Running smooth on a 154mV CPU undervolt, but msi afterburner crashes my games (Helios 300 G3-572-75L

da447m
da447m Member Posts: 17 Troubleshooter
edited November 2023 in 2019 Archives
So far, so good. 150mv undervolt using XTU (it amounts to 154 according to CPUID HWMonitor). Temps sometimes spike to ~80°C CPU and 75°C GPU, but otherwise can't complain as they seem to stay around 72 or so.

Not sure whether related, but after using msi afterburner to generate a curve for me, when I use the curve it simply crashes. It appears to be something related to game mode in win10, but I already disabled it following the hotel hero guy video. I didn't unlock the voltage since it has some pretty scaring warnings and I prefer to play on the safe side. Honestly I'd like to see some more temp dropping, more than actual overclocking the GPU (already repasted with kryonaut).

GPU undervoltage within XTU seems to amount to nothing at all. Tested up to 80mV GPU, but can't see any noticeable difference. How to proceed from here?

Curve generated by msi afterburner:

Answers

  • xapim
    xapim ACE Posts: 7,253 Pathfinder
    edited July 2019
    da447m Dont use xtu use TS instead 

    https://www.youtube.com/watch?v=XTPMwQKlfPE&t=5s
    https://www.youtube.com/watch?v=N3v3o21M8gg&t=27s
    https://www.youtube.com/watch?v=nYW7ROFx2Kk


    https://www.userbenchmark.com/UserRun/11532543

    UserBenchmarks: Game 43%, Desk 61%, Work 40%
    CPU: Intel Core i5-7300HQ - 63.5%
    GPU: Nvidia GTX 1050-Ti (Mobile) - 41.9%
    SSD: WDC WDS200T2B0B-00YS70 2TB - 71.4%
    HDD: WD WD10SPZX-00HKTT0 1TB - 93.7%
    RAM: Kingston HyperX DDR4 2666 C15 2x16GB - 76.8%
    MBD: Acer Predator G3-572

    I'm not an Acer employee. (just here to help in the best way i can)
    If my answer fixed you issue please accept it for any other users who search for it would find it quickly thanks :)
    If you want to learn more about undervolting/optimizing windows join the Predator fb group and youtube channel:

    Owner/Admin (HOTEL HERO/Red-Sand/Opoka Opoka)
    https://www.facebook.com/groups/PredatorHelios300
    https://www.youtube.com/channel/UCNJwGUHxSJ8FKqAhnOqQuAw
    Acer support:
    https://www.acer.com/ac/en/US/content/service-contact
    http://www.acer.com/worldwide/support/  


  • strider16
    strider16 Member Posts: 123 Skilled Fixer WiFi Icon
    XTU (and TS btw) only affects CPU and its integrated GPU, it may help keeping low power usage when doing anything except playing games, so you really only will see changes in nvidia gpu using afterburner.

    I think your frequency for gpu are quite high, I prefer to keep mine in the cool side, applied a small underclock, capping gpu at around 1900 MHz, and a bigger undervolt, so it hovers between 975 ~ 1010 mV
    Nothing peaks at the same time...
    https://www.userbenchmark.com/UserRun/33054103

  • da447m
    da447m Member Posts: 17 Troubleshooter
    strider16 said:
    XTU (and TS btw) only affects CPU and its integrated GPU, it may help keeping low power usage when doing anything except playing games, so you really only will see changes in nvidia gpu using afterburner.

    I think your frequency for gpu are quite high, I prefer to keep mine in the cool side, applied a small underclock, capping gpu at around 1900 MHz, and a bigger undervolt, so it hovers between 975 ~ 1010 mV
    About XTU/TS: Yep, I'm aware but I meant to say it didn't make any difference in temps. Presumably your system still keeps sending power to the integrated GPU even when idle, and capping that should help overall CPU temp.

    As for the rest, I finally managed to not get toasted. The thing is afterburner auto overclocking threw some 200mV LESS to all points of the whole curve with a setting of frequency (which is what happens when you simply raise he curve frequency).

    I just made some more direct thing, a more steep raise like below




    As you can see, I kept everything stock up to 1607MHZ (850mV), then slowly increased frequency for the voltage > 850mV and then steeper from around 1700 and 1790MHz to 1923MHZ, to achieve that at 900 mV.

    Things run ok now, all temps stay mostly below 66, spikes to 70+, but around low 60s. But this is what is reported by afterburner. Using CPUID HWMonitor, it register historical maximum spikes, and those can go ~80°C CPU and 75°C GPU.

    I should only notice that msi afterburner itself keeps changing my upper boundaries. It was initially supposed to be 1911MHZ maximum (flat line from there), at that level my voltage reached the maximum of exact 900mV. Then it changed alone to 1923MHZ. When I saw that I raised the whole straight line part to 1923MHZ and msi afterburner alone changed it to 1936MHZ. It seems that it gives some 950 mV around that clock. I'm keeping those and wont touch then for the time being, I had exactly one system reboot after maybe 10 hours straight while my system was open in the menu of a game and idle for some hours.

    To sum up: undervolt of both CPU and CPU CACHE = 154mv, along with that curve, plus repasting with kryonaut (didn't change any factory thermal pad), this gives very acceptable temps not greater than 70°C overall, keeping at mid/low 60s, while playing some of my games at maximum/movie settings. Idle temps are at low 40/41. Before it reached ~90°CPU before thermal repasting.






