EVGA

Hot!Benchmarking FTW3 RTX3090...?

Author
Jsunn
iCX Member
  • Total Posts : 276
  • Reward points : 0
  • Joined: 2007/05/06 17:01:05
  • Status: offline
  • Ribbons : 2
2021/02/28 18:45:24 (permalink)
Hello All, 
 
I am getting ready to install the Optimus Absolute Water block on my EVGA RTX 3090 FTW3 Ultra Gaming card. 
I wanted to get some good baselines for comparison before and after the block is installed for temperatures and clocks. 
 
I was using HWInfo to log the temps and clocks for port Royal runs, but I wasn't quite sure what values were really important.  
The chart is very busy, and I'm not a pro like Jayztwocents or Steve from Gamersnexus. 
 
Here is a basic chart I put together as an example, so wondering if I can reduce the items I am tracking, maybe limit it to GPU temp, Mem temp and GPU clock and Mem clock and skip all of the other temp values?
 
I want to get a good before and after I install the water block, but also want to make sure I capture the data that is needed. 
 
Here is an example chart from a Port Royal run. Temp scale is on the left, Clock Speeds are on the right.

Thanks everyone!
-J

Attached Image(s)


i7 8700K @ 5GHz 1.35v 
32Gb 3600MHz
Asus Maximus Formula X 
EVGA RTX 3090 FTW3 
Watercool all the things
https://builds.gg/builds/...ustom-built-case-26023
#1

16 Replies Related Threads

    arestavo
    CLASSIFIED ULTRA Member
    • Total Posts : 6447
    • Reward points : 0
    • Joined: 2008/02/06 06:58:57
    • Location: Through the Scary Door
    • Status: offline
    • Ribbons : 56
    Re: Benchmarking FTW3 RTX3090...? 2021/02/28 19:02:35 (permalink)
    GPU2 is on the back of the card, along with Mem1 IIRC - so those might not have as big a reduction (without an actively cooled backplate) in temp as GPU1 and the other VRAM temps.
     
    Overall your graph looks good to me.

    EVGA affiliate code: 9ZWDWFNW6A
    (Don't forget to upload your invoice or no credit is given!)
     
    FOLD ON
    #2
    Dabadger84
    CLASSIFIED Member
    • Total Posts : 2810
    • Reward points : 0
    • Joined: 2018/05/11 23:49:52
    • Location: de_Overpass, USA
    • Status: online
    • Ribbons : 9
    Re: Benchmarking FTW3 RTX3090...? 2021/02/28 19:23:58 (permalink)
    Just a point of focus, if you want solid before & after temps, you should do a full Port Royal Stress Test (it's 20 loops of the test, takes around 36 minutes).  This will usually result in temperatures leveling off and "reaching steady state", giving you an idea of at least close to what your gaming temperatures would be.
     
    A single run of PR can really give you some misleading numbers, an example:
     
    on my card, after some heatsink modifications on the backplate etc, a single run of PR resulted in these temps (maximum/peak is on the right, idle in the middle) :

    And here's a full Stress Test run with the same setup/clocks etc:

     
    I would recommend if you want to do a full comparison of before and after, to do a test that's at least 30 minutes of solid load so that you can hit steady state & really get an idea of just how much that block lowers your temperatures.
     
    From what I've seen, high quality full coverage blocks like the one you're getting typically knock the memory temps down in to the 40s under full load.

    Save up to 10% on eVGA.com purchases, use my Associates Code: 47AQQ6KPU2IZNXY
    My system on ModRigs: https://www.modsrigs.com/detail.aspx?BuildID=42686 
    System Specs:
    5950x @ Stock (for now) - Asus Crosshair VIII Hero - eVGA 1200W P2 - 2x16GB G.Skill Trident Z Royal @ 3600MHz CL14
    eVGA RTX 3090 K|NGP|N :D - Samsung Odyssey G9 49" Super-Ultrawide 240Hz Monitor

    #3
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/01 03:46:56 (permalink)
    Dabadger84
    Just a point of focus, if you want solid before & after temps, you should do a full Port Royal Stress Test (it's 20 loops of the test, takes around 36 minutes).  This will usually result in temperatures leveling off and "reaching steady state", giving you an idea of at least close to what your gaming temperatures would be.
     
