RTX 3090 - Temperatures

EsotericMrE

E
Hi Folks,

I've been having some issues playing ED and I was hoping someone out there with an RTX 3090 might be able to help me out.

tl;dr - My RTX 3090 FE's Hot-Spot temperature reach to temps above 100c during gameplay. Eventually, the card cuts out.

I was wondering if someone out there w/ a RTX 3090 that runs their game in "ULTRA" settings might be willing to share about 30min or so of their temperature sensor readouts using HWiNFO64.

I'd like to be able to prove to the S.I. that built my machine that my temperatures are not normal and that the card needs a RMA.

THANKS!
 

Morbad

Morbad
I have access to some RTX 3080s, but they won't have comparable cooling, or any cooling problems, not any more anyway.

The 3080 and especially 3090 FE have notoriously bad GDDR6X temps, but hot spot temp is more rarely an issue.

On the RTX 3000 series, hotspot temp can be either the hottest part of the GPU itself, or the hottest voltage regulator reading. 100C+ here is a sign of an issue, especially if the differential is more than 25C or so above the GPU temp reading.

It's possible that there is a problem with the thermal interface material used. There have been cases where entire pads were left off. It could also be a mounting pressure problem. That said, even entirely 'normal' FE samples can be problematic...I ended up watercooling a 3080 FE setup I built for my brother because the GPU was overheating; nothing was technically wrong with it when I took it apart either, it's just an ill-conceived design.

Anyway, instability related to temperatures should be pretty clear signs of a build, or component selection, issue if this is a pre-built.
 

DeckerSolo

DeckerSolo
I have an RTX 3090 and run EDO at Ultra everything at 4k, and yes it runs very hot.
Unfortunately I'm not able to share accurate air cooled temps as I have very recently watercooled the card with the EK Quantum Vector Full Cover and the Active Backplate.
I remember the Memory temp was getting the hottest at about 98-100c, with the hot spot not far behind which is what prompted me to go to water cooling, but the card did not cut out/crash the game. Water cooling has brought the temps down to around 65-69c.
You could try undervolting the card via MSI afterburner and also make sure the case is well ventilated.
 
Last edited:

EsotericMrE

E
DeckerSolo said:
I have an RTX 3090 and run EDO at Ultra everything at 4k, and yes it runs very hot.
Unfortunately I'm not able to share accurate air cooled temps as I have very recently watercooled the card with the EK Quantum Vector Full Cover and the Active Backplate.
I remember the Memory temp were getting the hottest about 98-100c, with the hot spot not far behind which is what prompted me to go to water cooling, but the card did not cut out/crash the game. Water cooling has brought the temps down to around 65-69c.
You could try undervolting the card via MSI afterburner and also make sure the case is well ventilated.
Click to expand...
Thanks!

Unfortunately, I'm getting these temps with a custom water loop w/ two radiators. I'm still getting a delta of over 30-degrees. The GPU Temp & Memory Junction are within safe temps (60 - 70 at peak) but, the Hot Spot will hit 105c.
 

0threeleven

0threeleven
EsotericMrE said:
Thanks!

Unfortunately, I'm getting these temps with a custom water loop w/ two radiators. I'm still getting a delta of over 30-degrees. The GPU Temp & Memory Junction are within safe temps (60 - 70 at peak) but, the Hot Spot will hit 105c.
Click to expand...
You should NOT be getting those temps with water cooling. What are your system specs? Mainboard, CPU, make of the video card? Are you using software or hardware to control your fan speeds? What speed is your water pump at (LPH/GPH)? Is your reservoir topped off or is everything AIO?
 

Dragsham

Dragsham
Utubers are saying to put a cpu sink n fan across the memory chips that should have thermal tape over em if not put the thermal tape over each chip and fit a sink n fan ...problem solved.
Also undervolting but that's a whole new ballgame.
Afterburner app can apply a undervolt and fan curve.
Glad my paltry 3060 ain't having the same issues

o7
 

Morbad

Morbad
EsotericMrE said:
Unfortunately, I'm getting these temps with a custom water loop w/ two radiators. I'm still getting a delta of over 30-degrees. The GPU Temp & Memory Junction are within safe temps (60 - 70 at peak) but, the Hot Spot will hit 105c.
Click to expand...

Something is extremely wrong.

Whoever put the block on that card either didn't correctly apply the TIM, didn't use the correct thickness pads on the memory (preventing good core contact), didn't tighten the block all the way down, or left off a VRM pad entirely.
 

DeckerSolo

DeckerSolo
EsotericMrE said:
Thanks!

