Guys, how exactly you do this undervolt in Afterburner? Do I need to turn down every step of the curve? And, if yes, how much you lowered it in every step? Sorry... Never made vga undervolting before. :(

ClowReed
Guys, how exactly you do this undervolt in Afterburner? Do I need to turn down every step of the curve? And, if yes, how much you lowered it in every step? Sorry... Never made vga undervolting before. :(

You can search for the 3080 undervolt video on Youtube made by WccftechTV.

I have a 3080 FTW3 Ultra incoming and not sure if I should just do this fix on day 1 or run it stock. Don't want to risk it until EVGA releases a bios fix.

Ok so a little backstory and then I'll get on with it... So...
 
My first 3090 FTW3 Ultra I got the red light on the 3rd PCIe plug closest to the die. 
RMA'd no issues, EVGA reps and service were superb. 
This card had the black screen high fan crash twice before it killed itself. 1000W gold EVGA PSU
This occurred in OCT.
 
Since then I have been using my current 3090 heavily with moderate issues. 
In total, I have had 10+ black screen 100% fan crashes. Each time I think the card is going to kill itself. But each time I have been able to pwr off the system and recover with minimal issues. 
I have replaced the 1000W PSU with a brand new 1200 watt platinum. While playing The Division 2 today I had the same crash... black screen 100% fans. So.... I will experiment with undervolting more to see if it will fix the issue.
 
Games that have black screened and crashed -
 
American Truck Sim - 1 crash (killed the first card)
Witcher 3 - 3 crashes
Division 2 - 4 crashes (1 on the new PSU)
Half Life Alyx - 1 crash
Warframe - 3 crashes 
 
 

LocutusHUN

MatthewAMEL
 
Are we talking only 3090's? I have a 3080 FTW3. Received in November. I play a lot of older games (Fallout:NV, SupCom, Homeworld Remastered). I have put hundreds of hours in on the older games since I received my card. I've never had a crash or black screen. I do not OC my card. My most recent TS: 17205, TSE: 8890.

Good question there. I am also using my 3080 FTW3 Ultra since november, and have no such issues listed here.
I would also like to assume, that this topic is only about 3090's.

And I've had my 3090 since September.
 
It's not always the card.
 

donnie123
Ok so I have my 3080 FTW3 Ultra since around 2 months now, its a 2014 black lips, it has an uneven power draw, and the 3rd plug on stock is reaching max 91w. Although the card is hitting 400w normally, and is spiking to 419w sometimes with 105% power setting in precision x1.
 
The clocks are boosting up to 2100 on stock bios with +105 on core, but during gaming in cyberpunk 2077 or warzone the clocks are sustained at around 2070 mhz. This is with zero voltage added. 
 
I've had so far zero crashes, backscreens or whatsoever problems, I been playing long hours of League of Legends, CBP2077 and Warzone. I've checked and seen voltage spike like somebody mentioned..
 
Should I consider myself lucky so far? Well knocking on something wooden rn. 

The problem is not at high power, it's at low power states.  If you've been able to play League without crashing, you do not have the issue.

a3britton
After playing around with undervolting I have some more input on my experience.
 
I have played around 10 hours with my current undervolt. BL3 and Division 2 mostly. I have had ZERO hard crashes. 1 crash to desktop for each game. (attributed to my undervolt I suspect. Tuned slightly after each crash)
 
Today while playing Division 2 I had another hard crash black screen 100% fan. This happened after I removed my undervolt. I set MSI to 107% pwr and added a +75Mhz OC to core and +250Mhz to mem. 
 
I proceeded to reapply my undervolt and for another 3 hours or so and no crashes whatsoever. 
 
IMPORTANT NOTES:
My card is a FTW3 Ultra 3090
I have a brand new 1200w PSU Platinum rated from Thermaltake
I have replaced the cooling solution with EVGA's own Hybrid kit
I have installed the official Stock EVGA Hybrid FTW3 Ultra BIOS.
My fans for the radiator currently run of my Motherboard. I have found this to be a more stable option.
 
 
 
 Currently set my card at 1.000v 1935Mhz and it really seems to be solid. This profile sort of mimics the core clock performance of stock settings for my card. While using about 40 less watts and 5-6c cooler.
 
I will continue to keep testing and playing around with this and report back. Maybe a week or two of testing and I will have a better idea of how this profile performs while tweaking as needed. 
Using HWiNFO64 I can see my voltage maintain and stick to 1.000v. The total PWR draw seems to top out at 390-405w.
 
