Monday, September 21st 2026

NVIDIA RTX 5080 Reportedly Dies After DLSS 5 Test in Modded GTA 5

An MSI GeForce RTX 5080 16G Ventus 3X reportedly stopped working after its owner injected an unofficial version of DLSS 5 into a heavily modded copy of Grand Theft Auto 5. Reddit user "wax469" said the card had been running fine for about a month before the crash. According to the post, the screen began flashing uncontrollably about five minutes into a session, and the system refused to boot afterward. Swapping to his Ryzen 9 9950X3D2's integrated graphics let the PC start normally, and reinstalling the RTX 5080 brought the failure back. A replacement card of the same model has apparently worked without issues since.

It's worth being cautious here. NVIDIA has not released DLSS 5 for GTA 5, so the injection relied on unofficial tools, and the user says the card was running at stock settings with no overclock. There's no photo evidence of physical damage, and it's entirely possible a manufacturing defect was already present and just got exposed under load. Whether DLSS 5 had anything to do with it remains unconfirmed. The user was running a heavily modified version of GTA 5 Legacy, and adding unofficial DLSS 5 seems to be what eventually led to the failure.
Still, this isn't the first time DLSS 5 has been tied to a hardware scare. Tom's Hardware has reported that DLSS 5 is linked to a significant increase in power draw. We also reported recently on an RTX 5090 whose 16-pin connector melted during official DLSS 5 testing in NBA 2K27, with power draw spiking past 600 W. That case, too, had more questions than answers. For now, the advice is to stay cautious with unofficial DLSS 5 ports and mods, especially when graphics card prices are at an all-time high right now.
Sources: Reddit, via Videocardz
Add your own comment

53 Comments on NVIDIA RTX 5080 Reportedly Dies After DLSS 5 Test in Modded GTA 5

#2
windwhirl
Begs the question of whether those DLSS5 patches and mods are somehow injecting stuff they shouldn't inject into whatever logic controls the card's power management, because otherwise the card should manage its energy usage in a way that doesn't lead to damage from excessive current/voltage/whatever, regardless of whether DLSS5 is active or not.
Posted on Reply
#3
Vayra86
windwhirlBegs the question of whether those DLSS5 patches and mods are somehow injecting stuff they shouldn't inject into whatever logic controls the card's power management, because otherwise the card should manage its energy usage in a way that doesn't lead to damage from excessive current/voltage/whatever, regardless of whether DLSS5 is active or not.
Should... but then there was this event not too long ago

www.pcgamer.com/new-world-making-gpus-unwell-again/
Posted on Reply
#4
Hecate91
It's like the New World game bug again, pushing cards beyond their limits, even though the firmware shouldn't be allowing power to get pushed hard enough to kill the GPU.
Then again these DLSS5 patches can't be too different than what "leaked" out, it seems more like DLSS5 was launched too soon since the 60 series is possibly delayed, this feature also melts the 12v2x6 connector.
Posted on Reply
#5
Let3r
Hecate91It's like the New World game bug again, pushing cards beyond their limits, even though the firmware shouldn't be allowing power to get pushed hard enough to kill the GPU.
Then again these DLSS5 patches can't be too different than what "leaked" out, it seems more like DLSS5 was launched too soon since the 60 series is possibly delayed, this feature also melts the 12v2x6 connector.
And what happened with furmark a while back.
Posted on Reply
#6
RejZoR
windwhirlBegs the question of whether those DLSS5 patches and mods are somehow injecting stuff they shouldn't inject into whatever logic controls the card's power management, because otherwise the card should manage its energy usage in a way that doesn't lead to damage from excessive current/voltage/whatever, regardless of whether DLSS5 is active or not.
No. Firmware logic for hardware operation that keeps graphic card alive (ie not pump insane voltages, clocks or amps into core and keep everything within temperature) should be entirely separate from any driver related stuff. Even if user sets fan speed so low it can't sustain itself it shouldn't melt to death, it should just downclock so hard it wouldn't overheat or in worst case simply disconnect from PCIe and shut itself down.

What is likely happening is that NVIDIA never really anticipated so much of GPU logic will be operating all at once, raster, tensor and RT. Still doesn't explain how any of this would bypass firmware protection logic. Then again this isn't their first massive blunder... looking at their famous connector they are pushing so hard...
Posted on Reply
#7
nageme
said the card had been running fine for about a month before the crash
So it's not a data point, it's noise.
Posted on Reply
#8
Visible Noise
nagemeSo it's not a data point, it's noise.
Reddit noise.
Posted on Reply
#9
Darmok N Jalad
windwhirlBegs the question of whether those DLSS5 patches and mods are somehow injecting stuff they shouldn't inject into whatever logic controls the card's power management, because otherwise the card should manage its energy usage in a way that doesn't lead to damage from excessive current/voltage/whatever, regardless of whether DLSS5 is active or not.
Granted, this is a one-off incident that could be attributed to many things, but you are correct in that modern GPUs (and CPUs) should be able to protect themselves appropriately. Not because ignorant users will try anything, but because modern GPUs (and CPUs) "overclock" themselves based on available power and thermals (aka, the "boost clock" spec). DLSS5 might be activating more silicon than ever before, and the algorithms for managing power draw might not be managing the thermal budget appropriately. We also saw a case of a 5090 drawing over 600W running DLSS5, which it should NOT be able to do. This could be a nasty side effect of companies pushing their silicon to its upper limits. Time will tell if these are just sensational stories or if the cards are indeed drawing too much power when running DLSS5.
Posted on Reply
#10
Hecate91
RejZoRNo. Firmware logic for hardware operation that keeps graphic card alive (ie not pump insane voltages, clocks or amps into core and keep everything within temperature) should be entirely separate from any driver related stuff. Even if user sets fan speed so low it can't sustain itself it shouldn't melt to death, it should just downclock so hard it wouldn't overheat or in worst case simply disconnect from PCIe and shut itself down.

