CPU i7-12700H, RING EDP OTHER RED, DELL G15 5520, motherboard 05J7DY.

[mql]

Hi, I will post my post in this forum because the question is basically popular and no specific solution is known. RING EDP OTHER RED during load and episodic during idle time. UEFI/BIOS will be edited with ru.efi, so all recommendations from specialists will be considered prospective.

 

[unclewebb]

As I mentioned in the other forum you posted in, 12th Gen H series CPUs do not allow access to any of the FIVR settings. I do not know how to solve the EDP RING throttling problem. This is a common problem for many 12th Gen computers with locked CPUs. Only the unlocked HX and K CPUs allow access to the FIVR settings.

Likely the current limit (IccMax) for the cache / ring is set to some very low value. This can cause EDP throttling. The RING will run at the same base frequency as your processor runs at. That is like having turbo boost disabled but only for the RING. Turbo boost will still work OK for the CPU cores.

Can HWiNFO report what IccMax is set to for the cache / ring?

Try watching the Cache Ratio reported in the FIVR monitoring table.

Does the Cache Ratio immediately drop when running Cinebench?

 

[mql]

As I mentioned in the other forum you posted in, 12th Gen H series CPUs do not allow access to any of the FIVR settings. I do not know how to solve the EDP RING throttling problem. This is a common problem for many 12th Gen computers with locked CPUs. Only the unlocked HX and K CPUs allow access to the FIVR settings.

Likely the current limit (IccMax) for the cache / ring is set to some very low value. This can cause EDP throttling. The RING will run at the same base frequency as your processor runs at. That is like having turbo boost disabled but only for the RING. Turbo boost will still work OK for the CPU cores.

Can HWiNFO report what IccMax is set to for the cache / ring?

Try watching the Cache Ratio reported in the FIVR monitoring table.

View attachment 309232

Does the Cache Ratio immediately drop when running Cinebench?

Thanks for the answer.
So, about IccMax. If OverClocking Feature 0x0 (default) is not enabled in UEFI,

0x8DAE8         One Of: OverClocking Feature, VarStoreInfo (VarOffset/VarName): 0x1D9, VarStore: 0x3, QuestionId: 0x1574, Size: 1, Min: 0x0, Max 0x1, Step: 0x0 {05 91 69 07 6A 07 74 15 03 00 D9 01 14 10 00 01 00}
0x8DAF9             One Of Option: Disabled, Value (8 bit): 0x0 (default) {09 07 04 00 30 00 00}
0x8DB00             One Of Option: Enabled, Value (8 bit): 0x1 {09 07 03 00 00 00 01} (X)

then the value of ThrottleStop FIVR IccMax is 160A, but if it is enabled, then as you can see, IccMax is 168A. But changing the value of IccMax from 160A to 168A does not in any way change the behavior of EDP OTHER in the RING window.

I use Alienware-Command-Center-Application to maximize PC fan speed, it also turns on high power mode at the same time. Thus, the maximum CPU frequency is maintained at ~4.7 MHz when the PC is idle, but during the Cinebench test, it drops to ~4.1 MHz, which corresponds to the maximum turbo frequency of all P-core processors. As you can see in the picture, at this moment, the Cashe Ratio also decreases, respectively, from 40 to 36.

Can HWiNFO report what IccMax is set to for the cache / ring?

Maybe I didn't quite understand what you wanted to know, because in my case HWiNFO doesn't have a value like IccMax for Cache/Ring. I'm attaching a picture from HWiNFO just in case.

I wanted to add that I unpacked the bios files of several popular brands and studied them in comparison with DELL bios. If you tell me what to look for in the BIOS, I will check what values DELL and other manufacturers have set and give an answer.

update:

By way of experiments, adjusting the VccIN_AUX Supply DC Voltage and (VccGT) Supply DC Voltage manages to slightly increase the TS Bench result, but it does not affect EDP OTHER RED.

 

[unclewebb]

But changing the value of IccMax from 160A to 168A does not in any way change the behavior of EDP OTHER in the RING window.

The core IccMax and the cache/ring IccMax are two separate settings in the CPU. Intel's 12th Gen H series CPUs have blocked access to the FIVR settings. I have not heard of anyone finding a way to unlock these settings. I have zero experience modifying a BIOS so I cannot help you with this.

Here is what HWiNFO shows for my 10th Gen K series CPU.
IccMax for the ring can be both read from the CPU and written to the CPU.
A 12th Gen H series CPU does not allow either reading or writing IccMax information.

