MelonGx
Stop PSU arguments.

Someone said EVGA has 10A fuse on FTW3’s each 8-Pin.
Since the red lamp came before driver 457.09, it’s probably the 10A fuse’s blowing issue now.

3090 consumes 940W peak power within 1ms in specific patterns
->
940/3=313 > 12*10=120
->
Three 10A fuses all blowed
->
Red lamp

marcula
CPU: 6700k @ 4.5 GHz
Mobo: Asrock Z170 Extreme7+
PSU: Seasonic Snow Silent Platinum Rated 1050w (5 years old)
 
Installed the 3090 with no issues. It was replacing my FTW3 1080 ti.
 
Card was mounted in the top x16 slot, and case was open to keep an eye on things because I'm paranoid whenever I install new hardware (heh...) So plenty of breathing room for it. Ambient temp in my room was like 72 F / 22 C. 
 
Updated to latest Nvidia drivers, installed Precision X1 and updated the firmware when prompted to. Changed the RGB color but didn't touch OCing or fan control. 3090 was idling around 30 C.
 
Ran a few games and tried rendering some stuff in Adobe Premiere. No issues and performance seemed great. Highest temp I saw was 63 C.
 
I'd had the card in for all of 4 hours at that point. I was idling on the desktop while chatting to a friend on Discord when suddenly... The screen went black, I heard the fans on the 3090 ramp up to jet engine mode, and I could faintly smell burnt plastic/electronics. Quickly powered the PC down and gave it a few minutes to cool off.
 
Powered back up and the card had a single red light with no RGB.
 
I was able to get into bios and windows using my 2nd monitor plugged into my mobo using the igpu. The 3090 wasn't recognized at all. Tried running DDU and reseating the card. No dice. Flipped the OC bios switch while powered down. Nothing.
 
So my brand new, $1800, not even out of the box for 24 hours GPU is a brick. I reinstalled my 1080ti and it's working just fine.
 
I called tech support last night and the guy I spoke to was great. Started the RMA process, went with the standard option so I just gotta ship my card back to get the replacement. Hopefully it was just a random lemon from the bunch and not indicative of a larger issue with FTW3s. But there's been several other threads, and this guy's video which all pretty much describe the same issues: (edit- can't post links?) youtube dot com watch?v=qUVdJxdqUNY
 
I'm also skimming these other threads talking about certain PSUs not being able to handle the 3090's transient power spikes. My PSU is a bit older now (Seasonic Snow Silent 1050w) but it's got more than enough wattage, and Seasonic isn't known for being crap... And never had an issue when my 1080ti was installed. So people saying "x PSU (doesn't) work with a 3090" seems like pretty worthless data, unless they've personally tested multiple PSUs with the same card.
 
 
 
 

arestavo

Intoxicus
Seems like a defect. Your card wouldn't be bricked by not getting enough power. Too much power would be a far more likely cause of a brick.
...

Cards draw power, PSUs don't push power - which your third sentence seems to reference. So, the PSU only provides what the card pulls - not more, and not less unless the PSU electrically cannot provide what the GPU is trying to draw.

Nereus

jankerson

arestavo

Intoxicus
Seems like a defect. Your card wouldn't be bricked by not getting enough power. Too much power would be a far more likely cause of a brick.
...

Cards draw power, PSUs don't push power - which your third sentence seems to reference. So, the PSU only provides what the card pulls - not more, and not less unless the PSU electrically cannot provide what the GPU is trying to draw.

Right, unless there is an issue like a short or a problem with the PSU then that would be different.
 
But under normal circumstances yes the PSU supplies the power requested of it by the hardware.

So.. the problem would be the card, not the PSU, or at least the card demanding more power than the PSU can provide. Wouldn't a firmware update be able to cap these spikes?
 

jankerson

Nereus

jankerson

arestavo

Intoxicus
Seems like a defect. Your card wouldn't be bricked by not getting enough power. Too much power would be a far more likely cause of a brick.
...

Cards draw power, PSUs don't push power - which your third sentence seems to reference. So, the PSU only provides what the card pulls - not more, and not less unless the PSU electrically cannot provide what the GPU is trying to draw.

Right, unless there is an issue like a short or a problem with the PSU then that would be different.
 
But under normal circumstances yes the PSU supplies the power requested of it by the hardware.

So.. the problem would be the card, not the PSU, or at least the card demanding more power than the PSU can provide. Wouldn't a firmware update be able to cap these spikes?

No.
 
Transient power spikes are not a new thing, they started to become more apparent in GPUS starting with the 1000X series and the Radeon Vega 64.
 
Graphics cards go from 0 to 100% in microseconds, so the power draw will spike for a few microseconds.
 
Like when a electric motor starts up, it draws more power to get it going then levels off.
 
Same thing with GPUs
 
With the PSUs it's just plain and simple poor design on the secondary side of the unit. Some of it is caused by cost cutting as we all know cost seems to be a large factor with people. So the companies have to have multiple tiers of PSUs based on cost and quality will also go up or down depending.
 
