EVGA

Hot!2080ti performance cut in half

Author
XiloTheOdd
New Member
  • Total Posts : 10
  • Reward points : 0
  • Joined: 2024/02/18 17:43:25
  • Status: offline
  • Ribbons : 0
2024/02/18 18:01:30 (permalink)
so before i started my project to put a water block on my GPU i benchmarked the card to get an average performance score. no excessive overclocking, just gave it 130% power draw capability (which i do in winter and temps remain fine.). baseline i used was 3dmark and the GPU score by the end was about 14500. respectable and average. then i began the tear down to remove the cooler, the plate under the cooler and then to add the water block when i realized, this waterblock wasnt for a 2080 ti but for a normal 2080 RTX. disappointed i put everything back together, re thermaled with the new pads and a new graphene? strip in place of goop. as the last time i gooped it the stuff all went to the side of the GPU processor. not really helping at that point.
 
decided to run another bench test to make sure everything would be fine temprature wise and it was down to 60C peak. was about to celebrate when the score came out 7100, i had no idea **** had happened. the hotspot temp was still hitting its 109C high point (no idea where that sensor is located) and now i'm worried about many possibilities of what went wrong with reassembly.
 
i've updated drivers
thermals arent an issue
clock speeds are hitting past the stock boost clock
despite giving the card the normal 130% power draw, it is not drawing more than 70% on average (though the HWM will note a spike to 120% at the start of any bench test then it will never go that high again.)
 
i've checked power cables.
i've checked bios settings to make sure the cards slot is still on its fastest setting.
thermals are not the issue.
PSU has worked for this card for years now and is still able to provide ample power.
 
i'm at a loss for what else i should check before tearing the card apart again and checking components. i didnt find anything where i did the work (i swept with a magnet and sifted through anything that grabbed then went over it by hand.) so nothing is completely detached but may be loose for all i know. i just dont know what and where would cause the GPU to function like this. any thoughts on things i can use to diagnose the card be it app or possible things to try that dont involve another computer i dont got the spare hardware to slap together to see if its software on my PC somehow or not.
#1

21 Replies Related Threads

    safan80
    iCX Member
    • Total Posts : 322
    • Reward points : 0
    • Joined: 2006/12/06 20:30:58
    • Status: offline
    • Ribbons : 2
    Re: 2080ti performance cut in half 2024/02/18 18:46:21 (permalink)
    If you want to do an apples to apples comparison use the same drivers as you used before but use the DDU driver reinstall method before reinstalling the old drivers. If you will need to take the gpu off the waterblock and make sure everything is making proper contact, you have no choice really.
     
    The best place to get DDU is https://www.wagnardsoft.com/display-driver-uninstaller-ddu-
    Make sure you use DDU's method for rebooting into safe mode and deleting current installed drivers
     
    Make sure you go into setting first to enable rebooting into safer mode, creating a restore point and to NOT delete c:\Nvidia as that way you can go back to your old drivers without having to redownload them. If you deleted the install files for your old drivers download them first.
     
    please take note of what I wrote above and watch this video after to see how DDU works
    https://youtu.be/FPqUeU09cf8
    #2
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/18 20:56:08 (permalink)
    oh im not on the waterblock its back to the old air cooler. the water block was not correct and wouldnt have made contact will all the components, just most of them.
     
    i'll have to dig up the old drivers as you mentioned and try that again. i have a new issue though, i reseated the graphics card in its slot (just in case the slot got dust in it or otherwise) and now i cannot control the fan speeds. they have a strange minimum speed for the temp its sitting at and wont go all the way down or off no matter what i do in the overclock softwares fan curve settings.
     
