Overclock.net - An Overclocking Community - Reply to Topic
Thread: [Official] NVIDIA RTX 2080 Ti Owner's Club Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
04-05-2020 09:44 AM
sultanofswing
Quote: Originally Posted by Imprezzion View Post
Well balls. It really did not like that.. card was running fine 2085 @ 1.093 on 67c ish when my PC just turned off completely and restarted.

Didn't do any damage to the card ofcourse, heck it'll take a lot more then that to kill a card in such a short run, but it tripped a protection somewhere and judging by the loud relais click on shutdown it was my PSU.

So, are you kidding me. A full Seasonic Focus Plus Gold 750w can't keep up with a 5.1Ghz 9900K which at best pulls 240w in a stress test and a 380w 2080 Ti? I mean, I have 12 fans and a LOT of RGB hooked up to it but still... it's a 750w ffs. I might try to connect the card to 2 separate PCI-E leads as it's on a single split 8+8 cable now.

And also, on the FTW3 (Ultra) BIOS it also does the exact same.. Power draw and FPS is way lower on just a very high curve on offset 0 compared to offset +150 and then 1 bin up in the curve.
Difference is like, a LOT.

I have 2 profiles next to eachother with the exact same clocks, just a different curve. You can see the difference in power limit and such as I apply each of them and MSI AB doesn't log my FPS here in the test i was running but it went from 103 fps to 111 fps and back switching between the profiles.

So yeah, curve with offset is MUCH better but also puts WAY more load on the poor thing and it cannot run the same clocks for me. 2070Mhz is nowhere near stable at the "higher load" of the better curve. It barely runs 10 minutes without throwing a crash to desktop or a DirectX error. Well, it never gets boring like this in these isolation times!
It's your power supply, The Focus series unless you have the revised version had a well known issue with Higher wattage cards. The transient response of the card boosting would cause the power supplies to freak out and just shut off.
04-05-2020 09:18 AM
J7SC At the end of the day, for gaming with a 2080 Ti, I doubt many folks would notice the extra MHz (say above 2070) unless they have a FPS overlay.

For the record, when I first got the GB Aorus XTR WB cards, each was tested with an AIO on a Z170 test bench, and they would still run min of 2190 MHz and 2145 MHz respectively, even at 46 C (w/23 C ambient). Since then, a much more extensive GPU loop for both cards typically keeps the cards well below 40 C even for hours. As mentioned before, I rarely touch the voltage slider, still run the stock Bios (max 375w +) and never got around to trying the voltage curve. FYI, I just posted related 'collage' graphs in another thread here https://www.overclock.net/forum/61-w...l#post28396312
04-05-2020 08:33 AM
Mooncheese
Quote: Originally Posted by Imprezzion View Post
It's not actually possible for us to edit the BIOS directly due to new checksum and encryption on these BIOS files compared to GTX9xx and older.
Maybe virtualize a Windows in Linux and see if a OC applied in the virtual windows will stick on Linux as well? I really really see that as a 2% chance of working but if we don't try random stuff we'll never know

BTW for you other guys on water / chilled. Did you notice a big difference between what clock was stable at what voltage compared to air and higher temperatures?
I mean, i've just been doing so much testing with the Curve discussion from previous posts and also just testing my card in general and i really seem to hit a pretty hard wall at 2070Mhz 1.068v. I mean, it's "hot" at 66-67c load but it is rock solid stable. Wierd thing is, even 2100Mhz @ 1.093v just will not even run 1 minute of Borderlands 3 at 67c and crashes all over the place but when it's cold, say idle temps 27c, and it warms up when just starting a game it runs 2115Mhz just fine untill it drops to 2100-2085 and eventually 2070Mhz when it gets up to temperature.
Nope, my card also exhibits a hard wall at 2100 MHz above 40C. Like it will do 2100 MHz and pass all benchmarks, presumably because the core starts out at ~27-31C and typically doesn't get hotter than say 40-42C whilst benching (Superposition will get it up to 43-44C because it's nearly at 340W the entire time) but once over 40C, it doesn't matter if I do +100 on the voltage slider in MSI AB, it's going to CTD + Display Driver Failure, guaranteed. So I'm also seeing a hard-wall at 2100 MHz with my 300a. The consolation is that my card will do 2070 MHz @ 1.013v which is fantastic because bringing the voltage down drops wattage consumption quite a bit. The difference for my card is substantial, much more than running the card at 2100 MHz with as much voltage as I can pump into it. In Firestrike for example, simply turning up core freq to +120 MHz and +500 MHz on the memory it did 37,800 Graphics (2070 MHz core actual) @ 1.069v throughout the duration of the test (going by OSD). Now, dialing it in on the freq / voltage curve so that it does 2070 MHz @ 1.013v it does 38,800 Graphics. Closely monitoring the wattage consumption while trying to bear out Sultanofswing's voltage starvation hypothesis (which works on his card because he has KP) with +100 on voltage slider and 2070 MHz core in Timespy Demo it was pulling on average 30-40W more than with the undervolt, and the clocks were dipping way harder and more frequently.

