News Igor's Lab uncovers 'hotspot issue' affecting all RTX 50-series GPUs — says it could compromise graphics card longevity

[CelicaGT]

There is a whole list of entry level “mistakes” on the 5000 series cards. Letslist them:

-Inability to monitor power
-Inability to manage power
-Lack of adequate safety features
-Reduction of safety features which were already inadequate (from 3 shunt resistors down to 1)
-power cables with almost no headroom in terms of spec.
-ROP-Gate
-inadequate quality control

And now, hot spots. It’s almost like Nvidia was hoping your GPU might not last that long beyond warranty, and you’d have to buy a new one.

Not entry level. These are "Cost reduction" measures. You know, in order to deliver a quality product to the consumer at a reasonable price while delivering a wonderful return to our faithful stakeholders!

(Sarcasm, for those who don't read between the lines)

 

[YSCCC]

I have an ASUS TUF RTX 5080 and I was playing MSFS 2024 which pushes both CPU and GPU pretty hard. The 5080 topped out at 68c and memory was just a tad lower. I will keep an eye out again.

I think the hotspot issue is on the PCB power delivery portion, not memory and GPU temp reported by software. So your chips might be fine but the VRMs cook themselves

 

[palladin9479]

Oh noes, the 50 series card having power delivery issues .... say it isn't so.

That 12VHPWR cable implementation kinda demonstrates how bad poorly designed the power delivery system is.

 

[saunupe1911]

I pulled apart mine to change paste and switch from pads to putty and can say everything is well spread out and covered with thermal pads. The power delivery on those cards is also way overbuilt since they have a standard 400W profile and support 500W IIRC. I think the only thing you might be concerned with is just standard material degrading over time.

This is why I never considered selling my EVGA 3080 once I got my 4090 FE. Helluva piece of hardware that's now used in my always on Plex server

 

[jesselafantaisie]

Same. I'd be interested to know how my EVGA 3080 FTW 3 is hot spotting (if at all) so I can see about trying to reduce any heat concerns.

Evga precision app shows all the temps

 

[atomicWAR]

This is getting sad. Nvidia mindshare is eroding in real time. I thought I might consider the 60 series in a couple years but now it may wait till the 70 series to see if Nvidia has straightened its act up otherwise I'll be looking at AMD, at least when every they can release a card better than my 4090 say +60% or so. I may be waiting for awhile...

 

[Heat_Fan89]

I think the hotspot issue is on the PCB power delivery portion, not memory and GPU temp reported by software. So your chips might be fine but the VRMs cook themselves

Hopefully ASUS boards were designed to counter that. In their marketing material they reference: "Certified military-grade TUF chokes, MOSFETs and capacitors for stable and durable GPU power. A premium phase-change GPU thermal pad melts to fill gaps, ensuring superior heat dissipation."

 

[YSCCC]

Hopefully ASUS boards were designed to counter that. In their marketing material they reference: "Certified military-grade TUF chokes, MOSFETs and capacitors for stable and durable GPU power. A premium phase-change GPU thermal pad melts to fill gaps, ensuring superior heat dissipation."

https://www.techpowerup.com/review/asus-geforce-rtx-5080-tuf-oc/4.html
from the teardown it appears it have at least pads for the power delivery components even to the backplate, but if it's like 90C-75C, it's still likely to not survive real long for the cost of the card.

Combining with the 12V 2x6 melty connector, missing ROPs and still unstable driver is really lame for a $999 MSRP card costing some $1300+ for street price

 

[Heat_Fan89]

https://www.techpowerup.com/review/asus-geforce-rtx-5080-tuf-oc/4.html
from the teardown it appears it have at least pads for the power delivery components even to the backplate, but if it's like 90C-75C, it's still likely to not survive real long for the cost of the card.

Combining with the 12V 2x6 melty connector, missing ROPs and still unstable driver is really lame for a $999 MSRP card costing some $1300+ for street price

The problem is that Nvidia, nuked the Hotspot sensor on those cards, per Linus (LTT). I noticed that when I was monitoring the temps, I noticed the Hotspot sensor was missing, and I never paid attention to it until this article. So we don't know whether they are in a normal range or outside of it.

My card cost $1484.99 but that is ASUS' MSRP on their website.

 

[Mr Majestyk]

50xx Super series needed in record time to fix the litany of problems. Hardware sucks and software still sucks, a perfect all-rounder for your gaming nirvana.

9070XT supply is trending up in Australia, will be buying one in June.

 

[Notton]

These are PNY and Palit cards, so my expectation weren't as high as those VRM temps were.
I'd like to see how Asus/MSI/Gigabyte and Nvidia FE fair.
If the problem is limited to PNY and Palit, it's as easy as avoiding their cards.
But if it affects everyone, you know it's a design flaw from Nvidia.

 

[endocine]

There is a whole list of entry level “mistakes” on the 5000 series cards. Let’s list them:

-Inability to monitor power
-Inability to manage power
-Lack of adequate safety features
-Reduction of safety features which were already inadequate (from 3 shunt resistors down to 1)
-power cables with almost no headroom in terms of spec.
-ROP-Gate
-inadequate quality control

And now, hot spots. It’s almost like Nvidia was hoping your GPU might not last that long beyond warranty, and you’d have to buy a new one.

