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

[mql]

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.


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.


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


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


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!

Are you sure that the methods you recommend can improve CPU performance?

Benchmark V: 17.01.64

Benchmark V:19.01.64 (beta)

Benchmark V:19.01.64 AVX2 (beta)

The CPU-Z benchmark reaches ~102 W

ThrottleStop benchmark reaches ~115 W

 

[career]

Are you sure that the methods you recommend can improve CPU performance?

No, i'm not sure about that. I said that it just helps maximize my own perfornamce, in my case. It's just my configuration that well tested by me.
My notebook's performance is limited due to issues with temperatures of p-cores #2 and #4 causing throttling under load.

Same 12700H but your notebook has more TDP and your numbers should be higher.
What would you suggest to maximize performance?

 

[mql]

No, i'm not sure about that. I said that it just helps maximize my own perfornamce, in my case. It's just my configuration that well tested by me.
My notebook's performance is limited due to issues with temperatures of p-cores #2 and #4 causing throttling under load.

Same 12700H but your notebook has more TDP and your numbers should be higher.
What would you suggest to maximize performance?

First of all, cooling, I don’t know if other laptops have the ability to take in air from above and below, but the DELL G series has such an opportunity.

Secondly, suitable thermal paste, my experience shows that not all thermal pastes advertised on forums have sufficient characteristics to dissipate more than 90W of heat on the processor die.
Thirdly, editing the BIOS, but it requires knowledge, because what I changed is not available in the standard BIOS menu.

 

[career]

First of all, cooling, I don’t know if other laptops have the ability to take in air from above and below, but the DELL G series has such an opportunity.
Secondly, suitable thermal paste, my experience shows that not all thermal pastes advertised on forums have sufficient characteristics to dissipate more than 90W of heat on the processor die.
Thirdly, editing the BIOS, but it requires knowledge, because what I changed is not available in the standard BIOS menu.

Thanks!
Think in my case bios mod will be too tricky. Thermal replacement and cooling enhancements are the only what's left to try.

 

[mql]

Sorry to come back with a question regarding the ThrottleStop CPU stress test.
Could ThrottleStop, at an average power ~90W, load CPU cores unevenly during a stress test?
As you can see from the CPU's test time of around 10 minutes, it is the P-core 4 that is the most thermally throttling, although it is not the fastest CPU core.

Update

I ask because a month ago I reapplied the same thermal paste, only this time in a very thin layer, not exceeding the height that formed between the copper plate of the cooler and the black coating.

The example is not from my computer, but it helps you understand what I mean.

After replacing the thermal paste, the CPU test results did not show intense P-core 4 thermal throttling.

I don't understand what has changed, is this a sign of thermal paste degradation or is it due to software changes such as Windows 11 and driver updates.

 

[unclewebb]

Could ThrottleStop, at an average power ~90W, load CPU cores unevenly during a stress test?

I would say the exact opposite is true. The TS Bench should be putting a very even load on each core. It runs the exact same algorithm on each thread of each core. When the TS Bench is set to 20 Threads, it starts up 20 threads with each thread doing the exact same thing.

although it is not the fastest CPU core.

I am not sure what this means. The ThrottleStop FID data shows that all P cores in your screenshot are being throttled to the same speed. I have heard that some Intel CPUs can throttle cores individually but that is not the case for your CPU. In your case, any P core can trigger thermal throttling at any time. When that happens, all P cores throttle and run at the same speed. They are all tied together. It appears the P cores do not have the ability to throttle individually.

If you click on the Mod heading at the top of the ThrottleStop monitoring table, this will show the ID number of each core. If you reposition the table you should be able to see some of the P cores and some of the E cores at the same time. When thermal throttling of a P core is in progress, it would be interesting to see if the E cores get throttled too. The E cores should not be throttling but anything is possible.

thermal paste degradation

I think this is a very common problem with many popular thermal pastes when applied direct die to mobile CPUs. I have not done any thermal paste testing on mobile CPUs. One user told me he tried several popular thermal pastes and ultimately, long term, they all deteriorated.

Genuine Honeywell PTM 7950 is not like traditional thermal pastes. It has a better reputation of lasting longer when applied to mobile CPUs. Give that a try if you have not already tried it.