Linked is a 10 minute snapshot of me playing Division 2.
You can see the Core clock peaked at 1950 - but I maintain 1935Mhz pretty solid. dipping here and there and raising here and there.
 
This behaves just like stock essentially. Anyway - I hope this helps anyone else experiencing issues. I'm not sure this is even viable tho. Most users will not put in the time or have the necessary knowledge to do this. I would prefer a new card that does not behave this way.

Good information. But to be honest, it makes no sens to buy a $2000 card and you can't use it to its full potencial.
 
I'm doing a little overclock on mine and i'm going to continue. If fail, i will do RMA. 
 
And to be more honest. If i had a card with Blackscreen problems like your, i would have already opened the RMA Ticket.
 
 

An update....

After a 4 hour gaming session of League today, I hard crashed at 1.062 V....after the card was fine for about 2 weeks.

Guess I'm lowering the voltage limit further.....

I sense an RMA in my future unfortunately....

I've got a RTX 3080 FTW3 Ultra, and I've been experiencing reboots at high loads. At first I thought it was just Cyberpunk 2077, so I'd tried the usual: DDU, reinstall driver/install older driver, but that didn't help. So, I thought maybe it was just that one game, as I'd been playing several other games before it released, including the Resident Evil 2 and 3 remakes, RE7, Red Dead Redemption 2, and Wolfenstein Youngblood. No issues with those. And I didn't have any problems with my previous RTX 2080, which I had sold about a month and a half after I'd got my 3080. My temperatures are at max load: ~60C on the CPU and GPU. I have pretty good airflow in this case, with plenty of fans, and I keep it cleaned regularly.
 
I have also disabled Windows' Startup and Recovery/System Failure/Automatic Restart, to see if I could see any BSODs. When my PC reboots like this, however, I don't receive any BSODs--just a reboot, like the reset button had been pressed. In the Windows Event Viewer, it only displays an Event ID 41 Kernel-Power, with no timestamps or bugcheck codes.
 
So I downloaded OCCT to stress test my components. The CPU, Linpack, RAM, VRAM, and PSU tests all passed, but as soon as I ran the 3D (GPU) test, my PC would reboot. Other benchmarks and stress tests wouldn't cause a reboot, such as the ones included with 3DMark. I also used memtest86 to check my RAM, and it also passed the full length test. And of course I also set my CPU and RAM overclock to all stock/default/auto settings in my motherboard BIOS. This also didn't help. 
 
Also ran sfc /scannow and DISM.exe /Online /Cleanup-image /Restorehealth and found no system file integrity errors. 
 
I tried using lower power settings, but that didn't help, either. I tried reseating my GPU, plugging and replugging the power connectors. No luck. So then I wondered it my PSU was faulty, even though I'd installed it at the same time as my RTX 3080. I'd bought this Seasonic Prime TX-1000 specifically for this RTX 3080, in case my previous Seasonic Prime Titanium 750W PSU was underpowered. But I went ahead and swapped the older PSU back into my PC, along with all the power cables. Oh and I'm using separate cables for each of the video card's power connectors---in fact, both of these PSUs only use single cables; no daisy chained cables here. Well, that didn't work, either. I'd still get PC reboots with OCCT 3D stress test. 
 
Then, I put my 1000W PSU back in, and decided to bypass my uninterruptible power supply (UPS) and just plug my PC directly into the wall. That seemed to fix the problem, as I was able to pass the OCCT 3D stress test then. I'd also bought that UPS, a Cyberpower 1500VA/1000W UPS, just recently. I had been using a 600W Cyberpower UPS before that. But I had the idea to use both of my UPS's, and plug everything but my PC into the smaller one, and just my PC into the larger one. That also seemed to work, as I was able to pass the stress test and play Cyberpunk 2077 without issue for at least a week. But then I got another reboot while playing Cyberpunk, along with reboots while running the OCCT 3D stress test, so I unplugged my PC from the UPS and went with plugging directly into the wall again. 
 
That also seemed to work, for a while...until I eventually started to get reboots in both Cyberpunk and OCCT's 3D stress test once again. So...I don't think it's any of my PSUs or UPSs. 
 
I came across this thread, and tried lowering my power limit even further than I had, down to 65% as someone else had tried. I got more reboots while trying OCCT's 3D test again. I also tried an undervolting method mentioned in this thread, and it didn't work, either. Actually, before I ran the stress test, I tried the Unigen Heaven benchmark, which hadn't crashed before, while all this has been happening...but my PC rebooted while running it. So it doesn't look like undervolting has helped me, either. So I reset the GPU back to stock settings again.
 