Unfortunately, I'm getting these temps with a custom water loop w/ two radiators. I'm still getting a delta of over 30-degrees. The GPU Temp & Memory Junction are within safe temps (60 - 70 at peak) but, the Hot Spot will hit 105c.
Click to expand...
You definately have something wrong with your water loop, there is no way if the waterblock is fitted correctly to you GPU you will be seeing high temps like that.
Yes you may have a rogue GPU, but more likely the thermal paste and/or thermal pads are not contacting your GPU components to the waterblock correctly.
Is the water pump even running.
I only have one 360mm radiator with three fans for my 3090, with no problem keeping temps in check.
Other commenting here also have low temps.
 
Last edited:

LumpTheNoLeggedDog

LumpTheNoLeggedDog
EsotericMrE said:
Hi Folks,

I've been having some issues playing ED and I was hoping someone out there with an RTX 3090 might be able to help me out.

tl;dr - My RTX 3090 FE's Hot-Spot temperature reach to temps above 100c during gameplay. Eventually, the card cuts out.

I was wondering if someone out there w/ a RTX 3090 that runs their game in "ULTRA" settings might be willing to share about 30min or so of their temperature sensor readouts using HWiNFO64.

I'd like to be able to prove to the S.I. that built my machine that my temperatures are not normal and that the card needs a RMA.

THANKS!
Click to expand...
Something is broken, you CAN'T have temps. this high no matter what you do, maybe you connected something wrong or your card or maybe MB is broken.
Is your power supply strong enough to support this card ? maybe fans on your card are getting not enough power, in your case you probably need 1000W PS
 
Last edited:

EsotericMrE

E
Thanks for the reply everyone... I didn't want to bog ya'll down with all the information but, here's the entire description of the issue with even more detailed information (it's ultimately what I sent to the system integrator that I bought the system from).

For starters, periodically, my monitor will go black and kick back on -- as if the driver was resetting. This seems to happen randomly during gameplay and while simply using the PC for non-gaming. It is also very quick, just a couple of seconds. Occasionally, it'll happen one or two times in a row. Some days are worse than others in terms of frequency.

The more egregious issue is that I'm losing all signal to my monitor at random times while playing Elite Dangerous. The computer will stay on and I am still able to hear audio (i.e. I can still talk to my friends on Discord). The only "fix" at the moment is to forcefully reboot my machine using the physical power button on the chassis.

--- TROUBLESHOOTING STEPS ---
- Clean install of Win11
-Swapped Displayport Cable
- 3DMark Benchmark and Stress Tests
---- Time Spy & Fire Strike Extreme - FAILED
------ Time Spy Results: https://www.3dmark.com/tsst/2675646
------ Fire Strike Results: https://www.3dmark.com/fsst/1944963
- OCCT Stress Tests
--- CPU & VRAM passed
- UNIGEN Superposition Benchmark & Stress Test (35min)
--- Passed all tests
- AIDA64 Benchmarks
--- Passed all tests

--- NOTES ---
The frustrating thing is that the benchmarks will push the temps but, I am unable to replicate the signal loss in anything other than Elite Dangerous. However, the flickering will happen randomly across all types of usage (gaming and non-gaming).

I was able to catch the signal loss event using HWiNFO64. Most of the GPU sensor readouts drop to zero when the signal loss occurs. Below is a Google Spreadsheet link to those results:

You will also notice on this crash (and all other stress tests) that the GPU Hot Spot temperature is dangerously high and can maintain temperature above 102c --- A delta of over 30-degrees.

----SEE ROW 954----
https://docs.google.com/spreadsheet...ouid=110636138285804410018&rtpof=true&sd=true

I have also included the additional tests in the containing folder on GDrive. There is a folder containing logs I captured in Windows10 and then logs after I performed a totally clean install of Windows 11.

ALL LOGS: https://drive.google.com/drive/folders/1mkrFuSIrn0QhDMaZFR1evrdA1Zx2hxVu?usp=sharing

You'll notice that the hot spot temperature rises above 100c in all the diagnostic and stress tests thrown at the card.

CPU: AMD RYZEN 9 5950X
GPU: RTX 3090 FE
MoBo: ASUS ROG Crosshair VIII Hero
PSU: EVGA 850W SuperNOVA P2
COOLING: Custom water cooling loop (CPU, GPU, VRAM w/ 1x 240 & 1x 360 Radiators)
MONITOR: ACER Z35P 120Hz
 

LumpTheNoLeggedDog

LumpTheNoLeggedDog
I think it is your PSU 850W. RTXs I believe have GPU temp target set to 80c by default. Flickering or loss of pic will be cause when one of your components will kick in their fans because of temp. raise and that causes less power (power shortage) to your GPU. When you play, all components need more power and your GPUs fans are not getting enough power at +80c your GPU fans should start running at 100%.
Get GPU Tweak II program if you don't have it yet. Monitor your voltage and fan RPM after your reach temps. over 80c.
Other possibility is your PCIe port is bad. Question, did you install your GPU on PCIe 1 x16 ? where it supposed to be
 
Last edited:

0threeleven

0threeleven
You have 2 radiators so you have 2 different loops more than likely, if it were a custom loop, it would be one large radiator (this is an educated guess). Running your GPU over 100°c for prolonged periods will ruin the video card. PSU will not cause the video card to overheat, what was stated before.