 

[A&P211]

what model laptop do you have?

 

[mql]

what model laptop do you have?

Dell G15 5520
Dell G15 5520 Setup and Specifications

My computer currently ranks sixth on the list CPU-Z validator Top 15 highest frequencies in the last 12 months with this CPU.

The core IccMax and the cache/ring IccMax are two separate settings in the CPU. Intel's 12th Gen H series CPUs have blocked access to the FIVR settings. I have not heard of anyone finding a way to unlock these settings. I have zero experience modifying a BIOS so I cannot help you with this.

Here is what HWiNFO shows for my 10th Gen K series CPU.
IccMax for the ring can be both read from the CPU and written to the CPU.
A 12th Gen H series CPU does not allow either reading or writing IccMax information.

View attachment 309361

Before going further, it is worth understanding what exactly this EDP OTHER throttling can affect in my case? If it is not difficult for you to provide a brief explanation or a reference to the source where this explanation is provided. Thank you.

 

[A&P211]

Run a r23 test

 

[mql]

Run a r23 test

Cinebench R23 installed from Microsoft store.
Outdoor temperature ~26-27°C
Indoor temperature ~28°C

Cinebench R23 tests with HWiNFO and other power monitoring programs running in the background. Multi Core result: 17380 - 17598.

Cinebench R23 test without power monitoring program. Multi Core result: 17971.

Single Core result: 1802.

updated
Retesting Cinebench R23 in a room with better air circulation. Although the ambient temperature has still risen to +28°, +29°C, open windows and air flow do their job. Multi Core result: 18204.

 

[unclewebb]

EDP OTHER throttling

This type of throttling slows down the ring bus inside the CPU. The on die cache memory chips will also work at this reduced speed. Bits and bytes moving around inside your CPU at a slower pace will reduce performance but for many tasks, you might not notice any difference. The reduced speed and power going to the cache chips can help reduce overall power consumption and it can improve stability. Intel's decision to deliberately throttle the ring is not all bad.

The FIVR is locked out in 12th Gen mobile H series CPUs so there is no way for ThrottleStop or any software to fix this throttling problem. Every 12th Gen H CPU will likely have the same EDP ring throttling problem. I am guessing that Intel set the IccMax current limit on the low side for the ring/cache. That is what causes EDP ring throttling.

Next time you are computer shopping, look for an HX processor. These mobile CPUs have unlocked FIVR settings as long as a manufacturer does not decide to lock out the FIVR in the BIOS.

 

[A&P211]

Cinebench R23 installed from Microsoft store.
Outdoor temperature ~26-27°C
Indoor temperature ~28°C

Cinebench R23 tests with HWiNFO and other power monitoring programs running in the background. Multi Core result: 17380 - 17598.
View attachment 309485

Cinebench R23 test without power monitoring program. Multi Core result: 17971.
View attachment 309486

Single Core result: 1802.
View attachment 309488

updated
Retesting Cinebench R23 in a room with better air circulation. Although the ambient temperature has still risen to +28°, +29°C, open windows and air flow do their job. Multi Core result: 18204.
View attachment 309506

I'm kinda surprised it got that high. I saw some reviews on the Dell G15. The video I saw showed the cpu on going upto 45w, which its the designed power output of the cpu. It still got lower benchmarks vs other laptop within the same specs in many of the game reviews, which means the cooling is not top notch.
I kinda like the overall design but its still a Dell.

 

[mql]

I'm kinda surprised it got that high. I saw some reviews on the Dell G15. The video I saw showed the cpu on going upto 45w, which its the designed power output of the cpu. It still got lower benchmarks vs other laptop within the same specs in many of the game reviews, which means the cooling is not top notch.
I kinda like the overall design but its still a Dell.

This particular computer has a BIOS change that I think combines the best BIOS settings from other manufacturers. However, the cooling system itself has not changed and corresponds to the original. As a drawback, I can note the much higher temperature of the third core of the Intel processor, which causes thermal throttling. In particular, the processor 17-12700h of this computer has a curved processor surface, which indicates that a layer of material is several tenths of a millimeter thicker over the location of the third core than over the remaining cores. But that's just my opinion.
I tried to remedy the situation by applying another thermal paste, but not all results were successful. Firstly, DELL had better thermal paste from the factory than the well-known and popular Arctic brand thermal pastes. It is possible that Thermal Grizzly Kryonaut is closer to the original in terms of performance. At the moment the computer has Thermal Grizzly Kryonaut Extreme installed, which does not solve the problem of overheating of the third core, but allows you to achieve previously published results.