Strangely, I've been able to play Battlefront II without any issue, and played a round of StarCraft II earlier today without any problems. And while I had been able to play Heroes of the Storm without crashing or rebooting, as it's a fairly lightweight game, I did get a reboot while trying to load a map today. So, it seems like the problem is getting worse. But, I'm not getting any red lights on the video card. 
 
I've already started the RMA process for the RTX 3080 and I'm just waiting for approval. Hopefully this fixes the issue. But I wonder if it could be my motherboard, or my motherboard's PCIe slot. I haven't tried the 3080 in the lower slot, yet. But I figured that if the PCIe slot is faulty, I'd have other issues, such as artifacting, or black screening, losing connection to the monitor, etc. I'm not getting any of those problems. I also haven't tried replacing my CMOS battery. But in the case of the battery, I think I'd be experiencing other issues (such as losing time and date, BIOS settings, etc.). I've experienced faulty motherboards in general, and I know the usual symptoms of that: reboots at idle, freezing, no signal to monitor, corrupt Windows file integrity, etc.--I'm not getting any of that here. 
 
I'm really at my wits' end here. I hope the replacement GPU solves these issues, but if not, then I'll know it's my motherboard (or something else, I guess..). I'm just really not looking forward to replacing it, since the stock on any that would fit my CPU is very low, and prices are higher than they would normally be. And this isn't the best time to build a new PC (the core components anyway, the CPU and motherboard and maybe also RAM), since there's new stuff coming out and scalping is still rampant. I was hoping to keep my core setup for at least another few years.
 
Just to round this off, here are my full specs:
CPU: Intel i7 8700K @4.7 GHz
Motherboard: Asus ROG Maximus X Hero (Wi-Fi AC) - Z370 chipset
RAM: 32GB (4 x 8 GB) Corsair Vengeance LPX DDR4-3200 CL16
GPU: EVGA GeForce RTX 3080 10 GB FTW3 ULTRA GAMING
Storage: Samsung 850 Pro Series 1 TB 2.5" Solid State Drive
Storage: Western Digital Caviar Black 4 TB 3.5" 7200RPM Internal Hard Drive
Storage: Samsung 970 Evo 1 TB M.2-2280 NVME Solid State Drive
PSU: SeaSonic PRIME TX 1000 W 80+ Titanium Certified Fully Modular ATX
Case: Corsair Air 740 ATX Full Tower Case
UPS: CyberPower CP1500PFCLCD
 
***just a quick edit to add some info on the Cyberpunk 2077 reboots: So when this first started happening, it wasn't happening all the time. In fact, at first I thought it was just a fluke. Anyway, it was only happening in Cyberpunk 2077 when I hit the Esc key to pause the game. It wouldn't happen in normal gameplay. It normally runs very smoothly with mostly high settings and the top level ray tracing settings, with DLSS on Balanced. It would just reboot whenever I hit pause, but not all the time. Sometimes I could be playing for a few hours and it would happen, sometimes when I first loaded into the game, and sometimes not at all. It wasn't happening on any other screens that would effectively pause the game, such as the inventory, character, or map screens (I had heard some people were getting this same issue in those other screens). But at one point it did crash/reboot when I had used a fast travel station, which is basically the map screen. More recently, I got a reboot before I got into the main menu; I hadn't even loaded a save. But it still wasn't happening in any other game or benchmark, until just recently, when I had a crash/reboot while running Unigen Heaven, and today when it rebooted while loading into a Heroes of the Storm game. But that is why I was interested in this thread, because with Cyberpunk, which is very graphics-intensive, I can see the wattage (using HWInfo and RTSS) drop from ~380W to ~140W, and the talk about low power states made me think of this. I'm not sure if that's in the same vein of the discussion or not, though. 
 
***edit 2: I know I could still try a few more things to rule out my motherboard (and I can always do that later, I guess), but the posts below mine make me feel better about starting the RMA process for this 3080 now. 
 
***waiting for my replacement 3080 to get here. Thought I'd try to play Europa Universalis IV (EU4) before I take my current 3080 out to ship back on Monday, since it wouldn't go anywhere right now on the weekend. EU4 is an very lightweight game, essentially just a map. My PC rebooted right as the main menu was about to load. So yeah, whatever the problem is, it's definitely worse than it was. 
 
***Update, 3/3/2021: I got my replacement in today. Black accents (not the red lips this time), serial number first 4 digits 2114, packaged in USA. Just for reference, my previous card was: serial number 2014, made in Taiwan. Everything seems to be okay; no reboots yet with OCCT 3D stress test--tested on shader complexity levels 3, 5, and then 8. I'll be fairly satisfied that the rebooting issue is fixed if I continue to pass gaming/benchmarks/stress tests over the next week or so. 
 