The GPU water block is either loose or there isn't any TIM (thermal interface material) like Artic Silver, that's what I use. Some all-in-one solutions use past or tape pre-applied while full coverage water blocks don't, the later cooling the GPU, memory and VRM's (voltage regulation modules) with tape on the VRM's and memory and past on the GPU chip.

Check your warranty, if you tear into it yourself or have a friend fix it, it could void the warranty. I wouldn't worry about all that other stuff except if your pump is running, fans are on and reservoir is full if non-AIO.
 

LumpTheNoLeggedDog

LumpTheNoLeggedDog
GPU like 3090 doesn't need additions cooling unless you will heavily overclock it. If he is running in default mode and hitting 100c, that means, something is wrong with his cards, PCIe is broken on his mother board or he is getting not enough power to supply his fans. Extra cooling may even worsen the situation because it needs power too and if that is the case then it is 100% PSU. I'm 100% sure adding more cooling wont solve this problem.
 

GJ51

GJ51
Air cooled ASUS 3090TUF

GPU 60 - 62C
Memory 80 - 82C
Hot Spot 71 - 72C

Under load @ 62% Utilization

3900X CPU
ASUS ROG Crosshair VIII Hero

Attached are 4 monitors

38" Alienware 3880X1600 @120Hz

Dell 27" 2560X1440
Dell 24" 1920X1200 (2) All 3 @60Hz

plus 2 usb monitors not running off the 3090

Running ED the board pulls between 315 - 345 W

Everything set Ultra or High
 
Last edited:

Morbad

Morbad
EsotericMrE said:
Thanks for the reply everyone... I didn't want to bog ya'll down with all the information but, here's the entire description of the issue with even more detailed information (it's ultimately what I sent to the system integrator that I bought the system from).

For starters, periodically, my monitor will go black and kick back on -- as if the driver was resetting. This seems to happen randomly during gameplay and while simply using the PC for non-gaming. It is also very quick, just a couple of seconds. Occasionally, it'll happen one or two times in a row. Some days are worse than others in terms of frequency.

The more egregious issue is that I'm losing all signal to my monitor at random times while playing Elite Dangerous. The computer will stay on and I am still able to hear audio (i.e. I can still talk to my friends on Discord). The only "fix" at the moment is to forcefully reboot my machine using the physical power button on the chassis.
Click to expand...

Yeah, it's overheating and unstable.

Any chance you can get some well-lit pictures of the card, edge on, so I can see where the waterblock interfaces with the card?

LumpTheNoLeggedDog said:
I think it is your PSU 850W.
Click to expand...

PSU shouldn't have anything to do with the temperatures the OP is seeing...radically high ripple can cause a slight increase in component temp, but nothing like this, and certainly not without showing other symptoms.

Something is wrong with the cooling. Ten-to-one the builder screwed up putting the block on...and the remaining odds are on some sort of flow rate problem. Either way, it's the system integrator's job to make right.

0threeleven said:
You have 2 radiators so you have 2 different loops more than likely, if it were a custom loop, it would be one large radiator (this is an educated guess).
Click to expand...

It's custom, so likely a single loop. Multiple loops requires more tubing and more pumps.

I've had as many as four radiators in a single loop, and I know people who have used more. Most cases don't have room for any single radiator large enough to cool multiple high-power components, so you put radiators where you can and link them all in series.
 
Last edited:

DeckerSolo

DeckerSolo
I'm not convinced the 850w PSU rating is your issue as I'm also using an 850w PSU with Ryzen 9 5950x and RTX 3090 FE no problems.
Of course the PSU could be faulty.
Is your motherboard BIOS upto date? I'm using as ASUS ROG Strix x570-f gaming, and ASUS have very recently posted and update to improve stability.
The excessive GPU temps with a waterblock fitted is definately an issue and points towards poor thermal compounds application and/or waterblock fitting.
I've got a temp probe in my water loop and can vary the radiator cooling fans based on water temp (resting water temp 41c), and can keep an RTX 3090 at 65/69c at full load with a 360mm three fan radiator no problems.
Monitor randomly flickering/going black could be a completely different problem not related to the GPU temps at all.
 
Last edited:

EsotericMrE

E
Thanks everyone!!! I really appreciate all your input and support!!

It looks like my next step is to take the machine back to the SI for them to take apart the GPU and inspect their work under the warranty.

If they give me any pushback, I'm glad I have temps from ya'll that I can point to.
 
You must log in or register to reply here.
Back
Top Bottom