    i havn't used DDU before but i've heard it mentioned often. after all this i may just start using it. i'm a if it aint broke dont update it kind of user. nvidia experience hadn't broken anything, so i just kept using it lol. but now it seems to have reached the end of its usefulness when it comes to drivers. i'll try this and something else i had thought of tomorrow after work.
    #3
    Sajin
    EVGA Forum Moderator
    • Total Posts : 49156
    • Reward points : 0
    • Joined: 2010/06/07 21:11:51
    • Location: Texas, USA.
    • Status: online
    • Ribbons : 199
    Re: 2080ti performance cut in half 2024/02/19 08:55:55 (permalink)
    109c on the hotspot isn’t right. Do another repaste.
    #4
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/19 09:08:45 (permalink)
    where is the hotspot sensor on the board? is it on the GPU core itself? why would the rest of it only reach 60C but the hotspot 109c? if you dont have an image a description of the location would suffice. like its by the transistors for power distribution etc.
    #5
    Cool GTX
    EVGA Forum Moderator
    • Total Posts : 30951
    • Reward points : 0
    • Joined: 2010/12/12 14:22:25
    • Location: Folding for the Greater Good
    • Status: offline
    • Ribbons : 122
    Re: 2080ti performance cut in half 2024/02/19 12:14:48 (permalink)
    Make sure you OC software is not running - turn it off & uncheck any box that starts the OC software on reboot - before you update your Nvidia driver
     
    RE: "new pads" &"new graphene? strip in place of goop"
     
    I've read reviews but never used a pad on a GPU.  What specific product did you use?
     
    What pads did you use?  If they are the wrong thickness or hardness you can have issues with proper contact between components & cooler
     
     

    Learn your way around the EVGA Forums, Rules & limits on new accounts Ultimate Self-Starter Thread For New Members

    I am a Volunteer Moderator - not an EVGA employee

    https://foldingathome.org -->become a citizen scientist and contribute your compute power to help fight global health threats

    RTX Project EVGA X99 FTWK Nibbler EVGA X99 Classified EVGA 3080Ti FTW3 Ultra


    #6
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/19 15:21:17 (permalink)
    i used the thermal grizzly pad. i watched a few channels that used em and stress tested them to see if the performance was there. the science is sound, its basically a reusable highly conductive pad because of the structure of the material. and i mean so far it works. but again when the card only draws 70% of the power it could, its not a challenge xD. but contact shouldnt be an issue. its all about getting the cooler tightened to the card. even standard goop was bleeding to the sides after my last repasting with almost nothing left on the GPU processor temps were, stable. on the high end but stable and easy to control. but yeah they were pads pre cut to the near exact size for a 20 series card. the one on the CPU works great so far. still fine tuning my voltage and and clock speeds vs temps on CPU though.
     
    as for what i've done as far as fixing the power draw, i did a clean install of video drivers after uninstalling all overclock software. this at least gave me back control of the fan speeds when i reinstalled my OC software. but i'm still seeing a low power draw for any kind of stress test of only 70%. to clarify for some, this isnt GPU usage, but power draw, i've given it a 130% limit an it doesnt go past 80% for more than a second. starting a stress test i'll see a spike of 112% but then it drops down to 70-80% range and my FPS tanks with it.
    #7
    Sajin
    EVGA Forum Moderator
    • Total Posts : 49156
    • Reward points : 0
    • Joined: 2010/06/07 21:11:51
    • Location: Texas, USA.
    • Status: online
    • Ribbons : 199
    Re: 2080ti performance cut in half 2024/02/19 15:31:23 (permalink)
    XiloTheOdd
    where is the hotspot sensor on the board? is it on the GPU core itself? why would the rest of it only reach 60C but the hotspot 109c? if you dont have an image a description of the location would suffice. like its by the transistors for power distribution etc.

    Yes, the hotspot is on the gpu die. Only reason I could think of the hotspot being that high would be a bad thermal paste job, or the die has degraded. I’m actually going though a hotspot battle of my own with my 4090.
    #8
    ty_ger07
    Insert Custom Title Here
    • Total Posts : 21167
    • Reward points : 0
    • Joined: 2008/04/10 23:48:15
    • Location: traveler
    • Status: offline
    • Ribbons : 270
    Re: 2080ti performance cut in half 2024/02/19 15:42:59 (permalink)
    Use thermal paste, not a thermal pad.  That should help a lot.
    Those thermal pads are good at spreading heat out laterally, but not good at passing heat through from one side to the other.  Their performance is mediocre.  They don't work good for high power applications.