At the same time, it should be recognized that the processor surface, which is bent by tenths of a millimeter, does not allow the cooling plate to be completely pressed against the processor surface. Which in turn ensures that the thermal paste between the CPU and the cooler is not completely squeezed out, allowing it to run longer without bleeding or disappearing.

This type of throttling slows down the ring bus inside the CPU. The on die cache memory chips will also work at this reduced speed. Bits and bytes moving around inside your CPU at a slower pace will reduce performance but for many tasks, you might not notice any difference. The reduced speed and power going to the cache chips can help reduce overall power consumption and it can improve stability. Intel's decision to deliberately throttle the ring is not all bad.

The FIVR is locked out in 12th Gen mobile H series CPUs so there is no way for ThrottleStop or any software to fix this throttling problem. Every 12th Gen H CPU will likely have the same EDP ring throttling problem. I am guessing that Intel set the IccMax current limit on the low side for the ring/cache. That is what causes EDP ring throttling.

Next time you are computer shopping, look for an HX processor. These mobile CPUs have unlocked FIVR settings as long as a manufacturer does not decide to lock out the FIVR in the BIOS.

Thanks for the explanation. This is enough for me to stop bothering with the issue of EDP OTHER throttling in RING. Because, in principle, nothing can be changed there, and it is not really known whether an increase in RING IccMax will increase processor performance. As I mentioned in the previous answer, the biggest problem is the thermal throttling of one core, which, in my opinion, is due to the design features of the CPU, because when using different thermal pastes, the third core is always throttled first.

At the same time, however, it should be noted that the thermal throttling of the third core still allows about 90W of continuous power consumption during the Cinebench R23 test. Therefore, we can say that the CPU has no defects, because there is no thermal throttling at 45W or 65W specified in the specification.

 

[unclewebb]

it is not really known whether an increase in RING IccMax will increase processor performance

An increase in the ring speed can increase performance for any benchmark or task that heavily uses the cache. Being able to run the cache at turbo speed instead of the base frequency of a processor does improve performance. Constant EDP ring throttling reduces performance but with an H series CPU, there is nothing you can do about it. For most real world tasks, the loss of performance is not that much.

Edit - I did some 10850K Cinebench testing. With all cores at a steady 5000 MHz and the cache at 4700 MHz, I scored 16937. When running the cache at the base frequency, 3600 MHz, my score dropped almost 500 points to 16439. A slow ring/cache definitely reduces performance.

 

[mql]

An increase in the ring speed can increase performance for any benchmark or task that heavily uses the cache. Being able to run the cache at turbo speed instead of the base frequency of a processor does improve performance. Constant EDP ring throttling reduces performance but with an H series CPU, there is nothing you can do about it. For most real world tasks, the loss of performance is not that much.

Edit - I did some 10850K Cinebench testing. With all cores at a steady 5000 MHz and the cache at 4700 MHz, I scored 16937. When running the cache at the base frequency, 3600 MHz, my score dropped almost 500 points to 16439. A slow ring/cache definitely reduces performance.

After receiving your answer, I was looking for information about the possibility of increasing the readings of CPU L1, L2, L3 Cache Read, Write, Copy by bypassing the blocked FIVR settings. At the moment, this is a double-edged sword, if you increase VccIN_AUX Supply DC Voltage, AC Loadline to the nominal values,

the CPU performance decreases, but the L2 and L3 Cache values increase in the AIDA64 test. Which in general does not increase the result of the Cinebench test, but rather reduces it.

I found the AIDA64 Cashe test for an XMG NEO 15 E22 laptop (i7-12700H/RTX 3080 Ti)

and am currently looking for information in the bios of this computer about how it differs from the DELL bios, because in my case the AIDA64 test shows much more modest results in L2 Cashe.


updated
After a long time of comparing the BIOS of different computer models, the only BIOS from which I can get useful information was and remains only the MSI laptops.

 

[WLF]

This type of throttling slows down the ring bus inside the CPU. The on die cache memory chips will also work at this reduced speed. Bits and bytes moving around inside your CPU at a slower pace will reduce performance but for many tasks, you might not notice any difference. The reduced speed and power going to the cache chips can help reduce overall power consumption and it can improve stability. Intel's decision to deliberately throttle the ring is not all bad.

