The ThrottleStop Guide

Discussion in 'Hardware Components and Aftermarket Upgrades' started by unclewebb, Nov 7, 2010.

  1. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,095
    Messages:
    5,862
    Likes Received:
    5,011
    Trophy Points:
    681
    I screwed up! I meant to say -250mV on the core and -125mV on the cache.

    Here is the Bob of All Trades video using a 9750H that shows this.
    https://tinyurl.com/y6almxqk

    At the 5:55 mark he takes the core from -125mV to -250mV. No crashes and he seems to get a boost in MHz after doing that. Too bad he has CPUID HWMonitor taking up screen real estate in his video. I would have liked to have seen the multi in ThrottleStop go up. I trust the TS multi more than some family members. :)

    That is normal. While Windows is booting up, ThrottleStop has not applied any of its settings. That helps ensure a smooth boot with less chance of a crash during boot. No more throttling flags lighting up after you boot up during normal use (or not so normal use) is great to hear.

    Now that your CPU is running great, how about enabling your C states again? I feel like a shady used car salesman trying to push those things on people. I like a cool running CPU and you actually get a 5% performance increase when running single threaded tasks with the C states enabled. They are free too. Something for nothing. Try running a few 1 or 2 Thread TS Bench tests with and without the C states enabled to prove this.

    I will see if I can find out what memory timings the Cinebench 8750H 1300 club are using.

    Go spread the word about TS. Every 8750H owner should give TS a try.
     
    Last edited: May 14, 2019
    minc3, bennyg, Papusan and 3 others like this.
  2. geust123455

    geust123455 Notebook Enthusiast

    Reputations:
    0
    Messages:
    15
    Likes Received:
    1
    Trophy Points:
    6
    -250mV on the core works. No crashes during boot or stressing but it made little to no difference. All the gains come from undervolting the cache so I stick with upping the cache more rather than the core. Very good to know that those warnings are normal and go away, so I don't have to worry about them :). I benched SSD speeds with crystaldiskmark and the differences of C states off and on were only within margin of error so I enabled them back on as per your suggestion and everything is running perfectly fine now. Thank you so much. I'll definitely recommend TS to everyone. Wish Intel recognized your work and gave you their full support.
     
    Last edited: May 14, 2019
  3. pantezuma

    pantezuma Notebook Enthusiast

    Reputations:
    0
    Messages:
    17
    Likes Received:
    13
    Trophy Points:
    6
    Hi @unclewebb
    Sorry to bother you, again.
    Do you have any idea why this could be happening? I mean TPL modifications only work after setting my computer to Sleep.
    I tried delaying the Task but no success...
     
  4. creyssonbr

    creyssonbr Notebook Guru

    Reputations:
    0
    Messages:
    65
    Likes Received:
    0
    Trophy Points:
    15
    Is the last version working with Intel I7 9750?
     
  5. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,095
    Messages:
    5,862
    Likes Received:
    5,011
    Trophy Points:
    681
    Some people have claimed that reducing the CPU Core offset voltage lower than the CPU Cache offset voltage works on some of the recent CPUs. I tried this trick on an 8th Gen U and I also did not see any change in performance, power consumption or core temperatures. I hope someone with an 8750H can post some tests to try to prove this one way or the other.

    I thought I just posted that information at the top of this page. I will post it again so you do not have to scroll up. If you want me to come over there and push the mouse button for you too, that is where I draw the line!

    I am pretty sure I know exactly why this is happening. I have heard of this problem on other laptops so it is not a unique problem. It sounds like one of the power limit registers is initially locked by the bios. When you start up Windows for the first time, you will not be able to make any changes to this register. After you do a Sleep - Resume cycle, this power register becomes unlocked so now ThrottleStop is able to adjust it without any problems. Hopefully you do not have Intel XTU or any of its remnants still installed. Unless you have access to a modified bios, there is probably nothing you are going to be able to do about this limitation. At least ThrottleStop works fine after you resume.
     
    Vasudev likes this.
  6. Xatanú

    Xatanú Notebook Consultant

    Reputations:
    54
    Messages:
    150
    Likes Received:
    165
    Trophy Points:
    56
    Hello again @unclewebb, I have a coupe of questions:
    1. I have set a profile with static voltage on both core and cache (same value), but I have observed that it keeps fluctuating all the time and is not constant. For example, if I set the voltage to 1.0537 (with all multipliers @ 46), the voltage keeps fluctuating between 1.0500V (sometimes a little bit lower) and 1.3000V. Is this normal? Is there anything I could do in TS to prevent this from happening and be able to get a constant voltage all the time, or it depends on BIOS and other factors and there is nothing I could do?