    ASRock Z77 • Intel Core i7 3770K • EVGA GTX 1080 • Samsung 850 Pro • Seasonic PRIME 600W Titanium
    My EVGA Score: 1546 • Zero Associates Points • I don't shill

    #9
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/19 16:14:13 (permalink)
    well i had this kind of result with paste so it is why i decided to try something else. but that said, i've had this temp before and had twice the performance. its not that it sits at 109c for more than a second, thats just where it'll spike to.
     
    if its in the GPU die itself then im guessing it'd be dead center. but ill worry about that after i figure out why im only seeing 70% power draw. thats the anomaly here. odds are if it is a piece of hardware like a transistor i bumped at fault then i'll be opening it anyway. i just want to run through everything else it could be before getting to that again. i'll have time on the weekend to crack it back open and rethermal everything.
    #10
    Sajin
    EVGA Forum Moderator
    • Total Posts : 49156
    • Reward points : 0
    • Joined: 2010/06/07 21:11:51
    • Location: Texas, USA.
    • Status: online
    • Ribbons : 199
    Re: 2080ti performance cut in half 2024/02/19 17:42:18 (permalink)
    Post a link to your 14500 score and 7100 score.
    #11
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/21 10:10:12 (permalink)
    so a few things i've learned talking with EVGA support.
     
    the hotspot temp, its in the memory somewhere, not the GPU processor. so the thermal pads you use for those aren't making good contact
     
    and my power draw issue can quite possibly be a loose or removed component which i suspected from the beginning. i'll know more this thursday after work and will post my findings here once i confirm it. since most people have a history of asking a question, solving it, then not being anywhere to be found, then 3 years later people show up wondering.
     
    i dont got links to the 14500 score but it is the average score for the same exact GPU model. as for the 7100 its consistent. all clock speeds are normal, temps (aside from mem temp being a bit high) are well within operational parameters, power draw is the only outlier sitting at 70-80% draw under load. not much else the score tells us otherwise other than there is a problem.
    post edited by rjohnson11 - 2024/02/22 23:39:37
    #12
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/22 17:37:37 (permalink)
    so yes, it is looking like more and more the GPU memory temp is the likely culprit to my power draw being throttled. its crazy though i didnt even move the old pads off it, secured it tight, and its still hitting a high temp. the 109C used to simply be a spike but it has become a normal high for it. and as a result, rather than reducing the memory speed, its reducing the total power draw instead, this includes for the GPU processor that is keeping a very good temp the entire time. and what the fans base their speed off of.
     
    i'm gonna try setting the minimum temp for 100% fan speed lower so it just stays at 100% while gaming and see if its affecting it at all. if i can bring the mem temp down to 90C i should see my performance come back.
     
    edit:
    nope even with fans at 100%, mem clock speed reduced to 6000mhz from 7000mhz, it goes from 50C to 106C in less than a minute. but when i take the load away, it cools that 106C down to less than 50C in less than a minute as well.
     
    also 1 fan is running 400rpm slower than the other when set to 100%.
    post edited by XiloTheOdd - 2024/02/22 17:57:09
    #13
    Sajin
    EVGA Forum Moderator
    • Total Posts : 49156
    • Reward points : 0
    • Joined: 2010/06/07 21:11:51
    • Location: Texas, USA.
    • Status: online
    • Ribbons : 199
    Re: 2080ti performance cut in half 2024/02/22 19:56:42 (permalink)
    Gpu hotspot and vram memory temps are not the same thing.
    #14
    rjohnson11
    EVGA Forum Moderator
    • Total Posts : 102187
    • Reward points : 0
    • Joined: 2004/10/05 12:44:35
    • Location: Netherlands
    • Status: online
    • Ribbons : 84
    Re: 2080ti performance cut in half 2024/02/22 23:40:20 (permalink)
    A friendly reminder that cursing or swearing is not allowed on the EVGA forums even if disguised or removed by the forum filter. 