Initially, I had a loud clicking noise coming from one of the fans, but I updated the firmware to the latest and it went away, although the standard fans ramping to 100% during the firmware update scared me...the clicking was very, very loud. But then I rebooted my PC myself, and the fans always spin up during reboots, just as they do during power on, and they didn't make any noises then. I ran the stress test and a few benchmarks and sometimes I'll hear a faint clicking/rattling when the fans ramp up, but it stops. I've checked that I don't have anything up in the fan shroud/heatsink, but I could check again I suppose. I've checked to make sure all fans are running at roughly the same RPM, and they are, so I don't think any of them were damaged. 
***

 I thought I had my issue figured out too. I went from a x99 classified system with evga p21000 power supply and a 3090rtx ftw3 ultra. System did the black screen wide open fans in low usage game several times.  I decided to build a 10900k system asus strix e gaming z490 maxtytan 1050w PSU  same gpu obviously...... I went no issues for about 2 weeks and then bam just did it last night low usage games guess I will RMA the card this time.

Rewire92
Another update.

Had a crash at 1.056 V....less than 24 hours after my first crash.  Just playing League again.

Limiting to 1.05 V.....

This may not work as a long term solution.

Damn, I was really hoping your solution would work long term. 

What’s the power draw on the PCIe slot. If it is under 80 using various workloads then you should be good.

Rewire92
Wish EVGA would tell us what's going on.

This does not make me confident on the card's longevity.

This is very true, i have a card (FTW3 3090) that has a great performance and i think i got lucky on the silicon
 
But just out of concern I ended up buying a 3090 strix that is on the way. If I don't get an EVGA return in the next few weeks, I will end up selling my card.

....another crash, only at 1.05V, less than 10 hours after restricting it to that.

Guess my card is going to die soon.

What a joke.  Two Thousand Dollars.

Gears 5 w/o the voltage curve:
Boost clock: 1905 Mhz
Max PCIe slot power: 47.9 watts
Max 8 pin #1: 108.7 watts
Max 8 Pin #2: 117.3 watts
Max 8 Pin #3: 110.3 watts
Peak GPU voltage: 1.081v
Avg framerate: 85.7
1440p, Ultra + Insane settings, SVOGI with 32 rays, VRS disabled

Gears w/ voltage curve:
Boost clock: 1900 Mhz
Max PCIe slot power: 35.6 watts
Max 8 pin #1: 86.3 watts
Max 8 Pin #2: 98.4 watts
Max 8 Pin #3: 71.6 watts
Peak GPU voltage: .875v
Avg framerate: 84.6
1440p, Ultra + Insane settings, SVOGI with 32 rays, VRS disabled

Without and with the voltage curve, I usually see a max of 12-13w getting pulled from the PCIe slot at idle or putzing around Reddit with Youtube in the background. In games, the most I've ever seen was around 65w getting pulled from the PCIe slot and that was without the curve. Now, it tops around 49w. The GPU voltage was all over the place without the VC: it'd peak at 1.081v, but would pingpong from .950 to 1.062 depending on the scene and would occasionally spike in just desktop usage if going from Reddit to a streaming service for a blip before settling back down. I dunno if the pre-VC numbers are good or not for what amounts to stock settings, though the uneven power draw across the 8 pins is a bit more evident with the VC in place.

Temps w/o the VC didn't exceed 70C under load but were around 40C at idle (set to maximum performance in the control panel, Windows has been set to the high power plan) and with the VC it never goes over 60c. Not a stellar result, but not horrific when one considers the 5900x has a NH-15s on it and my power-to-DB skews to quiet more than anything else (Cooler Master Cosmos II as a case, Noctua NF-S12a fans for the 120mm fans + the 200mm Noctua NF-A20 crammed in the front). Adding to that, with only an 850w unit, I figured OCing was a no-go since it seems Ampere (regardless of the board partner you're buying from) seems to have pretty significant power spikes without a VC in place and I figured I'd trip OCP if I was in the 450w range with just the card.

The last GPU I truly dabbled in OCing or adjusting voltages on was with was the GTX 570 SC I had and that was because it was a beastly sample (932 Mhz rock solidly stable at stock vcore). Since then and up until now, I pretty much left GPU boost to do it for me on every card I've owned since (GTX 670, GTX 780, GTX 1070 BE, GTX 1080 Ti BE, RTX 2080 Ti BE).

I digress. I dunno if I have a problem free sample, or if the paranoia that it'll catastrophically fail at some point in the near future is warranted.