The FIVR is locked out in 12th Gen mobile H series CPUs so there is no way for ThrottleStop or any software to fix this throttling problem. Every 12th Gen H CPU will likely have the same EDP ring throttling problem. I am guessing that Intel set the IccMax current limit on the low side for the ring/cache. That is what causes EDP ring throttling.

Next time you are computer shopping, look for an HX processor. These mobile CPUs have unlocked FIVR settings as long as a manufacturer does not decide to lock out the FIVR in the BIOS.

@unclewebb If we found a way to unlock FIVR, do you know what would fix EDP throttling? Someone on Github recently showed me how to modify the BIOS to essentially get the MSI Advanced BIOS with 1000s of settings on my TUF F15, so I have access to turbo ratio limits and other stuff now in the BIOS (can't modify them in Throttlestop but still should be similar).

 

[unclewebb]

Increasing IccMax to the max can fix a lot of EDP throttling problems. Not sure if that is possible with your laptop.

Be careful what you change. One wrong change in an unlocked BIOS can create an expensive brick.

 

[mql]

@unclewebb If we found a way to unlock FIVR, do you know what would fix EDP throttling? Someone on Github recently showed me how to modify the BIOS to essentially get the MSI Advanced BIOS with 1000s of settings on my TUF F15, so I have access to turbo ratio limits and other stuff now in the BIOS (can't modify them in Throttlestop but still should be similar).

Everyone can use their own methods for modifying uefi / bios, however, in my experience, I recommend using methods that allow you to first save the default settings, which you can later load as a template and return the edited uefi section to the default values. . In my case, this is ru.uefi which allows you to save a file with values and then load and save changes if necessary.

Increasing IccMax to the max can fix a lot of EDP throttling problems. Not sure if that is possible with your laptop.

Be careful what you change. One wrong change in an unlocked BIOS can create an expensive brick.

I fully support your warning. At one time, after a careless change in the RAM settings in the BIOS, I received a brick that only beeped and signaled that the RAM was invalid. At that time, the ancient method of disconnecting the battery and starting the PC only from the power cord helped. Not sure if it will work now.

So far, it has not been possible to find parameters in bios of various PC manufacturers that would remove RING EDP OTHER for 12700H and at the same time preserve or improve CPU performance. But strange as it is, turning on (Embedded Controller) EC Turbo Control Mode slightly boosted CPU performance. I consider this a strange behavior, because the information I find almost always indicates that the Embedded Controller limits more than frees the CPU.

 

[oinkypig]

disable smart app control for EDP errors. windows search-smart app control. Only applies in this case if throttlestop is running - under med-heavy workload, and the cpu frequency drops. Dont understand How no1 mentioned this already.

 

[WLF]

Smart app control is off already, although I don't see why a function that blocks untrusted apps would throttle your CPU power for all apps only when you're specifically at 99% battery or less.

Smart app control is off already, although I don't see why a function that blocks untrusted apps would throttle your CPU power for all apps only when you're specifically at 99% battery or less. I'm going to look for settings like @unclewebb recommended in my BIOS this weekend, although I will be careful about what I change.

 

[oinkypig]

Well, Something that has worked for me is after you have disabled VM and core isolation, and smart app control is to delete the throttlestop INI file and start throttlestop. FIVR registered after that for me at least.

 

[WLF]

Well, Something that has worked for me is after you have disabled VM and core isolation, and smart app control is to delete the throttlestop INI file and start throttlestop. FIVR registered after that for me at least.

Not sure what you mean by that, but Intel 12th gen laptop CPUs are locked down so that FIVR settings are only accessible from the BIOS (and some, like undervolting, are impossible). I have tried disabling VM and Core Isolation and it didn't change anything. This is a completely different issue than you think it is, it has to do with ASUS's BIOS using the embedded controller to throttle the laptop CPU to 25 W when the laptop is at less than 100% battery and the discrete GPU is active.

 

[PowerThrottle]

Generally, this issue is not usually caused by a low IccMax value because manufacturers don't usually tweak the IccMax value too much. If it has been adjusted, being able to increase it in some way might solve the problem. However, if it hasn't been altered, the issue will likely persist. I personally don't believe the problem is related to IccMax. As I have mentioned in my own guides, if there isn't an electrical issue, you might have a problem with SVID Voltage Limit (VR Voltage Limit) or something like DRAM RAPL. It's likely you won't be able to address these without advanced BIOS options and access. I wish you the best.