Also,
-the drivers are buggy,
-the product specs are lame, insufficient vram, overpriced for what they are, no real gen on gen performance improvement. the 5080 should have had 24GB, especially for the asking price, if not 32GB

This is the first "generation" of nvidia gpu in 20 years that didnt buy or want

 

[Diceman_2037]

i think this senile old goof needs to retire and stop causing alarm over nothing, the components in the area are rated to 150c.

 

[YSCCC]

The problem is that Nvidia, nuked the Hotspot sensor on those cards, per Linus (LTT). I noticed that when I was monitoring the temps, I noticed the Hotspot sensor was missing, and I never paid attention to it until this article. So we don't know whether they are in a normal range or outside of it.

My card cost $1484.99 but that is ASUS' MSRP on their website.

hum~ actually, the hotspot sensor was the GPU die hotspot, which of course itself is a problem, but if you read the igor's lab article, they are talking about the power supply rail/VRM region temperature, which is another issue and can be more serious.

It's a Sxxt show til now, looking like every single part is broken yet charges you an arm and a leg for it

 

[VizzieTheViz]

Ah yes, Pascal and EVGA. End of an era, and Pascal was a mistake NVIDIA will not likely make again. Great card btw, hold onto it.

I have an MSI 980ti in my old gaming pc that’s still doing fine.

N=1 and I’m not buying a 50-series card with all its potential issues but NVidia made a lot of really good stuff too.

 

[mitch074]

I have an MSI 980ti in my old gaming pc that’s still doing fine.

N=1 and I’m not buying a 50-series card with all its potential issues but NVidia made a lot of really good stuff too.

Now that I've changed most parts in my main rig, I'm putting back together the equivalent of my 2016 gaming rig - just for kicks, I tried running Black Myth: Wukong on it.
Hot damn, it actually RAN ! Playable ! Core i5 4670K@stock, 16 Gb DDR3, Radeon RX480 8 Gb reference design (with a new cooler, because noise).
All it's missing is a better CPU cooler so that I can OC the CPU again, but still...
They don't make'em like they used ta.

 

[orodruin]

I have a Gainward Phoenix RTX5070TI graphics card. I'm not sure if I should worry or not.

 

[Heat_Fan89]

i think this senile old goof needs to retire and stop causing alarm over nothing, the components in the area are rated to 150c.

Where did you find that information?

 

[Heat_Fan89]

hum~ actually, the hotspot sensor was the GPU die hotspot, which of course itself is a problem, but if you read the igor's lab article, they are talking about the power supply rail/VRM region temperature, which is another issue and can be more serious.

It's a Sxxt show til now, looking like every single part is broken yet charges you an arm and a leg for it

On my 2070 Super, HWMon shows the Hotspot as the "memory". However I am puzzled that the link you provided to tech-power-up shows the layout of the PCB and the VRM's along with the all the other components are all under the same heatsink and have thermal pads attached. Because even if the VRM's were running much hotter than the rest of the components it would raise the other temps as well.

I just hope it's all hyperbole on the part of Igor's Lab. Because if this is true then Nvidia will take a massive hit to their reputation. I know I would steer clear of their products if my 5080 gave up the ghost in 2-3 years.

 

[Heat_Fan89]

These are PNY and Palit cards, so my expectation weren't as high as those VRM temps were.
I'd like to see how Asus/MSI/Gigabyte and Nvidia FE fair.
If the problem is limited to PNY and Palit, it's as easy as avoiding their cards.
But if it affects everyone, you know it's a design flaw from Nvidia.

I was thinking the same before you mentioned this !

 

[Heat_Fan89]

I just checked the temps on my ASUS TUF 5080 and it idles at 28c - GPU and 40c - memory. Later today I am going to run MSFS 2024 and monitor the temps.

 

[VizzieTheViz]

I just checked the temps on my ASUS TUF 5080 and it idles at 28c - GPU and 40c - memory. Later today I am going to run MSFS 2024 and monitor the temps.

Does that card actually report the temps of the spots mentioned in this article?

The article states the reported temps were fine but the power delivery components were quite high and that there’s no reporting from the cards on those temps.

 

[DS426]

I have an ASUS TUF RTX 5080 and I was playing MSFS 2024 which pushes both CPU and GPU pretty hard. The 5080 topped out at 68c and memory was just a tad lower. I will keep an eye out again.

We're talking about temps at the power delivery area, not the cores and mem.

 

[vanadiel007]

In other news: my 7900X TX Merc 310 Black edition is still happily providing me with great gaming speeds, and I got the card back in the day for almost MSRP because everyone wanted the RTX 4000 cards and shunned away from the AMD cards.

Well, last laugh is on me!

 

[Heat_Fan89]

Does that card actually report the temps of the spots mentioned in this article?

The article states the reported temps were fine but the power delivery components were quite high and that there’s no reporting from the cards on those temps.

No, it just lists GPU and Mem. The hotspot sensor is no longer available.