    AMD Ryzen 9 7950X,  Corsair Mp700 Pro M.2, 64GB Corsair Dominator Titanium DDR5  X670E Steel Legend, MSI RTX 4090 Associate Code: H5U80QBH6BH0AXF. I am NOT an employee of EVGA

    #15
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/23 13:26:07 (permalink)
    Sajin
    Gpu hotspot and vram memory temps are not the same thing.

    one HWM was referencing one metric as hotspot, precision 1X was referencing an identical temperature as the GPU memory temp. i'd assume an EVGA software knows more than a non-EVGA software in that regard.
     
    anyway. i've re padded the memory, used a video reference to see where most people put pads on their 2080ti cards. good contact everywhere, i'm still getting high temps. i'm not even sure anymore of what could be causing this, if the cooler was bad, the GPU would heat up as well as the memory, but the GPU stays fine while the Memory gets hot as hell, at least according to my cards metrics. i have seen a constant power draw closer to 90% at least, but without even touching it its back down to limiting closer to 70% because either temperature on the memory or some other problem.
     
    i guess the important question here at this point as part of diagnosis, at what temperature for the MEMORY does the card throttle power draw? where is the sensor for the memory temp on the board?
    #16
    ty_ger07
    Insert Custom Title Here
    • Total Posts : 21167
    • Reward points : 0
    • Joined: 2008/04/10 23:48:15
    • Location: traveler
    • Status: offline
    • Ribbons : 270
    Re: 2080ti performance cut in half 2024/02/23 13:57:15 (permalink)
    Did you put paste on the core this time?

    ASRock Z77 • Intel Core i7 3770K • EVGA GTX 1080 • Samsung 850 Pro • Seasonic PRIME 600W Titanium
    My EVGA Score: 1546 • Zero Associates Points • I don't shill

    #17
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/23 13:59:21 (permalink)
    so this post will be the marked answer in case anyone comes looking. the issue was indeed the memory temp on the card hitting a hard limit of 109C. i did a little "tweak" to intentionally make it get hotter to test this theory, and when it hit 109C the bench throttled it down to 70% power draw. with my recent bench test, it peaked at 107c and i was able to achieve a 130% power draw for the bench and regain the old score of 14200.
     
    as for improving this temp any further, its probably not possible with this stock cooler anymore, either its warped from age, working on the card and cleaning it, or something else i cant think of. so this card may make it another year, hopefully for the new gen to drop so i can nab one of those. lol.
     
    and incase anyone was wondering about what i was talking about using instead of paste for the GPU core it is this: Thermal Grizzly - Carbonaut (urls are not allowed so, search it up on amazon if your curious.)
     
    before all this, my GPU would hit the low 80-82C range, due to the paste blobbing off to the side and very little being left on the processor itself. this sheet, stays in one piece and the only tough part is making sure it doesnt move when you secure the cooler to the card. i've managed to shave about 9C off that temp under load for the processor. so its really effective as the videos i've seen pushing it pointed out. to be fair thats goin from bad contact paste to this so, results may vary there.
     
    anyway thank you all for lending me some brains on this. since EVGA isn't in the market for making cards anymore i've been trying to make this one last and keep up with newer generations.
     
    edit: apparently i can't mark my own post as an answer...
     
    ^^^this is the ANSWER^^^
     
    for lack of a way to do it. unless a moderator can override.
    post edited by XiloTheOdd - 2024/02/23 14:04:15
    #18
    ty_ger07
    Insert Custom Title Here
    • Total Posts : 21167
    • Reward points : 0
    • Joined: 2008/04/10 23:48:15
    • Location: traveler
    • Status: offline
    • Ribbons : 270
    Re: 2080ti performance cut in half 2024/02/23 18:51:35 (permalink)
    I think paste on the core would cool the core better (in fact, short-term, I am sure without a doubt), and improve your contact issue with the memory thermal pads.  If you want to use a thermal pad on the core, I would suggest increasing the thermal pad thickness on the memory one size thicker from whatever you used in order to compensate.  Since you refuse to paste the core, you will need to adjust as necessary.  I guarantee you that your answer is NOT 'there is nothing I can do'.

    ASRock Z77 • Intel Core i7 3770K • EVGA GTX 1080 • Samsung 850 Pro • Seasonic PRIME 600W Titanium
    My EVGA Score: 1546 • Zero Associates Points • I don't shill

    #19
    XiloTheOdd
    New Member
    • Total Posts : 10
    • Reward points : 0
    • Joined: 2024/02/18 17:43:25
    • Status: offline
    • Ribbons : 0
    Re: 2080ti performance cut in half 2024/02/23 19:57:05 (permalink)
    ty_ger07
    I think paste on the core would cool the core better (in fact, short-term, I am sure without a doubt), and improve your contact issue with the memory thermal pads.  If you want to use a thermal pad on the core, I would suggest increasing the thermal pad thickness on the memory one size thicker from whatever you used in order to compensate.  Since you refuse to paste the core, you will need to adjust as necessary.  I guarantee you that your answer is NOT 'there is nothing I can do'.


