XPS 15 9550 temperature observations (undervolt + repaste)

Discussion in 'Dell XPS and Studio XPS' started by custom90gt, Dec 28, 2015.

Thread Status:
Not open for further replies.
  1. custom90gt

    custom90gt Doc Mod Super Moderator

    Reputations:
    7,425
    Messages:
    3,416
    Likes Received:
    3,717
    Trophy Points:
    331
    I recall your post about the VRM cooling with a fan. I was thinking something much simpler and a lot less elegant. Perhaps a sheet of copper that sits on top of the taller R22-604-111s (from the pic on page 27) with thin thermal pad between them and a thicker thermal pad to contact the VRM sensor. All of that followed up with a thicker thermal pad to sink it to the case. I dunno, I'm only pondering things in between studying for my cardiovascular test on tuesday.

    I don't think silicon thermal pads would pose a problem with electrical conductivity, but perhaps someone else can chime in on that. Also they may not transfer enough heat and may actually compound the problem.
     
    pressing likes this.
  2. pressing

    pressing Notebook Deity

    Reputations:
    282
    Messages:
    1,839
    Likes Received:
    922
    Trophy Points:
    131
    GonZ0 loved that fan idea!

    Thermal pad manufacturers typically provide electrical data in their sheets. For example, Fujipoly published some data on pages 15-16 below:
    http://www.fujipoly.com/usa/assets/files/Fujipoly 2015 Sarcon Catalog for web.pdf#page=8

    In the draft picture, I jotted down a quick and dirty schematic guessing the source of the key components; maybe that will be useful for someone some day. Some of those parts are special order-not easily identified. The picture should be considered a starting point for your own research.

    - I would guess the 3 TI87351 MOSFETs (PQ1000-PQ1002) and the 3 Alpha Omega MOSFETS ZS019QI (PU1200-PU1202) produce quite a bit of heat as my audio equipment has massive heatsinks attached to the MOSFETs.

    - You will also see several nearly-microscopic tantalum SMD caps in that area (PC1003...). Those can be sensitive to heat (had to replace a bunch of them in my car's instrument cluster last year due to heat - difficult to diagnosis and tough to solider anything at that size).

    Dell 9550 VRM.jpg
     
    einsteinchen likes this.
  3. Brian Anderson

    Brian Anderson Newbie

    Reputations:
    4
    Messages:
    4
    Likes Received:
    9
    Trophy Points:
    6
    A heatsink really won't help because there isn't any airflow apart from that generated by convection.
    What is needed is air movement over this area.
    I just ran a test where I placed a small fan under the Dell logo blowing air on the regulators and it kept the temps in check to a point that the GPU was the main cause of throttling:
    - VRM ~75C
    - CPU ~77C
    - GPU ~89C
    This was while playing Overwatch on high settings at above 1080 res capped at 60FPS.
    Without the fan the VRM hovered at 90C with PL1 at 15W.
    With the fan the VRM was around 75C and PL1 at 20W.
    Room temperature was 25C (77F).
    My new main logic board (replaced recently due to the audio problem that is common among these machines) seems to have less difference between VRM and the other components as my first one, and the temps leveled at 90C instead of 100C - but maybe that's because of the new (v13) BIOS? Too many variables to tell with any degree of confidence. I'm happy that the VRM temp is roughly tracking with the GPU temp though! Less throttling overall. :)
     
    pressing likes this.
  4. GoNz0

    GoNz0 Notebook Prophet

    Reputations:
    254
    Messages:
    4,062
    Likes Received:
    1,391
    Trophy Points:
    231
    Sat on my phone so I can't be arsed to go back and start quoting but whoever it was that suggested cooling the vrm sensor needs to slap themselves!
    It is an ambient sensor so it monitors the area. I found it when Aida64 reported the vrm as memory so the developer could correct it. Cooling the vrm's would be great if the chassis could dissipate the heat. Who knows, I may give it a go one day to see how well it works.