    A single run of PR can really give you some misleading numbers, an example:
     
    on my card, after some heatsink modifications on the backplate etc, a single run of PR resulted in these temps (maximum/peak is on the right, idle in the middle) :

    And here's a full Stress Test run with the same setup/clocks etc:

     
    I would recommend if you want to do a full comparison of before and after, to do a test that's at least 30 minutes of solid load so that you can hit steady state & really get an idea of just how much that block lowers your temperatures.
     
    From what I've seen, high quality full coverage blocks like the one you're getting typically knock the memory temps down in to the 40s under full load.


    This is an excellent point. I will rerun the stress test and see what the plot looks like.
    My thought is that with all of the different temps that the card outputs, there may be too much on the chart to look at. Is there some key temps I can plot to clean up the chart?

    For example, if GPU clock only depends on the GPU temperature, then maybe I don't need to show all of the others?

    Thoughts?

    Thank you,
    Jason

    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #4
    Dabadger84
    CLASSIFIED Member
    • Total Posts : 2810
    • Reward points : 0
    • Joined: 2018/05/11 23:49:52
    • Location: de_Overpass, USA
    • Status: online
    • Ribbons : 9
    Re: Benchmarking FTW3 RTX3090...? 2021/03/01 10:30:58 (permalink)
    Jsunn
    Dabadger84
    Just a point of focus, if you want solid before & after temps, you should do a full Port Royal Stress Test (it's 20 loops of the test, takes around 36 minutes).  This will usually result in temperatures leveling off and "reaching steady state", giving you an idea of at least close to what your gaming temperatures would be.
     
    A single run of PR can really give you some misleading numbers, an example:
     
    on my card, after some heatsink modifications on the backplate etc, a single run of PR resulted in these temps (maximum/peak is on the right, idle in the middle) :
    snip
    And here's a full Stress Test run with the same setup/clocks etc:
    snip
     
    I would recommend if you want to do a full comparison of before and after, to do a test that's at least 30 minutes of solid load so that you can hit steady state & really get an idea of just how much that block lowers your temperatures.
     
    From what I've seen, high quality full coverage blocks like the one you're getting typically knock the memory temps down in to the 40s under full load.


    This is an excellent point. I will rerun the stress test and see what the plot looks like.
    My thought is that with all of the different temps that the card outputs, there may be too much on the chart to look at. Is there some key temps I can plot to clean up the chart?

    For example, if GPU clock only depends on the GPU temperature, then maybe I don't need to show all of the others?

    Thoughts?

    Thank you,
    Jason



    Maybe only plot the hottest of each temperature I.E. whichever Memory iCX is running hottest - or only do Hottest & coolest.  So GPU, GPU2, 2 Mems, 2 PWRs.  It would clean it up a bit.  
    The hottest sensors on each card type is different, 3080 XC3 has different warmest temps from the 3080 FTW3, 3090 FTW3 Hybrid has different hottest sensors from the Kingpin, oddly enough as well.  My warmest VRM is always PWR3 on the Kingpin, when I had an FTW3 Hybrid, the warmest VRM was always PWR4.  
    You could also color code it a bit better, have Memory Temps all be shades of green, PWR/VRM shades of red, for example, so it's easier to visualize "memory temps are all in this range, VRM temps are all in this range" type information.  I would definitely do Memory Junction Temp separately, because it's whichever chip is running hottest internally, from what I understand, which means it can move around.

    Save up to 10% on eVGA.com purchases, use my Associates Code: 47AQQ6KPU2IZNXY
    My system on ModRigs: https://www.modsrigs.com/detail.aspx?BuildID=42686 
    System Specs:
    5950x @ Stock (for now) - Asus Crosshair VIII Hero - eVGA 1200W P2 - 2x16GB G.Skill Trident Z Royal @ 3600MHz CL14
    eVGA RTX 3090 K|NGP|N :D - Samsung Odyssey G9 49" Super-Ultrawide 240Hz Monitor

    #5
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/01 18:07:14 (permalink)
    Take a look here, I took the advice from above, this is a port royal stress test, I only included a few parameters. 
    Let me know what you think, is this helpful? 


