installed the latest BIOS update and are running microcode version 0x12F. Perhaps that fix is part of the problem. I am still using an old BIOS version with microcode 0x116. I know BIOS updates are supposed to make things better but after testing a few of these updates from Intel, I concluded that newer microcode versions are not always better. The default voltage curve is being adjusted by Intel. I have not tested 0x12F so I do not know if undervolting still works the same.

MSI does not give more than 75 W to the processor

Some laptops restrict the CPU power whenever the Nvidia GPU is active. They usually program an embedded controller to set this lower power limit. ThrottleStop has no access to the EC power limit.

Use ThrottleStop 9.7.3 and turn on the Log File option when playing a game or doing some testing. The newer version logs the P and E cores separately.

very noisy

Modern laptops with HX processors are kind of a bad joke. If you try to run them at their full rated speed, they will sound like an airplane getting ready for take off. You are forced to significantly reduce performance if you want to slow the CPU fans down to a reasonable level.

 

[KomAndr007]

In my BIOS, I initially have the microcode version 11F. Is it newer than 116 or later? Yesterday evening I rolled back to the native 11F. And I got a blue screen in the game, that is, -165 mV on this microcode the processor no longer holds stably, in addition, there was thermal throttling - the temperature at the peak was 91 degrees. I made the multiplier 51 at -165 mV for now. In fact, if you do not chase high numbers in tests, then for games, probably 50 will be enough. I wonder if it affects FPS much. But I realized that in order to stay at 52, additional cooling is needed and it will be noisy, otherwise throttling cannot be avoided. A wonderful processor. But as I understand it, the new Ultra line is no better.

 

[unclewebb]

Microcode numbers are in hexadecimal format. Each digit will count up first from 0 to 9 and then it will continue to count up from A to F. Look up hexadecimal on Google to learn more.

0x116
0x117
0x118
0x119
0x11A
0x11B
0x11C
0x11D
0x11E
0x11F
0x120

Microcode 0x11F is quite a few versions after microcode 0x116.

Whenever the microcode version changes, the ability to undervolt can also change. An undervolt that works great with one microcode version might constantly crash with a different microcode version. Intel is always very secretive about what sort of changes each version includes. After all of the 13th and 14th Gen stability issues, Intel has been tweaking the voltage curve to try and fix these issues. What is good for the mases that do not undervolt may not be any good for those of us that do like to undervolt. An undervolt that is stable at the default maximum multiplier may need to be adjusted if you decide to lower the maximum turbo multiplier.

With microcode 0x116, I can run a stable undervolt of -150 mV on the P core and P cache. I have had no crashes in quite a while. My laptop is mostly for testing purposes and light duty use. It has lived a fairly sedate life except when I first got it. I used to like running the hell out of it when trying to see what it could do. I plan to do some more testing next winter.