]]>

The Official MSI GT73VR Owners and Discussions Lounge

Discussion in 'MSI Reviews & Owners' Lounges' started by -=$tR|k3r=-, Aug 16, 2016.

  1. Falkentyne

    Falkentyne Notebook Virtuoso

    Reputations:
    2,153
    Messages:
    2,458
    Likes Received:
    2,857
    Trophy Points:
    181
    cores 1 and 3 are 10c hotter than cores 2 and 4? already?
    That can't happen this quickly?

    Yes as I explained, you lap the side of the heatsink with the -cooler- cores to make all the heatsink balance properly on the chip.
     
    Donald@HIDevolution likes this.
  2. Shehary

    Shehary Notebook Deity

    Reputations:
    265
    Messages:
    819
    Likes Received:
    1,179
    Trophy Points:
    156
    Yes Lap means sand, I use 2000 Grit only....in my opinion it's best grit to use, it won't shave off too much copper so less chances that you can destroy the HS by wrong sanding.

    I will run a test again tomorrow and post the result, we both use different pastes, @Falkentyne is not big fan of Grizzly Kryo but I found it is one of the best non conductive paste with highest mk/w I think around 13 mk/w which no other thermal paste offers.
     
    Donald@HIDevolution and hmscott like this.
  3. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,010
    Messages:
    14,552
    Likes Received:
    17,675
    Trophy Points:
    931
    The undervolt should only be a few clicks in XTU + power shell startup script if you are on Windows 10, or a few clicks in TS.

    At load with stock paste -100mV undervolt is pretty consistently dropping temps 10c. So there's still more cooling available to you when you get around to it.

    I'm happy for you that it's at a good point now, and thank you for posting updates here, please continue to keep us updated.
     
    Donald@HIDevolution likes this.
  4. Falkentyne

    Falkentyne Notebook Virtuoso

    Reputations:
    2,153
    Messages:
    2,458
    Likes Received:
    2,857
    Trophy Points:
    181
    @Shehary I have TWO 5g tubes of Grizzly Kryonaut. So not like I didn't try it. And w/mk values can be misleading. Look at the difference between Conductonaut and Liquid Ultra. Completely hogwash ratings.

    I'm not a big fan of it because temps started out really good, then degraded, with core #1 and #3 slowly getting hotter than core #2 and #4 because of uneven pressure with too little pressure on core 1 and 3, combined with small surface direct die contact to the CPU=leads to quick pumpout and dryout because of thermal stress. On desktops, this takes MUCH longer to happen because the heat is spread across the entire IHS surface rather than in a concenterated area and you have much higher static pressure also.

    So I switched to Conductonaut, which helped a lot but I still had annoying core temp differences in prime95 non AVX testing (i use prime for heatsink testing because I'm weird).
    I realized that WITHOUT AVX or FMA3 in prime95, i should have perfectly balanced core temps and i was still getting 3-6c core temp differences depending on overclock.
    And this matters when you're overclocking because if a core gets too hot, you will get WHEA correctable errors. Or BSOD.

    I tried sanding the metal posts (yes i was able to eventually sand a tiny bit) but i found out these actually never even touched the mainboard (SORRY!).
    So I worked on the heatsink. used a little 400 grit on the side that was too "cold" (too much pressure) but not much; i found actually that this low grit didn't even do very much anyway.

    So I switched to 2000 grit, first attempt got all cores equal but core 1 was getting hotter than the others by like 2C. I didnt like this.
    Seeing 75/73/73/73 is just wrong. or 89/87/87/87 at 4.7 ghz.

    So i worked again and now got everything down to 0C-1C difference and I'm happy. And this is with liquid metal.
    No degradation over high stress (4.7 ghz 1.270v or 4.5 ghz 1.175v, prime95 small FFT, without AVX) testing over several days. It's stable.

    @Ivan994 what exactly did you get 10c core temp differences in? Was it prime again? If it was AIDA64 you can ignore that. AVX loads will always give temp differentials. Impossible to avoid.
    Picture example:

    Hotter top cores on 0 and 2? (some programs label this as 1 and 3, some as 0 and 2).
    Then you lap side of heatsink with the cooler cores (bottom) first (keep in mind orientation of HS when removed), then after lapping bottom first, then give entire HS an even clean polish.