    -J
    post edited by Jsunn - 2021/03/01 19:02:52

    Attached Image(s)


    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #6
    Dabadger84
    CLASSIFIED Member
    • Total Posts : 2810
    • Reward points : 0
    • Joined: 2018/05/11 23:49:52
    • Location: de_Overpass, USA
    • Status: online
    • Ribbons : 9
    Re: Benchmarking FTW3 RTX3090...? 2021/03/01 20:59:24 (permalink)
    How are you measuring liquid temp?  Just curious, I'm interested in getting something that can do that as well, purely out of the "what is it" aspect of the liquid temp, especially since I'm eventually going to have both my CPU & GPU in the loop. - See in the picture it's an Asus thing, is that part of an AIO or separate component?
     
    I'd say it's odd your temps seemed to level off then started going up, but with your ambient increasing (assumably because of the heat the radiators & card were putting out) it makes sense that they started going up again.
     
    Nice to see I'm right about PWR4 being the warmest VRM sensor on 3090 FTW3s though, it was on my Hybrid version as well.
     
    To clean up the main chart, you could remove most of the stuff, keep the warmest VRM/PWR, warmest Mem temp, GPU temp, the rest aren't really needed, that would declutter it quite a bit, however with Mem1 & 2 being so close, it'll be interesting to see which one goes down more once you have the block on.  My guess is whichever one of those is the backplate-side memory will be the one that runs the hotter of the two after the block, of course.
     
    Temps are going to be a lot different and less varied once you get the full coverage block on - I've seen a few people reporting their warmest temps only being mostly in the low 40s with quality blocks on, with the GPU core running even cooler.
     
    GPU clock is almost wholly determined by GPU core/die temp, but, with how tightly packed the VRAM is around the GPU Die, that & the PCB heatsoak they cause (because they're stacked on either side of the PCB, right on top of each other, thusly heating the entire PCB in that area), I would wager that cooling the Memory as best you can would further help lower GPU Die temps, which would then again result in higher sustained clocks, because of less down-binning due to temperature stepping that's baked in to Ampere and other recent gen NVidia GPUs.
     
    How do you make those charts, simply outputting the information in to a spreadsheet program then having it make a chart, or something more fancy?  Kinda wanna do this to better illustrate my "before & after" for when I finally settle on a "Final" configuration for my backplate cooling fancyness... well, it's not fancy anymore, it's a mess, cuz Thermal Paste, but it sure is working well :-D

    Save up to 10% on eVGA.com purchases, use my Associates Code: 47AQQ6KPU2IZNXY
    My system on ModRigs: https://www.modsrigs.com/detail.aspx?BuildID=42686 
    System Specs:
    5950x @ Stock (for now) - Asus Crosshair VIII Hero - eVGA 1200W P2 - 2x16GB G.Skill Trident Z Royal @ 3600MHz CL14
    eVGA RTX 3090 K|NGP|N :D - Samsung Odyssey G9 49" Super-Ultrawide 240Hz Monitor

    #7
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/02 18:05:44 (permalink)
    Dabadger84
    How are you measuring liquid temp?  
    How do you make those charts, simply outputting the information in to a spreadsheet program then having it make a chart, or something more fancy?  Kinda wanna do this to better illustrate my "before & after" for when I finally settle on a "Final" configuration for my backplate cooling fancyness... well, it's not fancy anymore, it's a mess, cuz Thermal Paste, but it sure is working well :-D



    Hi, 
    My motherboard has two headers for temperature, they are marked as "Water In" and "Water Out", but I have one monitoring my liquid temp and the other I have a standard 10kOhm temp sensor just sticking out measuring the air temp around my PC. 
    The water temp really isn't relevant now as I only have my CPU being cooled, but it will be interesting to see how it goes up once I have my GPU added to the loop.
     