    2. When testing the undervolting settings to check if they are stable or not, I normally perform some tests, prime95, TS benchmark, Cinebench, and also keep some time at idle. When running the TS bench, sometimes it displays an error count, but the system does not crash, and the rest of the tests with prime95, cinebench, etc go well without crashing/errors. Is this an indicator that the undervolt values are not stable or is it normal? May I keep those undervolt values if the system does not crash even when TS bench shows some errors, or should I look for more conservative values? (e.g. when setting adaptive offset of -160.2mV @ 43 in all multipliers all the tests are fine and apparently the computer is stable but I get a LOT of errors in TS Bench but the system does not crash. If I set a higher offset I don't get any errors in TS Bench)

    Thank you very much for all your help and hard work!
     
    Last edited: May 14, 2019
  7. unclewebb

    unclewebb ThrottleStop Author

    Reputations:
    7,095
    Messages:
    5,862
    Likes Received:
    5,011
    Trophy Points:
    681
    I have done zero testing of static voltage so you are on your own. Very few people seem to be using the static voltage option in ThrottleStop. On desktop motherboards, I believe the voltage is controlled externally by a separate voltage regulator. When you go in the bios and request a fixed or static voltage, you actually get something pretty close to that. It might droop a little when fully loaded but it usually stays very close to what you request. I do not think the static voltage that ThrottleStop lets you set is comparable to that. I would probably stick with the TS Adaptive voltage setting and use either a negative or positive offset. Usually you only need a negative offset but if you are running high multipliers, maybe you might need a little bit of extra voltage. If your laptop has a fixed voltage setting available in the bios, give that a try and disable the TS voltage control.

    Prime95 and most modern torture test kind of benchmark programs usually bombard a CPU with a heavy percentage of AVX instructions. The TS Bench is the opposite. It does not use any AVX instructions. It was specifically created to give a good workout to the other side of your CPU. The TS Bench and Prime95 are not exactly typical real world apps but I think being TS Bench stable might actually be more important than being Prime95 stable. Why? Very few real world apps torture your CPU with AVX instructions the way Prime95 or LinX (Linpack) testing does.

    The TS Bench performs millions of calculations and then goes back and performs the exact same calculations a second time. If the first result does not equal the second result, that is a problem and the TS Bench flags that as an error. Think about it. Either the first calculation was wrong or the second calculation was wrong or it is even possible that they were both wrong.

    First time through, 2 + 2 = 4 and second time through, 2 + 2 = 5 or 2 + 2 = 105. I think that's a problem. I have talked with a couple of guys in the last few months that were just like you. They were questioning if TS Bench stable was important. It did not seem important because just like you, they had no problem passing all of the main stream benchmarking programs. They decided to ignore the TS Bench results and off they went to play some games. Guess what happened? BSOD! The TS Bench was right. Their CPU was not 100% stable.

    CPU stability can vary as a CPU wears in and it can vary from one day to the next depending on room temperature, etc. If you are not against the wall temperature wise, give your CPU a hair more voltage so you can be 100% stable in all benchmark programs, including the TS Bench.
     
    c69k, Xatanú and Papusan like this.
  8. Papusan

    Papusan JOKEBOOKs Sucks! Dont waste your $$$ on FILTHY

    Reputations:
    28,247
    Messages:
    25,434
    Likes Received:
    45,614
    Trophy Points:
    931
    8700K running i7-8750H clock speed and 2666Mhz ram. 3.9GHz all 6 cores...
    [​IMG]
    [​IMG]
    Yees:) But I fixed it for you:vbbiggrin:
     
    Last edited: May 14, 2019
    Queen 6, Vasudev and tilleroftheearth like this.
  9. Xatanú

    Xatanú Notebook Consultant

    Reputations:
    54
    Messages:
    150
    Likes Received:
    165
    Trophy Points:
    56
    Thank you very much for all your help.
    I don't want to run a potentially inestable system and get a BSOD in the worst moment, so I will experiment with more conservative values and perform more tests, specially TS Bench until all seems rock solid.
    Regarding static voltages, I have seen @MrFox using them in one of his videos. If I was able to get the static voltage to stick and don't fluctuate I think I would use static instead of adaptive as it should be easier to tune up and less prone for the system to hang when idle for going too low in voltage at low frequencies. The ideal for me would be to have a voltage curve and be able to tune it up like I do in Synapse or Kernel Audiutor in Android custom kernels or in Afterburner for the GPU. I don't know if that would be possible to implement. If not, I think it would be very useful to be able to set a minimum voltage when using adaptive to prevent crashes at low frequencies when idle.

    I will keep experimenting.
     
  10. Queen 6

    Queen 6 Notebook Enthusiast

    Reputations:
    3
    Messages:
    11
    Likes Received:
    14
    Trophy Points:
    6
    8750H - 2666MHz RAM, 3.9GHz all cores, but mine's in the 3K+ club :p
    3100CB (No Taskbar).jpg

    Q-6
     
    Last edited: May 15, 2019
    Vasudev likes this.
Loading...

Share This Page