Anyhow, yeah 2100 MHz, seems you need an above average core and clean power delivery to attain that or more.

That said, my clocks of 2070 MHz @ 1.013v are solid, I just played for around 3-4 hours yesterday total without a crash.

Here's 42 minutes of Ghost Recon Breakpoint "bug testing" (radial menu disappears at end and in the beginning I try in vain to present a "press a to open parachute prompt":

04-05-2020 07:39 AM
Imprezzion Well balls. It really did not like that.. card was running fine 2085 @ 1.093 on 67c ish when my PC just turned off completely and restarted.

Didn't do any damage to the card ofcourse, heck it'll take a lot more then that to kill a card in such a short run, but it tripped a protection somewhere and judging by the loud relais click on shutdown it was my PSU.

So, are you kidding me. A full Seasonic Focus Plus Gold 750w can't keep up with a 5.1Ghz 9900K which at best pulls 240w in a stress test and a 380w 2080 Ti? I mean, I have 12 fans and a LOT of RGB hooked up to it but still... it's a 750w ffs. I might try to connect the card to 2 separate PCI-E leads as it's on a single split 8+8 cable now.

And also, on the FTW3 (Ultra) BIOS it also does the exact same.. Power draw and FPS is way lower on just a very high curve on offset 0 compared to offset +150 and then 1 bin up in the curve.
Difference is like, a LOT.

I have 2 profiles next to eachother with the exact same clocks, just a different curve. You can see the difference in power limit and such as I apply each of them and MSI AB doesn't log my FPS here in the test i was running but it went from 103 fps to 111 fps and back switching between the profiles.

So yeah, curve with offset is MUCH better but also puts WAY more load on the poor thing and it cannot run the same clocks for me. 2070Mhz is nowhere near stable at the "higher load" of the better curve. It barely runs 10 minutes without throwing a crash to desktop or a DirectX error. Well, it never gets boring like this in these isolation times!
04-05-2020 06:59 AM
freezer2k
Quote: Originally Posted by freezer2k View Post
Hi,

I have the EVGA GeForce RTX 2080 Super Black Gaming 8 GB (08G-P4-3081-KR).

I want to do [email protected] using Linux Ubuntu 19.10 or soon 20.04.

The whole folding part I already got working. But I really want to undervolt them to keep power costs and temps low. This worked wonderfully using Voltage/Frequency Curve in MSI Afterburner under Win10.
I assume theres nothing like this available in Linux?

Can I flash a custom VF-Curve BIOS on this card?

Thanks
I found this for Linux:
https://www.reddit.com/r/nvidia/comm...king_in_linux/

So setting a custom Power Limit. Still using the default VF-Curve, but limiting power? I guess it will run at lower frequencies then?

Wonder how the power/performance ratio is compared to MSI Afterburner OC-Curve and the lock-in of a certain Voltage+Frequency.
04-05-2020 06:42 AM
Imprezzion
Quote: Originally Posted by freezer2k View Post
Hi,

I have the EVGA GeForce RTX 2080 Super Black Gaming 8 GB (08G-P4-3081-KR).

I want to do [email protected] using Linux Ubuntu 19.10 or soon 20.04.

The whole folding part I already got working. But I really want to undervolt them to keep power costs and temps low. This worked wonderfully using Voltage/Frequency Curve in MSI Afterburner under Win10.
I assume theres nothing like this available in Linux?

Can I flash a custom VF-Curve BIOS on this card?

Thanks
It's not actually possible for us to edit the BIOS directly due to new checksum and encryption on these BIOS files compared to GTX9xx and older.
Maybe virtualize a Windows in Linux and see if a OC applied in the virtual windows will stick on Linux as well? I really really see that as a 2% chance of working but if we don't try random stuff we'll never know

BTW for you other guys on water / chilled. Did you notice a big difference between what clock was stable at what voltage compared to air and higher temperatures?
I mean, i've just been doing so much testing with the Curve discussion from previous posts and also just testing my card in general and i really seem to hit a pretty hard wall at 2070Mhz 1.068v. I mean, it's "hot" at 66-67c load but it is rock solid stable. Wierd thing is, even 2100Mhz @ 1.093v just will not even run 1 minute of Borderlands 3 at 67c and crashes all over the place but when it's cold, say idle temps 27c, and it warms up when just starting a game it runs 2115Mhz just fine untill it drops to 2100-2085 and eventually 2070Mhz when it gets up to temperature.
04-05-2020 06:26 AM
freezer2k Hi,

I have the EVGA GeForce RTX 2080 Super Black Gaming 8 GB (08G-P4-3081-KR).