    I am just using MS Excel to plot the charts. It take a bit of tweaking to get them set up, but seems to work pretty well. 
     
    And post pics of what you are doing with your backplate, I am putting on the Optimus backplate that feels like it is a 3lb block of aluminum. 
     
    Take care, 
    J
     

    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #8
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/02 18:11:47 (permalink)
    Hello Again, 
    Here is what I came up with. I will use this data as a first run at stock settings. The O/C switch on the card is swithed, so the fans are always spinning. 
     
    The charts show all of the data, my GPU is air cooled and my case is a custom open air design. I am in a cool basement so it makes for good temps. 
    You can see in both charts where my furnace kicked on as a small bump in all of the temps. 
    This was a 35 min run of Port Royal Stress Test. 
     
    I am going to run it again with a small overclock next and see how the temps and clocks compare. 
    Let me know if you have any suggestions. I have a new respect for the guys that do this all of the time, this stuff is tedious!
     
    ***Typo on the chart, the box with the "Max Ambient Air Temp" should say, "Max Water Temp" not, "MAX GPU Mem Junct".***
    -J

     
    post edited by Jsunn - 2021/03/03 16:30:46

    Attached Image(s)


    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #9
    Dabadger84
    CLASSIFIED Member
    • Total Posts : 2810
    • Reward points : 0
    • Joined: 2018/05/11 23:49:52
    • Location: de_Overpass, USA
    • Status: online
    • Ribbons : 9
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 04:49:26 (permalink)
    There's an enxtensive amount of pictures on the backplate modding I'm doing in the thread I have going for it: https://forums.evga.com/3090-Kingpin-amp-others-Get-that-backside-VRAM-cool-m3219773.aspx 
    It's not for the feint of heart, especially with the point I'm at now where there's thermal paste involved :-D
     
    From your memory junction temp I would say getting some pads & heatsinks on your backplate area will definitely help though.  Bit more complicated on the 3090 FTW3 than it is on the Kingpin from what I've seen, but the overall idea is the same, if you get a custom backplate or just a solid piece backplate... the stock eVGA 3090 FTW3 backplate has too many holes so it actually wouldn't work well at all for the heatsink idea.  Too much uneven pressure etc.

    Save up to 10% on eVGA.com purchases, use my Associates Code: 47AQQ6KPU2IZNXY
    My system on ModRigs: https://www.modsrigs.com/detail.aspx?BuildID=42686 
    System Specs:
    5950x @ Stock (for now) - Asus Crosshair VIII Hero - eVGA 1200W P2 - 2x16GB G.Skill Trident Z Royal @ 3600MHz CL14
    eVGA RTX 3090 K|NGP|N :D - Samsung Odyssey G9 49" Super-Ultrawide 240Hz Monitor

    #10
    kevinc313
    CLASSIFIED Member
    • Total Posts : 2662
    • Reward points : 0
    • Joined: 2019/02/28 09:27:55
    • Status: offline
    • Ribbons : 16
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 06:21:59 (permalink)
    Excellent work, great graphs.
    #11
    phroze
    iCX Member
    • Total Posts : 390
    • Reward points : 0
    • Joined: 2018/09/17 20:09:17
    • Location: WA State
    • Status: offline
    • Ribbons : 0
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 11:34:35 (permalink)
    Just as an FYI, I use this for temp monitoring. Just placed it on my outlet of my distro block. Works great.
     
    https://www.amazon.com/gp...itle?ie=UTF8&psc=1

    Case: Lian Li O11 Dynamic XL
    Mobo: Asrock X570 Taichi
    CPU: Ryzen 5900x
    GPU: EVGA RTX 3090 FTW3 Ultra
    RAM: Crucial Ballistix OC to 3800 16 18 18 1:1
    PSU: EVGA SuperNova G2 1600w
    Cooling: Custom hardline loop: optimus block, ek, primochill, lian li, HW Labs, custom stuff
    #12
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 12:30:57 (permalink)
    Hello All, 
     
    I have completed my data collection prior to installing my water block. 
    My last run was a couple of small overclocks. I've really only been able to OC my board at 200 MHz on the memory and 120 Mhz on the core. 
    I can go a little higher, but I will get an occasional crash. 
     
