*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Discussion in 'Desktop Hardware' started by Mr. Fox, Nov 5, 2017.

  1. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    4,083
    Messages:
    11,557
    Likes Received:
    9,118
    Trophy Points:
    931
    If that actually happens id love to see two of those monster kingpin cards SLIed up with full waterblocks on them, omg :Do_O:confused::eek:
     
    Papusan and Johnksss like this.
  2. Johnksss

    Johnksss .

    Reputations:
    10,574
    Messages:
    18,990
    Likes Received:
    11,373
    Trophy Points:
    931
    I said that because people think they should be able to get what people on LN2 get and that is just not realistic thinking. They already tried the add more voltage on water and it produce worse results because the cards need to be far in to the - range to show improvement. Remember a while back when I said I have been getting lazy? Because I shouldn't even be doing anything on air or chilled water and all benchmarks should be done on nothing lower than Single Stage or higher, but this stuff gets rather expensive and one needs to take a break at times on the wallet.

    We have like 10+ of those cards right now at my local Microcenter right now.

    You would think so. Get in while the getting is good, but apparently something is wrong. And of course no one wants to give the real answer as to shareholders this and that and blah blah blah....

    I been said to do that. it's cool to do the new stuff, but if you don't stay on top of that, then you lose points fast. Most of those guys on top have many points using lessor hardware. The categories with the most submissions is where it's at. You stack up there, then it's allot harder to get knocked down the list.
     
    Papusan, Raiderman, Rage Set and 2 others like this.
  3. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,993
    Messages:
    6,576
    Likes Received:
    5,652
    Trophy Points:
    681
    Didn't Buildzoid or someone do a hard mod on a Pascal card that enabled physical dialing in of the core voltage by just turning a knob/dial - so bypassing any kind of firmware or software voltage control? Wouldn't that be possible with Turing?
     
  4. Johnksss

    Johnksss .

    Reputations:
    10,574
    Messages:
    18,990
    Likes Received:
    11,373
    Trophy Points:
    931
    That would be correct. They usually use controller pots to control the core, memory and pll voltage while monitoring temps and voltage with a VMM/DMM. Along with removing the power limits and insulating the card. These mods suck on anything not using extreme cold.
     
    Robbo99999 likes this.
  5. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,993
    Messages:
    6,576
    Likes Received:
    5,652
    Trophy Points:
    681
    Ah, what happens if you use those mods when not using extreme cold? If you're able to set the core voltage manually & accurately, then why is extreme cold important, couldn't you just set a realistic voltage that your cooling methods can deal with (like say 1.1 - 1.2V for the core for Mr Fox's behemoth water cooling)?
     
  6. Johnksss

    Johnksss .

    Reputations:
    10,574
    Messages:
    18,990
    Likes Received:
    11,373
    Trophy Points:
    931
    To be honest, you would think so, but from prior testing....Not really.
    The best way to try to explain it is this...
    The cpu and gpu core have a built in unseen temp setting in what can best be explained as stages.
    Example:
    Stages are as follows
    30C
    0
    -30
    -80
    -120
    And at each stage you are allowed to use more voltage or in some cases less or the same voltage.
    So say you wanted to hit 2300 on core, and the internal temp is set to -47C to -63C. Then you would need to between this range or else the test will crash or the gpu will crash at idle or any other host of anomalies because the core is considered to warm. Once your in the sweet spot for that 2300 range then you can generally bench almost all benchmarks. And as you go up that threshold changes again. Along with voltage needed to run a certain bench. This is why it does not scale well on air or even water. Chilled water gets you better results as does water running at -30C from a home made water chiller. Cards scale with temps, but then you hit that brick wall and no amount of voltage is going to help. The cores basically says. No. You need to be here and we will allow you to use that voltage and go further.
    That is an example not a complete science.

    Another example:
    Just like some users will have a lower voltage with a higher over clock and their score is much higher, but their temps are in the - range or around 0C under load. then the second you hit say 3C under load it crashes.
     
