The 3090's are power hogs, but I doubt the cinematic would crash the card from not having enough power (850w should be plenty given the specs). Could really be any of a number of things.  Hard to diagnose without repeatable errors.
 

Update your driver
Update OS
Update PX1/Afterburner
Reset card to factory defaults (no over/underclock)
Check integrity of game files
Purchase quality 1500va UPS from APC (eg: https://www.apc.com/shop/us/en/products/APC-Back-UPS-Pro-1500VA-for-Gaming-120V-Pure-Sinewave-LCD-Black-3-USB-charging-ports-10-NEMA-outlets-4-surge-/P-BGM1500B)
 

Fennario
Update your driver
Update OS
Update PX1/Afterburner
Reset card to factory defaults (no over/underclock)
Check integrity of game files
Purchase quality 1500va UPS from APC (eg: https://www.apc.com/shop/us/en/products/APC-Back-UPS-Pro-1500VA-for-Gaming-120V-Pure-Sinewave-LCD-Black-3-USB-charging-ports-10-NEMA-outlets-4-surge-/P-BGM1500B)

 
ah yes, a UPS for gamers...
 

You should look at the Windows Event Viewer to see what errors are associated with the crash.
I'm not particularly impressed with the current Seasonic PSUs, or for that matter, Gigabyte's GPUs.  I

tmmy34
Only kernel 41

Has this only happened once? If so, I wouldn't worry about it. Kernel 41 errors can happen for any number of reasons, including a random flipped bit.

I would start with a fresh install of the game and also a clean install of the most current nvidia driver.  Also make sure windows is fully up to date before you do the fresh driver install.
 
If you are still experiencing issues, you could have a hardware issue / or it could be tied to a bad windows update install.. If you are willing and able, a fresh windows install may help.  Before you do that, you can always check/test the PSU, RAM, SSD..  Overheating can cause this as well, so make sure all fans/AIO coolers, etc are properly working.  .   I don't think it is the GPU, but you never know, but good luck.
 
 
 

when the GPU usage goes up really fast it can cause a PSU OCP if your PSU is close to the edge, AN 850 T should have the head room not to do this but 3090s can spike crazy sometimes. 

OMG, no a card being 2x8pin does not mean it pulls more power from the slot. That has nothing to do with how the card may load balance.

If the machine actually powered down that is PSU OCP. Some games cause higher power spikes than others. Some PSUs hit OCP more easily than others.

If it doesn't do it anymore I wouldn't worry about it. If it does it again buy a new PSU. I've had good luck with my RM850x and RM1000x. Neither have hit OCP running a 1kw XOC bios, so they should easily handle a stock 3090. There are plenty of other good PSUs and there's a thread floating around this forum with people listing working PSUs with 30 series cards.

talon951
OMG, no a card being 2x8pin does not mean it pulls more power from the slot. That has nothing to do with how the card may load balance.

We've seen how well this gen load balances.  We've seen power spiking during a cutscene, too. 

850W with a 3090? In all honesty sense the 3090 CAN draw over 500W of power.. having only a 850W isn't that great of an idea. I am using a 3080 FTW3 with a 1300W... not saying go that high but should be at least a 1000w min to be safe. On top of that the the 10850K can draw nearly 300W by itself under Load. Add on fans, HD's, etc.. 

Given the fact the computer shuts off, and the PSU's DO degrade over time, Feel very confident it's a PSU issue. 

BTW, BF5 on a 3090, isn't taxing, which explains why the GPU isn't shutting off, as the game doesn't tax GPU's to 100%. What's your GPU usage hit?

nomoss
We've seen how well this gen load balances.

This gen doesn't have problems load balancing. Some of EVGA's cards have a problem load balancing. This person has a Gigabyte card. It is not affected by "this gen's" supposed load balancing issues.

So rma psu or change psu if that happened once?
 
 
ps:
Maybe it was some kind of outage ? I repeated launching game maybe 50 times again and no shutdown.

ty_ger07

You said load balance. We aren't talking about power spikes when talking about load balance. NVIDIA's power monitoring is too slow and based on averaging, and thus power spikes is a separate topic from power balancing. Better balancing should cause power spikes to be better distributed on average, sure (since one input isn't constantly near the upper edge compared to the others; just waiting to be pushed over the edge), but not necessarily certain.

Technically, talon said "load balance."  And then you just explained why load balancing could be a concern during power spikes.  
150+150+75 on a 3090 - I'd be undervolting the heck out of that card.

nomoss
Technically, talon said "load balance." 

nomoss
We've seen how well this gen load balances.

This gen doesn't have problems load balancing. Some of EVGA's cards have a problem load balancing. This person has a Gigabyte card. It is not affected by "this gen's" supposed load balancing issues.

nomoss
And then you just explained why load balancing could be a concern during power spikes.  

Yes. Load balancing can negatively affect the behavior during power spikes. If you have an affected EVGA card, it's not great. But this person doesn't have an EVGA card.

This generation as a whole has a power spike issue; by design. Some cards it matters, some cards not so much.

This generation as a whole doesn't have a load balance issue. Only certain cards (from EVGA, for example) have a load balance issue.

If you can't make sense of it, draw a venn diagram. This user has a Gigabyte card which is not affected by the problem you said that it is affected by.

ty_ger07

nomoss
Technically, talon said "load balance." 

nomoss
We've seen how well this gen load balances.

This gen doesn't have problems load balancing. Some of EVGA's cards have a problem load balancing. This person has a Gigabyte card. It is not affected by "this gen's" supposed load balancing issues.

nomoss
And then you just explained why load balancing could be a concern during power spikes.  

Yes. Load balancing can negatively affect the behavior during power spikes. If you have an affected EVGA card, it's not great. But this person doesn't have an EVGA card.

This generation as a whole has a power spike issue; by design. Some cards it matters, some cards not so much.

This generation as a whole doesn't have a load balance issue. Only certain cards (from EVGA, for example) have a load balance issue.

If you can't make sense of it, draw a venn diagram. This user has a Gigabyte card which is not affected by the problem you said that it is affected by.

Sigh - again - you said that’s the problem I was referring too. It’s also a problem that was reported on other cards (but EVGA got all the press.)
I said it’s a design issue, which you agree with.
I’m not sure what your point is, except to say that a Gigabyte card can’t have load leveling issues - which is a conjecture - just as my supposition that the spike could be causing his issue and could be because the spike is overwhelming the PCIE slot is a conjecture that draws some parallel from but not assertion that it is the same issue EVGA had. My conjecture draws from other info I’ve seen and issues I’ve had with Giagabyte products but not the specific ones in his build. Both of our conjectures would require testing to turn into a theory. How much effort do you really want to put into this? He should get a bigger power supply if it happens again is probably the right answer to his issue, if not a complete correction of the root cause.
If the point is to win an argument that’s entirely based on a one time Kernel 41 error and shutdown with no repeat then - ok, you win?