In the end people need to start taking the PSUs more seriously instead of the 1st thing they cheap out on. And actually do some real research looking at the professional reviews. They will spend a month researching keyboards or a mouse, but don't even bother to research a PSU at all before buying it.
 
The PSU is the most important part of the PC by far, but it's not fluff or exciting to post on line like RGB whatever to show off. So people don't even think about it and just buy the cheapest or whatever is $10 less than something else.

Nereus

jankerson

Nereus

jankerson

arestavo

Intoxicus
Seems like a defect. Your card wouldn't be bricked by not getting enough power. Too much power would be a far more likely cause of a brick.
...

Cards draw power, PSUs don't push power - which your third sentence seems to reference. So, the PSU only provides what the card pulls - not more, and not less unless the PSU electrically cannot provide what the GPU is trying to draw.

Right, unless there is an issue like a short or a problem with the PSU then that would be different.
 
But under normal circumstances yes the PSU supplies the power requested of it by the hardware.

So.. the problem would be the card, not the PSU, or at least the card demanding more power than the PSU can provide. Wouldn't a firmware update be able to cap these spikes?

No.
 
Transient power spikes are not a new thing, they started to become more apparent in GPUS starting with the 1000X series and the Radeon Vega 64.
 
Graphics cards go from 0 to 100% in microseconds, so the power draw will spike for a few microseconds.
 
Like when a electric motor starts up, it draws more power to get it going then levels off.
 
Same thing with GPUs
 
With the PSUs it's just plain and simple poor design on the secondary side of the unit. Some of it is caused by cost cutting as we all know cost seems to be a large factor with people. So the companies have to have multiple tiers of PSUs based on cost and quality will also go up or down depending.
 
In the end people need to start taking the PSUs more seriously instead of the 1st thing they cheap out on. And actually do some real research looking at the professional reviews. They will spend a month researching keyboards or a mouse, but don't even bother to research a PSU at all before buying it.
 
The PSU is the most important part of the PC by far, but it's not fluff or exciting to post on line like RGB whatever to show off. So people don't even think about it and just buy the cheapest or whatever is $10 less than something else.

 
I get it, but can there not be an enforced cap on that spike? I'm not an electrical engineer so I have no idea what can and can't be done in this situation.

I purchased a SuperNOVA 850 T2 for this build, so hopefully I'm not going to have any issues... 3090 FTW3 Ultra delivery expected first half of next week.
 
 

jdmwrxpower
When the card ramps up it is asking the power supply for more power. The issue is resistance. Resistance creates heat and limits current. The spike can actually blow the output on the power supply at the so instead the card is now operating on limited voltage and current. It's like a brown out for your video card. Enough power and current to run but operates dangerously below the threshold and causes electronics to fail. When you read about cheap power supplies you have to question the integrity of the circuits. Like one user stated when an electric motor starts it draws the most current that is why electricians install slow blow fuses because if you don't they will pop a normal fuse. Top tier power supplies are generally really good at handling spikes. The more headroom you have the better but weather or not you have 1000w or 1600w supply what matters is how much current can flow over the voltage rails which is all internal. What needs to happen is the card needs to be able to detect when it's calling for power and not getting enough to immediately shut itself down. Isolation safety circuit like an automatic E-stop so nothing is damaged. All I ever used was EVGA power supplies and they are great but I will caution you on using EVGA replacement cable sets they sell. Two of my friends lost a 2080ti after using the EVGA cables. After close inspection we found some of the pins were not crimper all the way and they were almost black which is indictive to arcing in the electrical fiels and electrical engineering world. Always use the factory cables that came with the PSU. Just my thoughts

jacoffey85

Nereus
Thanks for that explanation, awesome! As it happens, I do have individually sleeved cables from EVGA ...I've been using them for a few years without issue so far.. hmm
 

Usually the problem comes from aftermarket cable kits, I’ve personally never had issues with any of my included PSU cables.

Another thing to note, while I completely agree with getting a quality PSU for your PC, I’m realizing there’s a lot people in some of these threads related to PSU problems who don’t know how they actually work. One example would be people talking about OCP yet they’re a little off on how overcurrent protection is designed or what it is even trying to protect. I’m seeing a few terms getting tossed around such as “transient” but based on the context I don’t think some people using it truly understand it. Look for professional reviews on PSU’s. I’m an Electrical Controls Engineer, but even I can learn a thing or two about PSU from in depth reviews about designs since it’s not something I do every day.

kurtley
Geez, this post scares me. I shipped my 2080 ti through the Step-Up program last week, and my 3090 FTW3 Ultra should be here next week. 

jankerson
 
Your T2 1000 will be fine.

kurtley

jankerson
 
Your T2 1000 will be fine.

Oh good. Thanks, jankerson. So, the issues are PSU related then. Good to know.