 

[mql]

I would say the exact opposite is true. The TS Bench should be putting a very even load on each core. It runs the exact same algorithm on each thread of each core. When the TS Bench is set to 20 Threads, it starts up 20 threads with each thread doing the exact same thing.


I am not sure what this means. The ThrottleStop FID data shows that all P cores in your screenshot are being throttled to the same speed. I have heard that some Intel CPUs can throttle cores individually but that is not the case for your CPU. In your case, any P core can trigger thermal throttling at any time. When that happens, all P cores throttle and run at the same speed. They are all tied together. It appears the P cores do not have the ability to throttle individually.

If you click on the Mod heading at the top of the ThrottleStop monitoring table, this will show the ID number of each core. If you reposition the table you should be able to see some of the P cores and some of the E cores at the same time. When thermal throttling of a P core is in progress, it would be interesting to see if the E cores get throttled too. The E cores should not be throttling but anything is possible.


I think this is a very common problem with many popular thermal pastes when applied direct die to mobile CPUs. I have not done any thermal paste testing on mobile CPUs. One user told me he tried several popular thermal pastes and ultimately, long term, they all deteriorated.

Genuine Honeywell PTM 7950 is not like traditional thermal pastes. It has a better reputation of lasting longer when applied to mobile CPUs. Give that a try if you have not already tried it.

Thanks for the explanation, it's definitely the thermal paste's fault. Regarding the fastest core, it was meant that it is not the core to which the maximum multiplier 47 is applied and, accordingly, it is not the core to which the highest voltage is applied on average.

 

[mql]

Unfortunately, I am unable to purchase a genuine Honeywell PTM 7950 within a reasonable period of time.
Therefore, I will tell those interested about the current solution.
Although this picture is not from my computer, it shows what the thermal paste looked like in my case.
In the middle part of the processor, on the sides, the thermal paste seems to be blown out to the sides with bubbles.

To prevent this from happening, I cut strips approximately 1 mm wide from the Gelid GP-Extreme thermal pad, which I glued to the sides of the processor core. I specifically ran the test for several days under extreme CPU load. After the test, I opened the computer to see what happened to the thermal paste. The thermal paste was not squeezed out on the sides of the processor, part of it flowed to one of the ends of the CPU, the total amount of paste located on the CPU core was at least twice as much as before.

Perhaps this solution will be useful for someone else.

 

[unclewebb]

Thermal paste pump out when used on mobile CPUs is real. The pics you posted are a perfect example. After a while there will be next to no thermal paste left covering the CPU cores.

 

[career]

When dealing with notebooks there are two mainstream issues with thermal interface you may run into:

• "pump out": makes the thermal paste layer thinner overall but can still maintain consistent contact pressure, especially if screws are being fastened tighter after several burn-in cycles.
• "bump up": causes migration and changes in material properties that are localized to a specific point of the die. Can not be equalized by fastening screws tighter.

Best detailed explanation ever i think https://www.reddit.com/r/XMG_gg/comments/15kuywx

 

[mql]

A month has passed since the thermal paste was applied to the processor chip using circuit boards that hold the thermal paste.

In the middle part of the processor, on the sides, the thermal paste seems to be blown out to the sides with bubbles.

To prevent this from happening, I cut strips approximately 1 mm wide from the Gelid GP-Extreme thermal pad, which I glued to the sides of the processor core.

At the moment this is the best solution I have. TS test result with 7680M.

120M / 5.937 = 20.21
7680M / 373.105 = 20.58

 

[career]

At the moment this is the best solution I have. TS test result with 7680M.

If you have MUX Switch switched to use discrete graphics only, you may stop showing annoying "EDP OTHER" at GPU by turning off Intel Xe in Device Manager, just click disable there. You may also set Speed Shift EPP to 0.

 

[unclewebb]

@mql
If your trick has made your thermal paste last for a month then you are doing something right. I think thermal paste issues on laptops is a bigger issue than most people realize.

Thanks for posting a screenshot of ThrottleStop's extended view monitoring table. I rarely see that view in screenshots so I assume that most people do not know about double clicking on the monitoring table to access that view. I made it just big enough so I could see all 20 threads of my 10850K. Great to see that it also works correctly on a 20 thread 12700H.