 

[mql]

Generally, this issue is not usually caused by a low IccMax value because manufacturers don't usually tweak the IccMax value too much. If it has been adjusted, being able to increase it in some way might solve the problem. However, if it hasn't been altered, the issue will likely persist. I personally don't believe the problem is related to IccMax. As I have mentioned in my own guides, if there isn't an electrical issue, you might have a problem with SVID Voltage Limit (VR Voltage Limit) or something like DRAM RAPL. It's likely you won't be able to address these without advanced BIOS options and access. I wish you the best.

I have UEFI level access to edit all the parameters listed there. I have attached the file. If you can indicate the parameters to edit, I will check.

 

[PowerThrottle]

I have UEFI level access to edit all the parameters listed there. I have attached the file. If you can indicate the parameters to edit, I will check.

Could you check with the HWINFO tool? What does the throttle type show as? It might be better for me to provide recommendations based on that.

 

[mql]

Could you check with the HWINFO tool? What does the throttle type show as? It might be better for me to provide recommendations based on that.

At the beginning of this Conversation, the issue of IccMax has already been discussed. A link to the test video is attached test.mp4 | Failiem.lv. As you can see at the end of the video EDP OTHER disappears. Maybe someone knows why it happens.

updated
I did another test, only this time for 12 minutes. The result is the same, EDP OTHER disappears after a couple of minutes of maximum CPU load. Link https://failiem.lv/u/fwe49sgfu?ak=e3f01

I will mention that the computer was connected to the TV all the time, including during the tests, and continually uses the NVidia RTX 3060 laptop video card.

Increasing IccMax to the max can fix a lot of EDP throttling problems. Not sure if that is possible with your laptop.

Please watch this video about CPU testing. I would like to ask your opinion why the EDP OTHER is constantly red at the beginning of the test, and after a few minutes it splits or disappears altogether for a while. Link test_2
youtu.be/test 1
youtu.be/test 2

 

[career]

Thanks to TechPowerUp for allowing me to register! I would like to express gratitude and warmest regards to mr unclewebb and all the TechPowerUp for software they made. It's just outstanding and helps a lot!

I have got a notebook with i7-12700H and the same issue with EDP OTHER RING throttling that appears under load. I tried everything i could to change this behavior of the notebook but the solution was really hard to come by. Recently i accidentially found a solution, can't kind of say it's "the very right one" but it works fine and you may find it below.

Other solutions have yet to be found. Possibly...

PREMISE PART.

My model is Asus Lingyao Pro 16 2022 (originally it's Vivobook Pro 16X but this one is limited version for China having some differences)
Processor is i7-12700H. Full specs here https://store.asus.com.cn/store/product-1494134727.html.

Surely i carefully looked over all reviews i could find before i decided to make an order.
The reviews posted at chinese resources were well made, and seemed to be true, again i did a lot of search before i bought this item.
I terms of trottling, reviews' authors reported decent results and benchmark numbers looked good.

Reviews can be found here:

华硕灵耀Pro16 2022测评：RTX3070Ti显卡加持的高端全能创作本

全能创作本是近年来笔记本的热门选项。它既满足了游戏需求、设计需求，又满足了生产力办公需求，还拥有比常规游戏本更轻薄的机身和更强的续航能力。各厂商纷纷涉足这一领域，其中华硕可谓是最出类拔萃的。特别是在…

zhuanlan.zhihu.com

轻薄本天花板？华硕灵耀Pro16 2022使用体验

随着用户需求的多样化以及技术的进步，笔记本电脑已经实现轻薄便携与高性能的两极分化。 近些年更是在先进技术的加持下孕育出了高性能轻薄本，彻底解决了高性能创意工作者的创作生产需求。在高性能轻薄本领域，华…

zhuanlan.zhihu.com

如何评价华硕推出的全新灵耀 Pro16 2022 款？ - 知乎

10月21日，全新的华硕灵耀Pro16 2022正式发售，磨具全新升级，12代I7标压处理+3060的显卡的配置，在同价…

www.zhihu.com

灵耀 Pro16 2022 评测：顶级 OLED 好屏幕，真正 RTX 生产力_接口_设计_支持

为了适应办公和生产力群体的需求，灵耀 Pro16 2022 的接口也是相当丰富，左侧是一个 DC 电源接口、一个 2.5G 有线网口、一个 USB-A 5Gbps 接口，以及一个 MicroSD E…

www.sohu.com

Unfortunately my item has overheating and throttling issues that i noticed right after when i started to test it.
So far i didn't manage to find out for sure if it is just thermal related issue or some kind of production defect.

MAIN PART.

You may see all my setting in ThrottleStop attached. The main idea is to get maximum performance.

Uefi bios it locked to make any changes like that.
Things like VM, core isolation, and smart app control are disabled.
I can't change IccMax. Current value is 160 i suppose.
FIVR is locked.

Below i'm going to refer to what's discussed here, because as it seems we have a common issues.

As a drawback, I can note the much higher temperature of the third core of the Intel processor, which causes thermal throttling. In particular, the processor 17-12700h of this computer has a curved processor surface, which indicates that a layer of material is several tenths of a millimeter thicker over the location of the third core than over the remaining cores. But that's just my opinion.

Confirmed. I have also noticed higher temperature of the 2nd and 4th p-cores. Sometimes under load i do not see a good temperature distribution among cores, especially because of indicated two p-cores, which is the reason of throttling. As for thermal and cooling part, i havn't touched this thing yet.

Does the Cache Ratio immediately drop when running Cinebench?

Confirmed. It drops immediately.
My numbers in Cinebench R23 are ~14500/1800.

the issue of EDP OTHER throttling in RING.

This thing bothered me all the time. Whenever it blinks red, i noticed throttling immediately.

I found the AIDA64 Cashe test for an XMG NEO 15 E22 laptop (i7-12700H/RTX 3080 Ti)

Agreed. I was going to buy it, but its chassis design and display were the reasons why i changed my mind. I watched a hour interview on YouTube when they presented and tested new XMG NEO, it's really awesome thing. It has additional water cooling system but even without it, only using air cooling, it passes benchmark noticeably better than others. Last time i checked it always took 1th or 2nd place in comparison tables of benchmarks charts.

In my opinion, a scale of how much throttling may affect to performance drop can be like this:

1. BD PROCHET - causes huge performance degradation, it's visually noticeable in any task you do, the most worst thing you may get ever
2. EDP OTHER in RING - causes drop in cpu and cache performance and probably much more, it's visually noticeable in any task you do, a worst thing too
3. THERMAL in CORE - causes not much drop in performance, sometimes it can be visually noticed but rare
4. THERMAL in RING - causes not much drop in performance, sometimes it can be visually noticed but rare
5. EDP OTHER in CORE - causes not much drop in performance, sometimes it can be visually noticed but rare
6. EDP OTHER in GPU - to be honest i didn't visually notice any performance drop in ordinary tasks

FINAL PART.

This way EDP OTHER in RING causes a noticeable drop in performance.

The first method to get rid of this is to set lower max setting for cache ratio in ThrottleStop, that's it.

Default values were 4/40
Current values are 4/36

In my case, 36 is the maximum value when i get EDP in RING disappeared. If i set it higher, it appears again.

Now i'm using this as remedy. The point is this method is not slow down the notebook, as i thought it would do in the first place.

There is another way to forget about EDP OTHER in RING. We don't touch cache setting, instead we change only one setting in TPL:
Miscellaneous -> check Speed Shift ->set 4 for Min (or any other value) -> then set 52 for Max
In my case if i set higher value then 52, EDP in RING appears again.
This method brings to lower cpu frequency and probably downgrade overall performance, that's why i decided to use first method.

Finally, having all of these settings applied, the only following throttling icons are displayed sometimes:
CORE THERMAL, RING THERMAL
CORE EDP OTHER, GPU EDP OTHER

What else, well, another good option for performance in ThrottleStop to tune is PROCHET.
My default value was 92. I increased it to 97 via offset value (see screenshot attached).
If i use higher value like 98, i will see BD PROCHET under stress test that ruins overall performance of the notebook very much.
So in my case the maximum safe value is 97.

PS. Cinebench R23 performed with cache ratio 4/36. If i switch back to default ratio 4/40, numbers become sufficiently lower.

CONCLUSION PART.

Currently i was able to fix some performance issues thanks to ThrottleStop and other software. I literally have no words to express my respect and gratitude to mr unclewebb for what he is done. Thanks!

A guessed issue with the abnormal curvature of the cpu, if it turns out to be true, can be fixed too. I know how to do it, it's possible, i trust.
Next i consider to switch to liquid metal for cpu and gpu that will require some preparation for metal surface of thermal plate.
If it happens i'll post my updated benchmark results.

Hope it helps to someone!