XPS 15 9570 Owners Thread

Discussion in 'Dell XPS and Studio XPS' started by el3ctronics, May 16, 2018.

  1. _sem_

    _sem_ Notebook Deity

    Reputations:
    95
    Messages:
    761
    Likes Received:
    266
    Trophy Points:
    76
    This likely means that the throttling wasn't so bad. Many benchmarks don't stress the CPU+GPU that much as to end up deeply PL throttling, what you'd se running eg Prime95 and Heaven at once. Which is also the case for most in real life. But some games and GPU-assisted number-crunching wares do produce such loads, so one also can't assume this is irrelevant for all.
     
  2. abujafar

    abujafar Notebook Evangelist

    Reputations:
    56
    Messages:
    357
    Likes Received:
    195
    Trophy Points:
    56
    Thank you very much for sharing your results @bullerwins.
    Are you able to push the undervolt more or is 150 your limit?

    Sometimes you don't see a change in temps BUT you are able to sustain a higher clock (at the same temp).

    It also seems that the 1.1.4 update did improve the situation.

    If you have time, could you please run the Realbench Stress Test for 10 min (not the benchmark) and record everything with HWInfo64 on a text file? Upload it on pastedbin. It would really be helpful for us.
     
  3. bullerwins

    bullerwins Newbie

    Reputations:
    10
    Messages:
    8
    Likes Received:
    9
    Trophy Points:
    6
    You're right, the fans might be running slower too, but i can't confirm.
    I can push to 160-165, even 170. 175 crashed. I backed down thinking too much undervolt might be the cause for the power throttle (maybe you guys can help me with that) As long as the system is stable the further the undervolt the better?

    Done realbench stresstest for 12min: https://pastebin.com/J0fBzP9U
    I can upload the .csv if the formatted is screwed

    edit: the test is done on battery but the best performance selected. If that is an issue (the realbench test had the same results just to check if that afected)
     
    pressing likes this.
  4. abujafar

    abujafar Notebook Evangelist

    Reputations:
    56
    Messages:
    357
    Likes Received:
    195
    Trophy Points:
    56
    @bullerwins Thanks. From what I can see, there is something strange in this test. It most likely is because it was run on battery.
    Unfortunately understanding what exactly is going on is not straightforward. It takes some digging in the log and some guesswork. Sometimes the same flag is triggered by two different things.

    At this point, I believe that there are several layers of "protection" and also "battery saving features" in play here. We have Dell's Bios, Dell's Power Management, Windows Power Management, Intel's DPTF. Maybe you set one of these to performance but not the others?

    Your ran was very *cold* (you never went above 70c in any sensor) but the CPU clock was highly limited (not because of thermal protection or thermally-induced power limits).

    Plese, rerun this on AC if you have time.
     
  5. abujafar

    abujafar Notebook Evangelist

    Reputations:
    56
    Messages:
    357
    Likes Received:
    195
    Trophy Points:
    56
    I am a believer in pushing as long as it is stable. Not sure how much undervolting the iGPU helps.
    (Undervolting reduces power limit throttling, not the contrary!)
     
  6. bullerwins

    bullerwins Newbie

    Reputations:
    10
    Messages:
    8
    Likes Received:
    9
    Trophy Points:
    6
    Will repeat it later this night. I thought the same. I will test stability going with a higher undervolt too. With CPU core, cache, igpu its pretty straight forward but to undervolt the gpu with msi afterburner I just eye balled the volt curve a bit, any tip on that?
     
  7. custom90gt

    custom90gt Doc Mod Super Moderator

    Reputations:
    7,285
    Messages:
    3,310
    Likes Received:
    3,537
    Trophy Points:
    331
    So I will tell you that the cache undervolt typically causes blue screens. Undervolting the CPU past a certain point seems to limit the turbo boost that it can obtain but seems to be fairly stable. With MSI afterbuner you can actually overclock the GPU and it should help it run cooler (it is like shifting the voltage curve).
     
  8. bullerwins

    bullerwins Newbie

    Reputations:
    10
    Messages:
    8
    Likes Received:
    9
    Trophy Points:
    6
    abujafar likes this.
  9. abujafar

    abujafar Notebook Evangelist

    Reputations:
    56
    Messages:
    357
    Likes Received:
    195
    Trophy Points:
    56
    Yes, this shows signs of a power limit enforced to the CPU.
    I am gonna dig into the data you shared to see if I can find what triggers the power limit.

    Do you happen to have a fan or something that you can direct at the center of the backplate's grills?
    If so, try to use it and repeat the test.
     
  10. abujafar

    abujafar Notebook Evangelist

    Reputations:
    56
    Messages:
    357
    Likes Received:
    195
    Trophy Points:
    56
    I went through the data.
    This stress test is pretty brutal. It stresses CPU and both the internal GPU and the discrete GPU.

    After about 2 minutes of sustained load on both CPU (all cores) and GPUs, the CPU PL is reduced from 56W to 15W (with a consequent drop in freq).
    The GPU is largely unaffected by this. Actually, it seems that whatever is happening, it's favoring the GPU and crippling the CPU.

    Pinpointing what causing this behavior is not easy.
    - The CPU and GPU temps are high but not dramatically high. (and the CPU does cool down because of the throttling)

    - Someone mentioned that the problem might be the RAM. But your ram is actually quite chilly.

    - The only thing that I see running hot is one Ambient sensor which breaks 90c and won't cool down despite the CPU throttling. (But I am not sure this is the source of the behaviour)
     
    Last edited: Jun 6, 2018
Loading...

Share This Page