What is likely happening is that NVIDIA never really anticipated so much of GPU logic will be operating all at once, raster, tensor and RT. Still doesn't explain how any of this would bypass firmware protection logic. Then again this isn't their first massive blunder... looking at their famous connector they are pushing so hard...
Or the driver issues Nvidia had for months with Blackwell, I wonder how much of the logic is kept from failing from drivers alone on modern GPU's, all it could take is a vibe coded driver to kill the core or VRAM.
nagemeSo it's not a data point, it's noise.
Until more reports of it happen, I recall people saying reddit wasn't enough proof with the 12 pin connector, now it failing is nearly a daily occurrence.
Posted on Reply
#11
Visible Noise
Darmok N JaladGranted, this is a one-off incident that could be attributed to many things, but you are correct in that modern GPUs (and CPUs) should be able to protect themselves appropriately. Not because ignorant users will try anything, but because modern GPUs (and CPUs) "overclock" themselves based on available power and thermals (aka, the "boost clock" spec). DLSS5 might be activating more silicon than ever before, and the algorithms for managing power draw might not be managing the thermal budget appropriately. We also saw a case of a 5090 drawing over 600W running DLSS5, which it should NOT be able to do. This could be a nasty side effect of companies pushing their silicon to its upper limits. Time will tell if these are just sensational stories or if the cards are indeed drawing too much power when running DLSS5.
My car was fine until I went out to eat at P.F. Chang’s. Then I got a flat tire.

Proof that chicken teriyaki causes flat tires. Prove me wrong.
Posted on Reply
#12
dtoxic
Someone need to test DLSS 5 to the limit and recreate this,but my take is card was already defective in some way and dlss just touched that particular thing,many others are using dlss 5 without any such problems,also i don't want to completely dismiss that it was posted on reddit,but then again it's reddit
Posted on Reply
#14
AusWolf
it's entirely possible a manufacturing defect was already present and just got exposed under load
I think this is the most likely event. It's hard to imagine that software could kill a GPU, although Furmark used to be notorious of that, so nothing is certain.

Also, if DLSS 5 is really capable of overriding a card's power limits and electrical safeties, it has to be investigated.
Posted on Reply
#15
wolf
Better Than Native
This is furmark adjacement in how it absolutely saturates a GPU, I can imagine playing around with it for a month is risky business on any card for various reasons. As a single case goes though, it's hardly a trend and may be enough of a community warning to make others more cautious.

Not forgetting too that it's the first release of DLSS5 modded to within an inch of it's life and shoehorned every other way.
Posted on Reply
#16
A Computer Guy
When are they going to learn. Power limit your card to 300W before running DLSS 5!
Posted on Reply
#17
Luke Skywalker 8472
AusWolfI think this is the most likely event. It's hard to imagine that software could kill a GPU, although Furmark used to be notorious of that, so nothing is certain.

Also, if DLSS 5 is really capable of overriding a card's power limits and electrical safeties, it has to be investigated.
Yes. It goes substantially over the spec. I read about some cases of damaged cards because of it, but shouldn't the card be protected in overpower scenarios?
Posted on Reply
#18
Visible Noise
A Computer GuyWhen are they going to learn. Power limit your card to 300W before running DLSS 5!
I run it at board limit on my 4090. There’s no reason to power limit outside of the power savings itself. I’ve never heard of an RTX 5080 that is incapable of running @360w with a 100% duty cycle.
Posted on Reply
#19
AusWolf
Luke Skywalker 8472Yes. It goes substantially over the spec. I read about some cases of damaged cards because of it, but shouldn't the card be protected in overpower scenarios?
That's strange. I wonder if it's a design flaw, or if it's by design. AMD cards work at slightly lower clocks when the RT cores are active, as the power limit is more evenly distributed across the die. Maybe Nvidia doesn't want the same to happen, and lets the cards eat more than they should when the Tensor cores are hard at work with DLSS 5 just to maintain clocks. Just a theory.
Posted on Reply
#21
Crackong
Ever since Nvidia started using AI to write their own codes, these things happens.
Maybe their "AI written DLSS5" is so good in finding all the loopholes in their "AI written firmware" trying to squeeze every bit of performance out of it.
So it is like the "DLSS5 AI" is attacking its own brother, the "Firmware AI".

The battle happens inside our GPU.
If the DLSS5 AI loses, we get bad performance.
If the Firmware AI loses, we get a burnt card.

Posted on Reply
#22
FR3D1
Ninja Cat
Event HorizonWhat a terrible way to lose a card.
Just your typical Reddit user with *another* dead RTX card. It’s a trend that seemed to be subsiding before the AI boom but is now noticeably on the rise again.

Regards,
Fred.
Posted on Reply
#23
TechConsoomer
Correlation != Causation. Clickbait/lazy reporting by TPU. If my GPU died while watching Youtube, will the media outlets send headlines like: "RTX 5090 reportedly dies after playing Youtube video"???
Posted on Reply
#24
Visible Noise
Luke Skywalker 8472Yes. It goes substantially over the spec.
Bullshit. Provide an authoritative source, not social media.
Posted on Reply
Add your own comment
Nov 22nd, 2026 01:00 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts