Stable during benchmark but not in game

[imwolo]

Hello everyone,

So I will describe my problem as much as possible. I have an i5 9300hf processor and I followed the unclewebb advice on how to set up i5 9300h (basically the same CPU). I increase PP0 Current limit to 256 and I maxed out ICCMAX on both core and the cache and now I'm playing around with undervolting (I'm using throttlestop 9.2). So before I did this I had PL2 Yellow on Core and EDP OTHER on the ring. They are both gone now after changes to PP0 and ICCMAX. So now to my undervolt I can do -210 mV on core and - 105 mV on the cache and it's perfectly stable during Cinebench R23 and during the inbuild benchmark tool in Throttlestop (Tested 6144M on 8 threads). Now when I start up the game after 2-3 minutes my system crashed (no BSOD) and had to restart. I also tried -85 mV on the cache and - 170 mV on core and it still crashed but when I tried -103 mV on both core and cache it was completely stable. I don't understand both benchmark tools I mentioned they used 100% of my CPU while the game was fluctuating from 30 to 80%. Did I do something wrong? Shouldn't it be stable while I'm gaming with the first values I mentioned?

I will be thankful for any advice.

 

[unclewebb]

Hello

An undervolt that is stable at full load might not be stable when the load is fluctuating. Laptops are more likely to crash when lightly loaded but most users still only do full load testing when checking stability.

Did you do some 1, 2 or 4 thread TS Bench tests? Did you try using the Random MHz option in this test? Did you run a simple benchmark like Heaven to put some load on your GPU while running any TS Bench tests or other stability tests?

When you are playing a game, the GPU will increase power consumption and will add extra heat into your computer. You need to run a variety of tests to prove that an undervolt is stable.

 

[imwolo]

Thanks for the quick answer! I will get right to the testing. I just have a few more questions for you if you don't mind @unclewebb . I found that if set both core and cache to -105mV it will crash but at -103mV it is stable so far. So I should be undervolting only core while I leave cache at -103 mV? And I wanted to ask about PP0 and ICCMAX. If I leave PP0 at 256 and ICCMAX maxed out on both core/cache will it have some negative impact in the short/long run or am I safe leaving them? Thanks again.

 

[unclewebb]

-105mV it will crash but at -103mV it is stable

Do not set the voltage to the bare minimum to be stable. Set the cache to -80 mV so you know your CPU is 100% stable. Now you can try adjusting only the core offset to a bigger number. Do some testing with Cinebench R20.

It is OK to max out the PP0 Current Limit and IccMax values.

 

[hipster_lenin]

Id suffer stutters in game if my core was undervolted too much, crashes were always largely related to unstable GPU overclocks. take that for what its worth if you have your GPU undervolted / ocd

When i first started undervolting the CPU I found that i could go to -155 on cache/cpu and no higher. dropping cache to -125 afforded me another -50mV on the CPU, but it'd be stable in benchmarks up to -100mV (delta) cache

 

[imwolo]

Hello again @unclewebb , so I took your advice, and with a little bit of tweaking, I was able to get -90mV on the cache and -180mV on the core. I used Prime 95 (Small FTT) and TS bench with Heaven on Extreme and it was stable and no errors in ts bench either. I did get 2 errors when I tried -100/-200. So that's means that I found limits of my CPU? No room for any more improvements? And also it is normal that during prime95 with no GPU test in the background my CPU peaked at 96 C? Thanks again.

 

[unclewebb]

So that's means that I found limits of my CPU?

Yes, you have found the limits. A CPU should never report any errors. You do not want 2 + 2 = 4 one day and for 2 + 2 = 5 the next day. That is the testing that the TS Bench does. It performs the same calculation multiple times on each core to make sure that your CPU gets the same answer each time.

No room for any more improvements?

Have a look in the Options window. Your computer has set PROCHOT Offset to 6. This controls when the CPU will start to thermal throttle. The Intel default is 0 which allows the CPU to start throttling at 100°C. A 6° offset tells your laptop to start thermal throttling at 94°C (100°C - 6) instead of the full 100°C. If you do not see a lock icon near this setting, you can try lowering this value. Many laptops use a value of 2 or 3.

Some laptops set the thermal shut down temperature right at 100°C. It is supposed to be 125°C but some laptops cut corners. If your laptop does this then it would be best to set PROCHOT Offset to at least 3 so you do not have to worry about a premature shut down.