Help undervolting i7-9750h

[not_a_real_name]

Hi, can you guys help me with throttlestop settings? i have a legion y540 laptop and i'm not that expert in undervolting.
i set both (CPU core and CACHE) offset voltage to 110mv and i'm not enabling (speed shift-EPP) because if i enable it and set the value to 0 the cpu frequency is always above 3.8 GHZ.
i also chose OK-save voltages immediately.
i didn't try it to see the results yet but are there any thing wrong with these settings and explain please the EPP thing.
thanks in advance.

 

[duskw4lker]

Hi, can you guys help me with throttlestop settings? i have a legion y540 laptop and i'm not that expert in undervolting.
i set both (CPU core and CACHE) offset voltage to 110mv and i'm not enabling (speed shift-EPP) because if i enable it and set the value to 0 the cpu frequency is always above 3.8 GHZ.
i also chose OK-save voltages immediately.
i didn't try it to see the results yet but are there any thing wrong with these settings and explain please the EPP thing.
thanks in advance.

hello!

The Speedshift EPP is perfectly fine with a value of 0, that is the same as using the windows power slider in "best performance". your CPU (and Windows as a result) will be more responsive because it will be clocking higher than with a higher value in speedshift (eg 128. 255 is the slowest thing ever, don't) . Those clocks are the expected behavior and not to be worried about.

Regarding the voltage for both stuff: how lower you can go depends on a lot of stuff (mobo, bios settings, power supply, blah blah) so it's up to you to test stability. there is no universal value to get "the best" result.

but try something like this:

1) baseline: Run Cinebench r20 in multicore test

• while having the TS window open and running the test, take notes of CPU package temp, CPU package power, limits reasons and the score obtained
• you should check your limits reasons in TS because:
  • you may need to edit the Long and Short power limits - PL1 and PL2, and turbo time limit in the TPL window
  • you may need to edit pp0 current limit also in the TPL window

2) if you are running an older version of TS than 9, run TS manually, don't set it to autostart with task scheduler. As you gain more confidence, later you can set it to boot up with Windows after you get your best results. Before that you will crash at least once testing voltages.

• If you crash, after the mandatory reboot, recovery, absolutely unnecessary panic and sweating (been there), delete the .ini config file in TS folder so you can start all over again. Or if you can, reset the values in FIVR before it crashes again while being idle. if your UV was unstable you may crash while idling, so the faster way to start again is to delete the .ini before opening TS.

3) set CPU Core to -0.050mv and Cache to -0.025mv as a starting or conservative UV and run Cinebench r20 multicore test.

• Take notes of CPU package temp, CPU package power, limits reasons and the score obtained

4) go lower by 0,005 in each value and keep testing and taking notes

• the idea is that you get higher clocks, higher scores, lower temps, lower power consumption and no limit reasons.
• If you get it right, this CPU can reach 40x multiplier with a BCLK of 99.768 mhz = 3990.7 mhz under full load.
• Don't go crazy on the Cache as it will make you crash more often than the Core, go with a 2:1 ratio (2cpu:1cache)

I'm by no means an expert in TS, but those steps are the base of this whole undervolting magic thanks to @unclewebb

 

[not_a_real_name]

hello!

The Speedshift EPP is perfectly fine with a value of 0, that is the same as using the windows power slider in "best performance". your CPU (and Windows as a result) will be more responsive because it will be clocking higher than with a higher value in speedshift (eg 128. 255 is the slowest thing ever, don't) . Those clocks are the expected behavior and not to be worried about.

Regarding the voltage for both stuff: how lower you can go depends on a lot of stuff (mobo, bios settings, power supply, blah blah) so it's up to you to test stability. there is no universal value to get "the best" result.

but try something like this:

1) baseline: Run Cinebench r20 in multicore test

• while having the TS window open and running the test, take notes of CPU package temp, CPU package power, limits reasons and the score obtained
• you should check your limits reasons in TS because:
  • you may need to edit the Long and Short power limits - PL1 and PL2, and turbo time limit in the TPL window
  • you may need to edit pp0 current limit also in the TPL window

2) if you are running an older version of TS than 9, run TS manually, don't set it to autostart with task scheduler. As you gain more confidence, later you can set it to boot up with Windows after you get your best results. Before that you will crash at least once testing voltages.

• If you crash, after the mandatory reboot, recovery, absolutely unnecessary panic and sweating (been there), delete the .ini config file in TS folder so you can start all over again. Or if you can, reset the values in FIVR before it crashes again while being idle. if your UV was unstable you may crash while idling, so the faster way to start again is to delete the .ini before opening TS.

3) set CPU Core to -0.050mv and Cache to -0.025mv as a starting or conservative UV and run Cinebench r20 multicore test.

• Take notes of CPU package temp, CPU package power, limits reasons and the score obtained

4) go lower by 0,005 in each value and keep testing and taking notes

• the idea is that you get higher clocks, higher scores, lower temps, lower power consumption and no limit reasons.
• If you get it right, this CPU can reach 40x multiplier with a BCLK of 99.768 mhz = 3990.7 mhz under full load.
• Don't go crazy on the Cache as it will make you crash more often than the Core, go with a 2:1 ratio (2cpu:1cache)

I'm by no means an expert in TS, but those steps are the base of this whole undervolting magic thanks to @unclewebb

thank you so much for your help