    Sent from my SM-G920F
     
  5. custom90gt

    custom90gt Doc Mod Super Moderator

    Reputations:
    7,425
    Messages:
    3,416
    Likes Received:
    3,717
    Trophy Points:
    331
    lol, you need to get off your high horse. You're the one that used a q-tip to cool the sensor down to locate it. You can either cool the sensor to trick it or go ahead and cool the rest of the VRMs down. The copper plate may allow you to cool everything down and sink it to the bottom. You also assumed that the VRM wasn't causing issues, when it seems that during gaming people are having throttling due to the VRM temps.
     
  6. GoNz0

    GoNz0 Notebook Prophet

    Reputations:
    254
    Messages:
    4,062
    Likes Received:
    1,391
    Trophy Points:
    231
    Not on a high horse but why would you want to cool the sensor and mask the problem until you break the mainboard when the vrm's crack?

    Should my none contact temp probe ever arrive from China I will see if I can get a temp from the vrm's as I expect they are running a lot higher.

    Sent from my SM-G920F
     
    pressing likes this.
  7. custom90gt

    custom90gt Doc Mod Super Moderator

    Reputations:
    7,425
    Messages:
    3,416
    Likes Received:
    3,717
    Trophy Points:
    331
    I think half the problem is we don't know how hot the VRMs are actually getting. I've read threads where someone uninstalled the intel thermal platform drivers and their VRM sensor readings were lower. I don't know how well setup the VRM sensor is to compensate for the spacing from the actual VRMs.

    If I still worked as a mechanical engineer I could throw some thermocouples on the VRMs and sensor and see what was happening, but I don't have access to any of that anymore.
     
    pressing likes this.
  8. pressing

    pressing Notebook Deity

    Reputations:
    282
    Messages:
    1,839
    Likes Received:
    922
    Trophy Points:
    131
    Brian - Super interesting solution! Could you elaborate on what you are saying here? Do you mean the Dell logo just below the screen? Do you mean you put a small fan just above the f7/f8 keys effectively laying horizontally on top of those keys? How big was the fan? Which fan was it?

    Which audio problem (lol)? Do you mean the crackling sound when adjusting brightness via f11/f12 keys?
     
  9. einsteinchen

    einsteinchen Notebook Geek

    Reputations:
    4
    Messages:
    83
    Likes Received:
    21
    Trophy Points:
    16
    I know that Dell uses used parts as replacements. Beside the throttling my XPS also suffers from quite a bit of coil whine. Chances are that I end up with a better board than before :) In the case Dell does not fix the problem after the second attempt I can return the notebook and get a refund (according to german law, I do not know how this is handled in other countries):

    Back to topic. Yesterday I repasted again and one VRAM chip definitely does not make contact with the heat sink. I put a thicker pad on this vram (had not enough pads for all four), but temp still go above 90°C, which still leads to throttling (1.8 - 2.3 GHz).
     
    Last edited: Sep 4, 2016
  10. Gerty

    Gerty Notebook Enthusiast

    Reputations:
    2
    Messages:
    11
    Likes Received:
    3
    Trophy Points:
    6
    Ive just had an Overwatch gaming session (low-medium settings) with hwinfo in the background...
    The lag kicked in perhaps 15 minutes in, but it was not as severe as usual.. after checking the numbers it is true that PL1 Power limit reached a minimum of 7W, and the cpu clock minimum was 800MHz.
    Maximum values for temparatures reported in the block containing "ambient" meters are:
    CPU: 82
    Ambient: 92
    Ambient: 77
    Temp3: 51
    Ambient: 70
    Ambient: 74
    Ambient: 54
    DIMM: 65
    CPU-fan ~5000RPM
    GPU-fan ~5000RPM

    I dont know what these specific ambient readings represent, but the 2nd in the list has a relatively low value of 77C, this should not be in the problematic range?
    Granted, this time the lag did not manifest in a completely wrecked framerate with cracking audio, but still in the unplayable range.
    Weirdly enough lately I am having issues with ping spikes (1000+) and rubber banding and im not sure if these are all related, or if the bluetooth/wifi is crapping up on me (awaiting on card replacement).
     
Loading...
Thread Status:
Not open for further replies.

Share This Page