    again i've pasted the core, with 2 different kinds of paste and the result is the same, my temps are higher because over time of it heating and cooling down it liquefies and boils over to the edge, leaving nothing more than a smudge on the contact area for the processor. this carbon pad has shown the better results in that regard. not sure why your not trusting of that. i'm looking at better numbers for the processor than i've ever looked at since before i started this for that portion of it.
     
    as for the memory, how thick we talkin for pads? i'm using 2mm which is what this cooler came with, got it all tightened down and only getting marginally better numbers. thankfully enough to have solved the issue of why power was being throttled. should i be stuffing it till it barely screws together? should i buy 3-4 different sizes of pad to figure out which one is thick enough? that makes no sense unless components of the GPU are shrinking with age, the old pads i could understand maybe losing contact over time, but they only lost contact after the 2nd time i pulled this GPU apart and then it was being difficult about getting proper contact again. so barring getting super fancy with some high precision calipers i dont have, these 2mm pads are making full contact, the only way they could make more contact is if i could wrap the whole memory chip in em. or have a cooler with a completely different design to make more direct contact with the memory than to put a metal plate in between the cooler and the PCB for memory and power delivery components. which, if i had a water block for the right model, would be the case. sadly i do not.
     
    so the answer is, there is nothing i can do, without spending an inordinate amount of cash for a chance to get things to drop 10C at the most because products for 2080ti are no longer in production. anything i find will be second hand or "refurbished" like the incorrect block i got that has clear signs of usage. its not worth the money to risk getting dead components to RMA back till i find one that is good. so, i'm working with what i got. this is probably the best it'll get. the 50XX series is right around the corner. i'll grab one of those with a waterblock already to go on it asap and that'll be the end of this 2080ti and it'll be a backup if something happens.
    #20
    ty_ger07
    Insert Custom Title Here
    • Total Posts : 21167
    • Reward points : 0
    • Joined: 2008/04/10 23:48:15
    • Location: traveler
    • Status: offline
    • Ribbons : 270
    Re: 2080ti performance cut in half 2024/02/23 20:16:29 (permalink)
    Again? "Again" you pasted the core twice?  "Again"?  You never said it before.
     
    You said you tried an incompatible water block, went back to an air cooler, installed a thermal pad in the process, and the temps aren't right.
     
     
    1) The best, thinnest, most perfect thermal pad that exists performs worse than thermal paste.
    2) All thermal pads are thicker than paste and everything else needs to be adjusted accordingly to achieve proper contact pressure.
    3) No one has any idea which thermal pads you used, whether they are appropriate thickness, and how good their contact pressure is.
     
    All we know for sure is that something isn't right and the solution, definitely, without a doubt, isn't 'this is as good as it gets'.
    post edited by ty_ger07 - 2024/02/25 15:33:33

    ASRock Z77 • Intel Core i7 3770K • EVGA GTX 1080 • Samsung 850 Pro • Seasonic PRIME 600W Titanium
    My EVGA Score: 1546 • Zero Associates Points • I don't shill

    #21
    Q56_Monster
    CLASSIFIED Member
    • Total Posts : 3881
    • Reward points : 0
    • Joined: 2007/11/02 11:03:10
    • Location: Philly 'Burbs
    • Status: offline
    • Ribbons : 5
    Re: 2080ti performance cut in half 2024/02/27 13:36:09 (permalink)
    Xilo, pastes don't "boil over".  You have a lot of great advice here in the thread, but as long as you're happy.

    Mobo: Asus Rampage VI Extreme CPU: i9-7980XE
    GPU: RTX 3090 KPE
    HD: Samsung 1TB 970 Pro NVMe M.2
    PSU: EVGA Supernova 1600 T2 OS: All of them
    Monitor: 48" LG CX OLED  Other Cooling Fun: F1 Gemini/Tek-9 Fat LN2 | 12k BTU A/C to Chiller Conversion 



    #22
    Jump to:
  • Back to Mobile