What thickness for thermal pads can I use on GPU VRM chips for Acer Predator Helios 300 (PH315-52)?

1246

Answers

  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    today i made some changes...i dont have any kind of throtteling, also i did AIDA64 test for about 40min. look at the picture

    What’s your temps in aida when you select all 5 checkboxes except disks?
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    With intel GPU test or without ?
  • zerophyte
    zerophyte Member Posts: 9

    Tinkerer


    I reccommend this kind of pasting over dots. My temperatures in aida 64 after 1h fans at 4500rpm 60'C
    So just put a line of paste on the die like that ands then let the heatsink squish and spread it out by itself?
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    @zerophyte yes just let it down so it squishes ...it will cover 100% off cpu and gpu. Like on the picture dont go to the end with the paste.

    (for PCs- dont use this tehnique on big square ryzon chips they need another tehnique.)
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    @belik i left it for 10min i dont have more time now.
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    @belik i left it for 10min i dont have more time now.
    I mean following

    1. If possible, check your ambient temp
    2. Put laptop on plain table, don't lift it up
    3. Set PredatorSense to Normal mode (i.e. factory 45W limit, -125 undervolt. no other tweaks like Throttlestop etc) and fans to auto, no coolboost
    4. Run Aida64 with 5 checkboxes On (CPU, FPU, Cache, Memory, GPU) for 20-25 min
    5. Post a screenshot like this one below



    Mine behavior is following

    At ambient 22C, first 15  min fans run at 3500+4000 rpm, temps reached 86C on CPU and 81C on GPU
    Then, after GPU temp reached 82C, fans swtiched to next profile 4000+4500 rpm and temperatures stabilized at 83C on CPU and 77C on GPU
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon


    Ambient temperature is 21'C -  factory 45W limit, -125 undervolt
  • zerophyte
    zerophyte Member Posts: 9

    Tinkerer

    So after applying 0.5mm pads to VRM with some noctua drops on the 3 mentioned chips with pad over top and applying noctua to both cpu and gpu, I have better results than before. 

    Back when I used arctic mx4, I was hitting high CPU temps in my he 84 to 91 degrees Celsius when playing Destiny 2. And that was with a cooler pad underneath running at max. 

    Now with noctua nt h2, it runs at a max of 74 to 76 degrees Celsius for the CPU without cooler pad running. With cooler pad on max fans underneath, I’m averaging 64 to 74 degrees Celsius. Quite an improvement I have to say. That or I did a better paste job on the dies this time. 
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020

    Ambient temperature is 21'C -  factory 45W limit, -125 undervolt
    I see, it looks that 17 version designed properly in sense of air intake and case grills because in 15 ver I have 82C CPU/77C GPU under stress test but when I rise back edge by several cm up (like put a book under) temps immediately drop to 78C CPU/72C GPU, i.e. 4-5C drop just due to lifting up


  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    zerophyte said:
    So after applying 0.5mm pads to VRM with some noctua drops on the 3 mentioned chips with pad over top and applying noctua to both cpu and gpu

    yep you did all right and results are fine
    is your ver 15"? as I explained above you can even not turn fans in cool pad on, you need just kind of mesh table and temps already will drop by 5C

    you can increase set undervolt to -165 or -170 and it will give several degrees improvements more
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    I see two differences more @Citizenkane
    1) your fans is at 4000+4500 while with this temp they must be at 3000+3500 or 3500+4000, maybe you activated fast profile instead of normal or smth 
    2) your cpu is 100Mhz lower than mine maybe it’s due to gpu takes more tdp or smth else anyway not a big difference 
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    belik said:
    you can increase set undervolt to -165 or -170 and it will give several degrees improvements more
    when i decrees to -155 CPU stress test is stable but if i start a game laptop restarts goes back to defaults. 
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    Which game? I used -170 during Doom 2016 even with memory at 3200Mhz (I e with extra stress on memory controller in cpu) and had no issues
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    Star Citizen
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    Even -170 is not sufficient to pass latest XTU stress test AVX2 at constant 4GHz, it requires  -175
    Mine is stable at -175 so far

    What’s your Timespy scores? Mine is below 
    https://www.3dmark.com/3dm/43555047
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    So you play games with -170 or -155 ?
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    I didn’t define final value yet
    The highest undervolt I can pass Timespy under extreme profile (i.e. 56W CPU 90W GPU) is -170

    All higher tells me that error occurred and shows 0 scores  =)
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    Your card 1660? It can’t go above 80W then
  • Citizenkane
    Citizenkane Member Posts: 48 Die Hard WiFi Icon
    Can you tell me the most stable undervolt so i can test it ?
  • belik
    belik Member Posts: 397 Seasoned Practitioner WiFi Icon
    edited February 2020
    Try if you can pass Timespy -170
    For me it’s fine so far maybe I will reduce it to -160 or -165 later just to have safety gap