*** The Official MSI GT75 Owners and Discussions Lounge ***

Discussion in 'MSI Reviews & Owners' Lounges' started by Ultra Male, Jun 23, 2017.

  1. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    I'm optimist. Don't make me sad before its time.)
     
  2. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    So, I put IC graphite pads on CPU and GPU, decided to put tiny amount of thermal paste just on corners to hold it in place, but maybe I shouldn't. 20180814_162203.jpg

    CPU runs cooler and difference between cores is less.

    Before I had two hot cores, now I have two slightly hotter and one slightly colder. Average temps are 4°C less and CPU package 6-7°C less. IC-prime95-fullfan.png
     
  3. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    Completely screw up with GPU though, I think I damaged graphite pad or thermal paste got somewhere. Now it 3°C hotter, so I gonna redo it.
     
  4. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    Nope. It's just for some reason GPU runs 4°C hotter with IC graphite pad, while CPU runs colder 6° with same pad and same fan RPMs...

    Maybe there's not enough pressure on GPU heat sink?
     
  5. Falkentyne

    Falkentyne Notebook Prophet

    Reputations:
    7,147
    Messages:
    5,427
    Likes Received:
    7,509
    Trophy Points:
    681
    You should not have repasted the GPU.
    The GPU was just fine already because the contact was already good, so the MSI original paste worked fine.
    The IC pad is going to be worse there. The GPU will improve if you use a good paste like Kryonaut, Phobya Nanogrease Extreme, or Coolermaster Mastergel Maker Nano.
    The GPU heatsink is flat, so is the GPU, so the MSI original paste works about as good as MX-4. The IC pad is worse than MX-4.
    Best thing to do now is buy a good thermal paste and re-do the GPU.

    For GPU thermal pads, use Arctic 1mm. Works great, and the GPU thermal pads are on a different plate and do not affect GPU contact. (unlike the CPU).

    On the CPU, the bga chip AND the heatsink are both convex, so the contact is worse. The thicker IC pad thus helps the convex surfaces keep contact better, thats why you had an improvement. Notice you still have core 0+2 being hotter, but now it wont degrade to 20C hotter than core 1 and 3. You can also try changing the CPU pads to 0.5mm pads and stretch them out a bit to compress them even more (You need arctic pads for that). The best way to compress 0.5mm pads is to apply them then do a combination of pressing with your fingers hard and stretching slowly, then when they stretch, then trim them.
     
    raz8020, selfmadeboss and prodj like this.
  6. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    That's what I said before I saw that there's no way of getting to the CPU without dismantling all the GPU heatsinks.
    My OCD tells me to do all that (I have 0.5mm arctic already), but my laziness says: it's good enough.
    So, I'll decide later..
     
    hmscott and Falkentyne like this.
  7. prodj

    prodj Notebook Consultant

    Reputations:
    15
    Messages:
    104
    Likes Received:
    24
    Trophy Points:
    31
    Removed graphite pad from GPU, repasted with Nano, and temp dropped by 10°C on full load full fan. Became even lower than it was with stock paste (what a surprise.)
    65°C in AIDA's 100% stress on max 5000rpm looks fine to me.

    The pad itself looks squished, so it had enough pressure. gpu pad.jpg
    So my guess is:
    Surface of the GPU's heat sink has not flat texture. It is not convex or something, but its surface seems to have parallel lines in one direction (like micro-trenches), I noticed that when I was cleaning it from old paste, easier to clean along these lines, but not the other way.
    CPU's heat sink on the other hand didn't seem to have bumpy surface in any way, so maybe that's why I have better temps with graphite pad. (not sure, any other opinions?)

    I think I will try nano paste for CPU, but I don't think that after gaining -6°C with a pad I can gain much more... we'll see.. but not today... too drunk and too tired.
     
    hmscott likes this.
  8. Falkentyne

    Falkentyne Notebook Prophet

    Reputations:
    7,147
    Messages:
    5,427
    Likes Received:
    7,509
    Trophy Points:
    681
    Nothing to do with that.
    You had better temps on the CPU because the contact was better. The CPU heatsink is convex. SO is the CPU silicon. And then you have VRM elevation on top of that. So you have very bad contact pressure on the top side and ok contact pressure on the bottom side. So the stock Lairdtech pad sucks. The Panasonic pad is thick enough to keep contact on all the cores without risk of pumpout or dryout (I believe the pad is 0.2mm thick, maybe 0.1mm I don't remember, but close to the same of the height of the BGA raised above the housing. So there was better contact with the PGS pad than the stock paste, thus better temps. Better than stock paste but not as good as a good fresh application of a good paste. But without sanding the heatsink flat you do run the risk of the good paste 'drying out' on the side of the cores facing the VRM's due to lower pressure, and trying to balance a convex heatsink+convex CPU is not easy without hardware modification. Using 0.5mm pads and stretching and compressing them to 0.2mm does improve things a lot.

    Using Nano on the CPU would give you at least 5C lower temps than the Soft PGS. I know. I've tested Nanogrese extreme, Kryonaut, Coolermaster Mastergel Maker Nano and Liquid Metal and the pad.

    On the GPU, contact pressure was already solid and even so the stock MSI paste (Lairdtech thermal stamp) is still better than that PGS pad.

    Here is what a sanded heatsink and stretched 0.5mm thermal pads will get you with Liquid Metal.

    lm_sanded.jpg
     
    raz8020, Papusan and prodj like this.
  9. selfmadeboss

    selfmadeboss Notebook Enthusiast

    Reputations:
    1
    Messages:
    10
    Likes Received:
    3
    Trophy Points:
    6
    Is overclocking the 7820 hk worth it. I just bought it and I want to improve the clock speed to about 4,5 GHz
    Right now I’m at 4,2 GHz stable (done via dragon center) even if I run Aida stress test for about 20 minutes there’s no throttling
    Since I don’t want to void the warranty and I haven’t applied any thermal paste I’d like to know if it is possible to achieve 4,5 GHz without any hardware modifications just simply with software or maybe with under volting

    What clock speed are you guys running your 7820hk?
     
  10. Falkentyne

    Falkentyne Notebook Prophet

    Reputations:
    7,147
    Messages:
    5,427
    Likes Received:
    7,509
    Trophy Points:
    681
    4.5 ghz on stock paste is doable but you're going to be hitting 90C on anything using all 8 threads even at 100% fan speed. Whether or not you can do it without a POSITIVE voltage offset is anyone's guess. You are not going to be able to undervolt and remain stable. Even with a repaste, if you do 4.5 ghz without an undervolt, your laptop is going to overvolt for you (and not display it properly either) by up to 150mv at full load and make you overheat, if you don't set IA AC DC loadline to "1" first, and then re-tune your voltages manually.

    If you don't have a bad chip, Liquid Metal should get you to 4.7 ghz, anywhere between 1.25v-1.3v real voltage (this is assuming IA AC DC loadline is set to "1" rather than Auto. Good chips can do 4.8 ghz at 1.3v+. Excellent chips may do 4.9 ghz at 1.4v on LM but you're going to overheat FAST on anything using all 8 threads, and if you pass 92C, the zener diodes may power trip the laptop (Don't ask).

    Due to the way the internal overvolting works on this laptop, you will not be able to exceed 4.5 ghz and stay under 100C, without setting IA AC / IA DC loadline to "1' in the Bios, otherwise getting something like 1.25v at *IDLE* will shoot up to 1.4v (!) at full load, and the VID will be showing something like 1.31v (instead of 1.4v) while you will be wondering why your temps are reading 100C....
     
    raz8020, selfmadeboss and Papusan like this.
Loading...

Share This Page