    I am more than happy with the performance, and I am hoping that I can squeeze a little more out of the card once I get the water. block installed. 
    I'll get another thread for that install!
    Let me know if you want to see any other info.
     



     
    post edited by Jsunn - 2021/03/03 12:34:09

    Attached Image(s)


    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #13
    Delirious
    EVGA Forum Moderator
    • Total Posts : 14669
    • Reward points : 0
    • Joined: 2007/11/15 13:34:04
    • Location: at my computer
    • Status: offline
    • Ribbons : 60
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 12:46:26 (permalink)
    Thanks for posting
     

    "Be quick to listen, slow to speak and slow to anger" 
    Affiliate Code XZUMV9TJW5
    education may be expensive but wait until you get the bill for ignorance
    A wise man once said that we can't make anyone feel or do anything. We can throw things into the wind, but it's up to each person to decide how they want to react, where they want to stand when things fall.
    #14
    gvarv
    New Member
    • Total Posts : 91
    • Reward points : 0
    • Joined: 2020/12/20 18:04:00
    • Status: offline
    • Ribbons : 0
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 15:47:17 (permalink)
    phroze
    Just as an FYI, I use this for temp monitoring. Just placed it on my outlet of my distro block. Works great.
     
    https://www.amazon.com/gp...itle?ie=UTF8&psc=1




    Thanks! Might have to get myself one of those when I expand my loop to include my GPU.
     

    NZXT Phantom
    Crosshair VI
    Ryzen 3900X
    3090fe
    32 GB RAM DR
    Asus MG279
    #15
    gvarv
    New Member
    • Total Posts : 91
    • Reward points : 0
    • Joined: 2020/12/20 18:04:00
    • Status: offline
    • Ribbons : 0
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 15:49:18 (permalink)
    Jsunn
    Hello All, 
     
    I have completed my data collection prior to installing my water block. 
    My last run was a couple of small overclocks. I've really only been able to OC my board at 200 MHz on the memory and 120 Mhz on the core. 
    I can go a little higher, but I will get an occasional crash. 
     
    I am more than happy with the performance, and I am hoping that I can squeeze a little more out of the card once I get the water. block installed. 
    I'll get another thread for that install!
    Let me know if you want to see any other info.
     




    I probably missed where you described the fan settings - are they stock? Custom? Set to a particular RPM?
     

    NZXT Phantom
    Crosshair VI
    Ryzen 3900X
    3090fe
    32 GB RAM DR
    Asus MG279
    #16
    Jsunn
    iCX Member
    • Total Posts : 276
    • Reward points : 0
    • Joined: 2007/05/06 17:01:05
    • Status: offline
    • Ribbons : 2
    Re: Benchmarking FTW3 RTX3090...? 2021/03/03 16:28:10 (permalink)
    gvarv
    Jsunn
    Hello All, 
     
    I have completed my data collection prior to installing my water block. 
    My last run was a couple of small overclocks. I've really only been able to OC my board at 200 MHz on the memory and 120 Mhz on the core. 
    I can go a little higher, but I will get an occasional crash. 
     
    I am more than happy with the performance, and I am hoping that I can squeeze a little more out of the card once I get the water. block installed. 
    I'll get another thread for that install!
    Let me know if you want to see any other info.
     




    I probably missed where you described the fan settings - are they stock? Custom? Set to a particular RPM?
     




    Hi,
    Yeah, sorry, I forgot to post the fan settings. 
     
    I have the BIOS switch in the "OC" mode. Other than that all of my fan settings are using the factory curves. I don't have any other fan settings changed. I kind of default to more silent operation as my open air case is right next to where I sit. 
     
    Let me know if you have any other questions.
    -J 

    i7 8700K @ 5GHz 1.35v 
    32Gb 3600MHz
    Asus Maximus Formula X 
    EVGA RTX 3090 FTW3 
    Watercool all the things
    https://builds.gg/builds/...ustom-built-case-26023
    #17
    Jump to:
  • Back to Mobile