    When you're finished, the copper should have a nice really bright orange shine to it, instead of looking "Dirty" like in the picture. Doesn't have to be mirror like you see in some of those professional overclocking modding videos and pics. Just better than shown.


    1498412583-cpu2-deforme_mod.jpg

    gt73vr-heatsink-cores_sanding.jpg

    tl;dr:
    2000+ Grit sandpaper: do the bottom (COOLER CORES) until bottom is shiny.
    Then do the ENTIRE heatsink evenly until top is shiny even though bottom is already shiny, do the entire HS with even full strokes.
     
    Last edited: Feb 14, 2018
    Donald@HIDevolution likes this.
  5. GENOCID

    GENOCID Notebook Consultant

    Reputations:
    64
    Messages:
    179
    Likes Received:
    147
    Trophy Points:
    56

    my one is up to 14c difference in cine at the moment but other cores are at 61c on max fans on 4.2ghz ac dc loadline 5 max voltage 1.195. (score 923)
    when i order new heatsink im gonna sand the old one and ill try record video how ive done it if succesfull
    step by step from @Falkentyne suggestions
     
    Donald@HIDevolution and hmscott like this.
  6. Papusan

    Papusan BGABOOKS = That sucks!! STAHP! Dont buy FILTH...

    Reputations:
    8,404
    Messages:
    16,694
    Likes Received:
    23,251
    Trophy Points:
    931
    Mind you... Gelid Extreme have a lot lower thermal conductivity numbers [W/(m*k)] 8.5. You can't say Grizzly Kryonaut's higher 12.5 W/m*k will give you +47% better thermal conductivity. Same also for ICD who have a lot lower value 4.5 W/m*k (Aka 270% lower thermal conductivity numbers vs Kryonaut). FYI. Phobya NanoGrease Extreme have 16 W/m*k. Grizzly Kryonaut ain't a good option if your heatsink have uneven finish.
    Second this.
     
  7. Falkentyne

    Falkentyne Notebook Virtuoso

    Reputations:
    2,153
    Messages:
    2,458
    Likes Received:
    2,857
    Trophy Points:
    181
    I think part of the problem is that these pastes were not designed for direct die contact. It's why Intel uses that pigeon poop (more like putty or IC7 than normal paste) on the inside of their IHS. Because it lasts for years and can handle such high temps. Normal pastes have trouble with such high temps on direct die contact because of the small surface. Some people tried pastes like Kryonaut on delidded CPU and then reattach the heatsink, and while it was better -at first- than the Intel goop, it didn't last very long. Lasted much longer on top of IHS attached to the heatsink, because the heat is now spread out, makes imperfections on the heatsink surface less important now.

    On direct die contact, any imperfection will cause bad core temp differentials because of the tiny surface. And the core are not spread out all around the die, either. the four cores are actually close to the middle! Around the edges are the system agent and Ring and GPU and some other stuff.

    1502110755-schema-and-cpu-scratches.jpg
     
    Donald@HIDevolution and GENOCID like this.
  8. Shehary

    Shehary Notebook Deity

    Reputations:
    265
    Messages:
    819
    Likes Received:
    1,179
    Trophy Points:
    156
    I think I cannot undervolt any more, anything above -30 @4.1 or -65 @ 4.0 rebooting the laptop.:tears:

    or may be now I m too old for this....:mad2:
     
  9. GENOCID

    GENOCID Notebook Consultant

    Reputations:
    64
    Messages:
    179
    Likes Received:
    147
    Trophy Points:
    56
    ive tried to put my cpu on 4.6 stock voltage ive got to cine but error then so test not completed but ive seen temps risin to 89 where other 2 were only at 70c! which i think is really good on 4.6 ghz stock voltage .but yet again test wasnt completed so aybe there still would be few degrees up and i couldnt get a score so it wasnt stable but still... if ill be able to fix this core difference i will use phobya nanogrease again as i thinks its very solid thermal paste just need to work around it to make it as best as possible as @Falkentyne s chip right now lol..
     
  10. Phoenix

    Phoenix Super Tweaker

    Reputations:
    10,064
    Messages:
    16,031
    Likes Received:
    21,060
    Trophy Points:
    931
    GENOCID = [​IMG]

    Soon your taptop will turn into this:

    [​IMG]
     
Loading...

Share This Page