The ThrottleStop Guide

I was going to say, @unclewebb was dead on. It's hard to believe, I know - it seems so stable. Yet when idle, that's when the crashing happens. Also, on battery power, I am pretty sure there is extra limiting going on, because even with TS active, my Alienware clocks go way down on battery, even though I still have the same performance plan selected. (probably something is going on, I SHOULD be able to get max performance on battery, shouldn't I... or not... lol)

Anyway, just to add one thing - if you really want to play around some more, and if the greater undervolt is truly stable under all conditions on AC power, you could also use the option of setting a different TS profile when it sees that you're on battery power. This is under the Options popup at the top right, you can activate it and tell it which profile to switch to for battery power (and whether right away, or only once battery goes down to a certain percentage). You'd choose to do it right away. Then you'd set that profile (#4 is the default for battery) with much less aggressive undervolt settings, perhaps higher EPP value, etc. Depending on how quickly it crashes when you switch to battery, that might let you do the higher UV on AC power.

 

I did what you suggested, here are the results:



I don't get why mostly opengl performance continued to drop after each test, despite undervolt settings... but maybe it's normal. No throttle at all by the way, power limits are disabled by default on TS, in fact this system almost never throttles.
Anyway, I don't see much difference from one another. I think I'll have to compare also max temp values from TS sensors.

Just a note, as i haven't said it before: I've let Windows to manage speedshift values, because on this machine it is already activated and I didn't want Win10 and TS to fight for cpu control. So, I'm always using Windows balanced profile with speedshift setting automatically set to 84 (it switches to 0 in rare cases only).

 

If ThrottleStop detects any errors, they will be reported instead of the time being reported. If it shows the time value after a TS Bench test, no errors were detected.

The TS Bench is not the ultimate stress test by any means. I think of it as more of a basic stress test. You should definitely be 100% TS Bench stable with 0 errors being reported in any of the TS Bench tests you try. I recommend doing both single thread and multi-thread tests. On a laptop when undervolting, being error free in both of these tests is important. Too many users concentrate only on full load or uber load testing. Partial load is where most laptops spend most of their time.

Always a wise decision.

 

Not the big performance differences. Don't put so much trust in the Maxon OpenGL benchmarks. Have always beeen flawed. Use the settings who give you best temp. Switch to the other settings or decrease your undervolt if you later can see stability issues.

Btw... Have you tested with other voltage settings? As for example -90 / -100mV for both core and Cashe.

 

I did run every test possible with ts bench and prime95. Along with testing few games, i found the stable undervolt for my laptop. I'm currently running -160mv on core and -140 on cache. In my case, having more uv on core reduced power consumption and temperatures by a little margin than having matching uv with cache. Sadly the disable and lock power limits setting didn't have any effect, maybe coz they're locked by acer? It's not that necessary to me though, since i use lower multipliers while gaming. Once again, thank you so much for this amazing piece of software

 

@unclewebb I've noticed that when the clock modulation box is checked the CPU usage of TS goes from 0.02% to 0.13% and the results on Cinebench are ~10 points lower (from 683 to 673).
I suppose that's caused due to TS being constantly writing to the clock modulation MSR to make sure it is at the set value, is there any chance of making this feature less impactful (or aggresive) on the CPU performance in the future? Thank you

 

I could be wrong, but I do not think you need to have the Clock Modulation button checked if you are using Speedshift.
Joe

 

Nah, clock modulation is not related to Speed Shift.

According to Intel documentation clock modulation throttling should only be used as a last resort method to cool the CPU. In my case when DPTF is enabled clock modulation won't go below 100 as DPTF throttles the power when the temps are high.
But when I disable DPTF, clock modulation throttling appears whenever the system is a bit hot and it won't go away long time after it has cooled down.

 

Dell has long been a fan of using clock modulation throttling. If you are interested in the early days of Dell and clock modulation throttling, here is a 24 MB pdf download that tells all. This document is what originally motivated me to write TS.

https://drive.google.com/open?id=1tDRwD215OkMyx9BzTQRq1esBOXJrC4Dw

Anyway, there is already a hidden option within TS so you can adjust the frequency that TS checks the clock modulation and set multiplier registers. In the ThrottleStop.INI config file you can add this line.

TimerInterval=1000

That line tells TS to check these registers every 1000 ms which is every second. The available range is 4 ms to 1000 ms. Default value is a secret. Play around with that value and hopefully you can come up with a better compromise that is right for your laptop.

 

I set the TDP Level Control to 2 from the 0 and I "Lock" it. My question is, how can I unlock it? I already reboot my laptop but the "Lock" and the "TDP Level Control" is grayed out.