Clevo Overclocker's Lounge

Discussion in 'Sager/Clevo Reviews & Owners' Lounges' started by Spartan@HIDevolution, Mar 4, 2016.

  1. hissy

    hissy Notebook Consultant

    Reputations:
    53
    Messages:
    162
    Likes Received:
    40
    Trophy Points:
    41
    Oh so it isn't normal? Then, maybe you or someone else here will be able to help?
    75с on core counts as thermal issues? with constantly full 115w power draw.And the only perf cap reason i ever seen there under load is TDP.
    Without tweaking it holds whatever advertised by nvidia as max boost clock, 1440 or something.
    Under light enough load it has no problems boosting up to 2100, but for these cases i lock it manually at lowest possible clock to get ~same performance with passive cooling x)
     
    Last edited: Dec 14, 2019
  2. hissy

    hissy Notebook Consultant

    Reputations:
    53
    Messages:
    162
    Likes Received:
    40
    Trophy Points:
    41
    [oops bugged]
     
  3. Prema

    Prema Your Freedom, Your Choice

    Reputations:
    9,358
    Messages:
    6,297
    Likes Received:
    16,457
    Trophy Points:
    681
    We are strictly talking benching here and not long therm gaming clocks scenario. So to get your max score up you have to overclock the vRAM to at least 1850Mhz from the Max-P stock of 1750Mhz.
    Also lock the core in the 2000Mhz region at something like 0.950-0.975mV and cut/straighten the voltage curve above that.
    There will be no throttle reason shown (in current software) as no flags are triggered by stuff the falcon security chip does internally, but clocks begin to drop very early on due to its additional temperature targets long before the 87c cut off point that triggers the firmware thermal flag.
    So if you are currently above the aquator make use of the lower temps outside and make use of the FN +1 full fan mode... ;)
     
    Scerate likes this.
  4. Scerate

    Scerate Notebook Evangelist

    Reputations:
    224
    Messages:
    687
    Likes Received:
    650
    Trophy Points:
    106
    good to know, gonna try that on my 2080, Right now i just undervolted it a bit with a +180mhz Offset and +800mhz on memory

    @jc_denton @Prema @jaybee83

    guess what booted up INSTANTLY with the b-dies

    upload_2019-12-15_11-52-37.png

    just put straight 3200 in without any other tweaks, only had to POST once

    so if anyone has some hynix Gskill 2x8GB 3000c16 too here is the spreadsheet i made to Keep myself informed about the changes https://docs.google.com/spreadsheets/d/13rdr3nHzz638g8kzK5_Unw_thCf04K0c9MmSzQdpUIg/edit#gid=0

    Edit:
    upload_2019-12-15_12-24-37.png
    worked too, took about 3 training sessions
     
    Last edited: Dec 15, 2019
    jaybee83 and jc_denton like this.
  5. jc_denton

    jc_denton BGA? What a shame.

    Reputations:
    10,873
    Messages:
    3,034
    Likes Received:
    5,759
    Trophy Points:
    581
    There's just something very special about B-Die kits :D Good job @Scerate !
     
    jaybee83 and Scerate like this.
  6. hissy

    hissy Notebook Consultant

    Reputations:
    53
    Messages:
    162
    Likes Received:
    40
    Trophy Points:
    41
    Well it's already set up like that, except memory. It won't hold higher clocks at all. With benchmark load it instantly drops down to ~1600 range with TDP cap no matter where and how i lock it.
    60c with max fans, but absolutely zero effect on performance.
    1850 memory helped this time though, for whopping 30 points!! x)
    What may i miss here? Specific driver? Something in bios?
     
    Last edited: Dec 15, 2019
  7. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,348
    Messages:
    57,892
    Likes Received:
    17,613
    Trophy Points:
    931
    Your vram clocks drop? What happens if you go for something inbetween clock wise?
     
  8. jc_denton

    jc_denton BGA? What a shame.

    Reputations:
    10,873
    Messages:
    3,034
    Likes Received:
    5,759
    Trophy Points:
    581
    Hmm, it almost sounds like ECC causing a driver crash. Try running FireStrike at stock core clocks, ideally with locked voltage so that there isn't variation due to temperature. Then rerun FS but add +100Mhz to the memory clock until you see instability and compare. Scores should be higher up to the point when ECC kicks in. However, it also helps to go above that and see whether it was just a memory strap. Sometimes you hit these and score gets worse, but as soon as you go above it scores improve.
     
    Papusan likes this.
  9. hissy

    hissy Notebook Consultant

    Reputations:
    53
    Messages:
    162
    Likes Received:
    40
    Trophy Points:
    41
    Sorry, apparently my phrasing confused you. Not english speaker, you see.
    @jc_denton You probably right about memory strap, but it was somewhere at lower clocks, i've tested it in small steps. Now as i went straight for 1850 vram clock, it increased performance as expected.
    Thing is completely stable as it is. Vram clock stays where it was set. GPU works properly and clocks up and down as it pleases depending on load and limits. Nothing crashes.
    But as you can see, it's graphics score is whole 800 points lower that what you're achieving here and it cannot run any faster because of power limit.
    Under heavy load it just throttles down to 1600 range (exact 2 or 3 curve points around 1605mhz@745mv); i've manually tweaked the curve in this range so it will crash if i increase it just one step more. In lighter moments it may increase slightly above 1700.

    I feel that for your score average core clock during test must stay somewhere in upper 1800's range or higher.. But this is clearly impossible with just simple undervolt, even if i'm so unlucky to get a bad chip, and Prema there god the best one, i think.
    So i came here and i ask you for advice, I'm at least want to understand what else may be impacting these scores so much.

    to not be completely useless, here's some possibly useful info:
    I found that War Thunder game might serve as good and convenient stability test.
    What's special about it, is it reacts to gpu instability faster that anything other i've tried, and does it without dying. It's graphics engine crashes, but instantly just restarts without even pausing the game... and crashes again and again, and you can keep it running in background while tweaking your overclock and instantly see if it's stable.
     
    Last edited: Dec 15, 2019
  10. jc_denton

    jc_denton BGA? What a shame.

    Reputations:
    10,873
    Messages:
    3,034
    Likes Received:
    5,759
    Trophy Points:
    581
    Can you try this and see if there is any change?

    1. Reboot
    2. Open Afterburner
    3. Add +150 Mhz to core / 500 Mhz to mem ( the values you know work)
    4. Open voltage panel and lock voltage/frequency to 0.975mV by pressing "L" (do not level out the curve or change it)
    5. Hit apply
    6. Run FS

    Sometimes having multiple breaks in the curve can cause some funny stuff to happen.
     
Loading...

Share This Page