    Last edited: Apr 11, 2019
    ssj92, jaybee83, Mr. Fox and 2 others like this.
  7. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,993
    Messages:
    6,576
    Likes Received:
    5,652
    Trophy Points:
    681
    I see, that's an argument revolving around 'the fact' that Turing cards don't respond well to increased voltage beyond stock levels unless they are extremely cold. So it's not specifically related to the fact that the voltage is being manually controlled through the controller pots, instead it's just that Turing cards simply don't respond well to increased voltage unless very cold. So even a card that had software increasable voltage up to 1.2V would show the same limitations, so nothing to do with the controller pots. @Mr. Fox was keen to get increased voltage limit for his 2080ti, so you don't think that manual controller pot mod would help any because he can't run his cards cool enough with his current setup, he'd need colder temps? Same would go for a King Pin card that could hit software controllable 1.2V then.
     
    Johnksss and Mr. Fox like this.
  8. Mr. Fox

    Mr. Fox Undefiled BGA-Hating Elitist

    Reputations:
    28,804
    Messages:
    35,076
    Likes Received:
    55,662
    Trophy Points:
    931
    The voltage mod that worked for Pascal does not work for Turing. Bullzoid burned a lot of calories on trying to restore voltage control and it never worked correctly. The NVIDIOTS went way overboard with foolish nonsense to ruin things for us with Turing GPUs. The first six minutes of this video will sum things up pretty well... long story short, the retards that engineered Turing effed things up real bad. It passes the litmus test for gamers because stock performance is excellent. For overclockers, it fails the litmus test because it's trash for gamer kids, not a great product for overclocking enthusiasts.



    No so much other than they botched things up with both hardware and firmware. They took their engineering blunders from Pascal and put them on steroids to magnify them. If the firmware wasn't such utter rubbish these GPUs wouldn't require extreme cooling as a partial workaround. Even with extreme cooling they're not as good as they would be had the Green Goblin not screwed them up so bad. I think Bullzoid characterizing it as a 'fricking pile of garbage' is very accurate. LN2 should be an extreme thermal solution used by a very tiny minority of overclocking enthusiasts to achieve freakishly extreme results; not a partial workaround to trick the GPU into behaving the way it should behave on water if it were not screwed up. We cannot really characterize this as anything other than a totally botched up product if overclocking is your interest. There is no excuse for having to run sub-zero cooling for such modest overclocking gains. If you want to get really pissed off, watch the whole video.

    The only thing the NVIDIA brand has going for it now is the fact that their GPUs don't suck as bad at overclocking as AMD's GPUs suck at overclocking.
     
    Last edited: Apr 10, 2019
  9. Johnksss

    Johnksss .

    Reputations:
    10,574
    Messages:
    18,990
    Likes Received:
    11,373
    Trophy Points:
    931
    1080 TI Kingpins go up to something like 1.4+volts using just software, but it doesn't help any when the card is not cold enough. Here is what seemed to be the ideal spot on chilled water. Any higher and it would just crash.
    1.41V memory and 1.17V core.
    Software would only show core at 1.093V max or most of the time 1.063V
    20180711_180945_rs.jpg
    Edit:
    Not sure what was going on with Buildzoid, but others are able to run higher voltage just fine.
    https://wccftech.com/nvidia-geforce-rtx-2080-ti-galax-hof-oc-2-9-ghz-3dmark-world-records/
     
    Last edited: Apr 10, 2019
    jaybee83, Robbo99999, Papusan and 2 others like this.
  10. Raiderman

    Raiderman Notebook Deity

    Reputations:
    625
    Messages:
    940
    Likes Received:
    2,254
    Trophy Points:
    156
    The new Radeon doesnt suck at overclocking, it just needed new drivers. I havent tried the updated firmware yet either, as there is a new one out. They are getting better, and better.

    image_id_2169279.jpeg
     
    Last edited: Apr 10, 2019
    ssj92, jaybee83, Rage Set and 2 others like this.
Loading...

Share This Page