I want to do [email protected] using Linux Ubuntu 19.10 or soon 20.04.

The whole folding part I already got working. But I really want to undervolt them to keep power costs and temps low. This worked wonderfully using Voltage/Frequency Curve in MSI Afterburner under Win10.
I assume theres nothing like this available in Linux?

Can I flash a custom VF-Curve BIOS on this card?

Thanks
04-05-2020 02:20 AM
Imprezzion
Quote: Originally Posted by EarlZ View Post
The cooler on that card is on a different level compared to the Gigabyte Gaming OC, its about 20-25c better.

I decided to do a repaste and it was worth the effort. The TIM Gigabyte used was already dried up and flaked. Used the Noctua paste that I had around and my temps are abou 11-12 lower on MAX load and it bearly breaks 74c with the division 2 now.
The Gainward Phoenix GS I have also had terrible paste. The paste itself was pretty good and gooey actually and not dried up but it had quite some airpockets in it and not the best coverage at all. Replaced it with Prolimatech PK-3 and it dropped temps like a brick.


Now I'm waiting for my X52 + G12 to show up to see how low I can get it to go with a AIO. Should I liquid metal it or?
I'm now on 2070Mhz 1.068v with EVGA FTW3 BIOS and it was around 70c on 100% fanspeed first in games like Division 2. Now it's on about 65c on 70% fanspeed.

Oh by the way, @86Jarrod and @sultanofswing this totally works and i can comfirm this completely with real-world tests.

I just use a very very simple performance test imo. Borderlands 3 on Badass 1080P 150% resolution scale standing still at the exact point the game launches in Sanctuary. Very repeatable as you always spawn exactly the same location and facing the exact same direction.

2100Mhz @ 1.093v with 0Mhz slider and just slamming the Curve up to 1.093v @ 2100 results in 113 FPS.
2100Mhz @ 1.093v with +200Mhz (2055 @ 1.068mv) slider and core voltage slider to +100mv and THEN doing the curve to 2100Mhz @ 1.093v results in 118 FPS. Quite a difference and I can repeat it time after time.

So yeah, only use Curve with an as high as possible core slider OC and +100mv on the slider. It gives more performance somehow compared to 0Mhz slider and then a big jump in the curve!.

Sidenote, i'm back to the KFA2/Galax BIOS as the EVGA FTW3 Ultra doesn't behave in terms of power limit. It is 373w which is only 7w less then the KFA2 380W however it shows 120-124% all the time and the KFA2 on a higher clock and higher voltage barely shows 102% so a 20% difference. The KFA2 doesn't throttle. The EVGA does at 1.093v.

And yes, the aircooler on the Gainward Phoenix GS is really struggling with 1.093v at any reasonable fanspeed. It runs around 66c stabilized after a long while but only with 100% fanspeed as VRAM and VRM heat is also put directly into the main heatsink so more voltage is 2x more heat as the VRM gets hotter as well.

Glad to see my Kraken G12 + X52 + Enzotech VRAM/VRM heatsinks coming in sometime next week. Until that time i'll just have to live with the noise
04-05-2020 12:02 AM
Mooncheese
Quote: Originally Posted by sultanofswing View Post
@86Jarrod I tried your method and got this result.

+90 offset slider(2145mhz), Then dragged 1093mv to 2160.

Weird thing is the second test showed it being all over the power limit. Hwinfo64 shows a peak of 422 watt, Weird since this Bios is good for 520 watt.

17,300 that is awesome.
04-04-2020 07:44 PM
EarlZ
Quote: Originally Posted by z390e View Post
Asus Strix OC RTX 2080ti

Ambient right now is 18 degrees C
The cooler on that card is on a different level compared to the Gigabyte Gaming OC, its about 20-25c better.

I decided to do a repaste and it was worth the effort. The TIM Gigabyte used was already dried up and flaked. Used the Noctua paste that I had around and my temps are abou 11-12 lower on MAX load and it bearly breaks 74c with the division 2 now.
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off