advice regarding optimization of faulty notebook cpu

[Gwaihir137]

hi,

I hope I'm posting to the right forum, as this seems to be a hardware related issue but I'm using ThrottleStop as a workaround! I have a notebook with a Intel Core i7-8750h and Nvidia RTX 2060 Mobile stock undervolted by -50mV, which has been behaving strangely lately. The fact that I'm not able to identify the cause is giving me headaches, After weeks of tinkering around, I personally think the cpu is faulty but I wanted to ask you guys about your opinion and the possibilities to get the best out of my notebook using ThrottleStop, which I have been using for testing and optimization, as there are a lot of guides out there. Especially for the 8750h, which tends to run hot.

Problems first occured when I played Valheim, were I experienced random CTDs after a short time period. I first thought it to be a temperature related problem, so I disassembled the notebook and reapplied the thermal paste, but the problems persisted. To keep it short

• When running cpu + gpu intensive applications or games, many of them will likely CTD after a few minutes. Most of the time the application just quits, sometimes I'm getting BSODs like WHEA_UNCORRECTABLE_ERROR or CLOCK_WATCHDOG_TIMEOUT (and others)
• My UEFI power management comes with 4 different profiles. When I choose "quiet" or "power saving" the CPU is limited to 2.2 GHz. Then there are no CTDs, notebook runs stable, but at a fraction of the performance.

What I did so far

• reinstall Windows, updating drivers etc etc
• running memtest, no errors
• some measurements using hwinfo64 and benchmarking tools. Temperatures are within the limit (cpu <90°C using prime95, gpu < 75°C using Furmark) and surprisingly zero crashes in whatever benchmark I used
• Removing the UEFI undervolt setting of -50mV, which seems to slightly decrease the amount of CTDs
• I'm mainly using "balanced" power mode (default) for day to day work, where the i8750h boosts up alle the way to 3.9 GHz. When I disable turbo boost in ThrottleStop or reduce the turbo ratios for all values to 22 (which effectively disables turbo boost), I'm experiencing the same crashes
• Extensive uv testing using ThrottleStop, which gives me a huge performance gain with reduced temps and power consumption. I mainly experimented with the FIVR voltage offset where I can go up to -200mV core voltage and -100mV Cache voltage. My settings run stable in all benchmarks, at day to day work and games which mainly utilize the CPU (like Minecraft) but the games which have had problems before will crash even more often
• Using "quiet" power mode and overriding the cpu settings with ThrottleStop leads to the same crashes, so I don't think the gpu is involved into the problem

Do you have any advice for me what feature of ThrottleStop I possibly could check out to reduce the amount of CTDs without having to choose "quiet mode"? Buying a new computer is not a option for me right now.

Sorry for the long post and thanks in advance!

Gwaihir

Here is some more technical information for you

Spoiler

After boot

FIVR

TPL

stress test using prime95

Overview of stable ThrottleStop UVs, each measurement was taken after 10 Minutes of Cinebench R23 Multicore

Profile​	Core 0 MHz avg​	Temp Core C Avg​	Temp Core C Idle​	Temp Core C Max​	VID V​	CPU Ppow W max​
Stock +0​	3160​	84​	54​	96​	1.298​	86,7​
UV -50 (core only)​	3316​	84​	48​	96​	1.234​	89,2​
UV -70​	3344​	83​	45​	95​	1.208​	89,6​
UV -100​	3452​	83​	43​	94​	1.170​	87,2​
UV -110​	3444​	84​	43​	95​	1.163​	85,4​
UV -120​	3486​	83​	43​	95​	1.156​	83,5​
UV -130​	3565​	84​	43​	94​	1.146​	82,1​
UV -140​	3525​	83​	43​	93​	1.136​	80,3​
UV -150​	3568​	83​	43​	93​	1.126​	77,9​
UV -170/-70 (core/cache)​	3556​	84​	43​	92​	1.127​	79​
UV -180/-80​	3563​	84​	43​	93​	1.112​	76,9​
UV -190/-90​	3604​	83​	43​	91​	1.107​	75,6​
UV -200/-100​	3660​	81​	43​	89​	1.092​	73,2​

more information about my undervolting experience on my blog (german): https://keepmydesktop.blog/2022/04/10/gaming-notebook-kuhler-schneller-leiser/

 

[unclewebb]

random CTDs

Is CTD "crash to desktop" while gaming or something like that?

Any crashes when undervolting are a good sign that your CPU is not getting enough voltage. It is not 100% stable. Try -100 mV for the cache and -150 mV for the core. If your games are still crashing then reduce your voltage settings some more.

Being full load benchmark stable is great but most crashes occur from lack of voltage when one of the cores is lightly loaded.

Different games are going to work different parts of the CPU and GPU differently. You might need different voltage settings depending on the game.

 

[Gwaihir137]

Thank you for your response(s)! I think the second answer was deleted since I last visited. I took some time for testing, just to be sure.

I'd like to clarify that these crashes to desktop or bsods started out of a sudden with the stock configuration of my laptop, which is -50mV on both core and cache. It also happens when I disable undervolting in UEFI entirely. After that I started undervolting and experimenting with ThrottleStop, because I tried to find a solution to this problem. That's why I think my hardware must probably be faulty.

Try -100 mV for the cache and -150 mV for the core

I did some testing with the settings you suggested, also with lower values, unfortunately it did not help on it's own (it lowered the temperature, which lead to longer boost times and thus increased the consumed power = more crashes?). I also cut my gpu speed to about 80% with MSI Afterburner, which saved me 10W with little to none fps loss. So far I have come to the conclusion, that only limiting the CPU speed to a certain level combined with undervolting fixes the issues with crashing applications. The stable threshold seems to be somewhere around 2.8 - 3.1 GHz.

So my stock clocks are 41, 41, 40, 40, 39, 39 (see screenshots in op). If I set these values to e.g. 30, 30, 30, 30, 30, 30, no more crashes. Have you come across similar issues before? Do you have any recommendations what else I could try to get my cpu stable without cutting the frequency?

 

[unclewebb]

@Gwaihir137
If your computer is crashing, I would set the offset voltage to 0 for the core and the cache. At default voltage, a computer should not be crashing. Your BIOS might have a -50 mV setting but that is not default voltage for an Intel CPU.

If you are still having stability problems with the offset voltage set to 0 then it could be a problem with the memory or GPU. You are going to have to do some more testing to try and isolate this problem.