  • da447m
    da447m Member Posts: 17 Troubleshooter
    strider16 said:
    XTU (and TS btw) only affects CPU and its integrated GPU, it may help keeping low power usage when doing anything except playing games, so you really only will see changes in nvidia gpu using afterburner.

    I think your frequency for gpu are quite high, I prefer to keep mine in the cool side, applied a small underclock, capping gpu at around 1900 MHz, and a bigger undervolt, so it hovers between 975 ~ 1010 mV


    (***** lost a whole comment after editing it)

    To sum up my previous reply: about XTU/TS I'm aware of that, but the system keeps sending voltage to the integrated GPU even when not in use, so one could shave off some degrees off of the CPU by undervolting the integrated GPU.

    My curve was supposedly tested, and generated, by msi afterburner itself after half an hour or more! But forget it, now I'm using this curve:


    With this I stay in the average of low 60s in both GPU and CPU. Notice that everything is stock voltage up to 1607MHZ.

    obs: msi afterburner itself can change your handmade curve. My initial curve was exactly that but with a straight line at 1911MHZ for 900mV. Then, prob after a reboot, it changed it to 1923MHZ. I then made the 1911MHZ leftover left by the auto repositioning to 1923MHZ(what I mean is that the very first point of the straight line was left in original 1911).

    After many hours playing with no issue, using computer for other tasks, idle, etc, my system simply shut down at menu of a game after idle there for like 20 min. I booted the system, just to find out msi afterburner did the exact same to my adjusted curve, first point at 1923MHZ/900mV and the rest straight line of 1936MHZ beginning at 912mV. Since this is low 900mV I'm keeping it. If I have problems again I'll report, but all temps stay mostly at low/mid 60s.

    Msi afterburner with rivaturner will show you the average, if you look into something like CPUID HWMonitor you will see historical temps, and this is where you can see spikes to ~80°C CPU and 75°C GPU. Not sure how relevant these spikes really are tho. So far, I'd say this is a successful unvervolt and stable, at least the CPU part.





  • strider16
    strider16 Member Posts: 123 Skilled Fixer WiFi Icon
    Unless you are using an external monitor, it will always keep integrated gpu on, since it is on it that laptop display is attached. But it is very power efficient anyway. I could apply -135 mV undervolt on mine (with -150 mV on its cpu side), and it pass stress tests this way, you could try it too.

    Now, your gpu undervolt is quite good! I see undervolt curve messed all the time here too, but it seems somehow random, and it never drifts too much from the values I adjusted, so I just left it there. It seems like the program let us apply any number, but the silicon itself having minimum step values.

    I prefer using hwinfo to read values, it is more complete and less prone to wrong readings, but for instant measures I left gpu-z running, it plots a detailed graph for gpu, but it includes temps for cpu too, so I can see if cpu and gpu temp spikes are coincident.

    Nothing peaks at the same time...
    https://www.userbenchmark.com/UserRun/33054103

  • da447m
    da447m Member Posts: 17 Troubleshooter
    strider16 said:
    Unless you are using an external monitor, it will always keep integrated gpu on, since it is on it that laptop display is attached. But it is very power efficient anyway. I could apply -135 mV undervolt on mine (with -150 mV on its cpu side), and it pass stress tests this way, you could try it too.

    Now, your gpu undervolt is quite good! I see undervolt curve messed all the time here too, but it seems somehow random, and it never drifts too much from the values I adjusted, so I just left it there. It seems like the program let us apply any number, but the silicon itself having minimum step values.

    I prefer using hwinfo to read values, it is more complete and less prone to wrong readings, but for instant measures I left gpu-z running, it plots a detailed graph for gpu, but it includes temps for cpu too, so I can see if cpu and gpu temp spikes are coincident.


    I'm using an external monitor, I'm using effective -72mV undervolt on the integrated gpu. I might try your total later on, but I've got a black screen death with it plus that first msi profile, when I removed the undervolt it didn't happen, maybe coincidence.

    I have a new curve now, just made some tiny little adjustments related to what I said above, just made the curve a wee bit less steep.

    [clock](voltage)

    - equal to stock curve up to 1607, then 1689(862), 1721(875), 1835(893), 1885(900), 1923(912) and flat from here. This amounts to a maximum 206+ MHZ at the top frequency for the same voltage. I'm also running 455+MHZ memory overclock.

    It might be that some games could complain, in that case you want to make the previous values less steep, but 1911MHZ using 912mV should work for pascal series as per my google-fu.

    UNINGINE Superposition scores it at 9011, with stock profile the score was ~8300. Same temps in both situations, going as high as 71°C at most.

    Oh, my CPU doesn't get too hot under CPU-Z stress test and similars, the problem is, while gaming, the CPU gets extra heat from the GPU. Both share the same heatsink and the GPU is closes to the fans, so this is why stress testing the CPU alone gives good temps but gaming makes it spike much hotter.

    ps: I set your -150mV integrated gpu undervolt, it seems that it makes a further tiny difference in temp drop but it is difficult to say. Bench around 1916/1940 temps slowly rising to 66°C after some 4 min of stress test (similar to before I'd say). Room temp 22º C. I expect this going much badly when avg temp here is 33ºC, so I'd try to shave off more temps, maybe with liquid metal, let's see. So far very satisfied with the helios 300 anyways, easy and cheap to work around the absurd previous temps.

Categories