ASUS RTX 3080 TUF 10GB LHR - unpleasant RMA experience.

Status
Not open for further replies.
V

vlad_8011

Limp Gawd
Joined
Oct 13, 2020
Messages
137
Hi folks.
Just wanna warn some people who are going to choose ASUS GPU's in the future.
To be clear, GPU is overheating, in games like CyberPunk, Red Dead Redemption 2 i seen resoults just like in picture under.
278836036_491243892706306_7545366107905822866_n.jpg
278983297_476322954278974_5938695354173353136_n.jpg
After changing case, around 10 different fans placement i can only say its the GPU fault, not the case. So i decided to RMA it. I live in Poland, here, road to RMA looks like that - you send RMA to shop, where you bought the product, they test it, and send further to (in this example) ASUS.
So shop stated they made 2 DAYS FURMARK and detected no issue. So they sended it to ASUS, today, after 2 weeks, 1 and only buyer leaved (GPU is already set to sell) i got reply from shop, that they got reply from ASUS - THEY DETECT NO PROBLEM.
So, just to be clear, 112 degree on memory, in almost open case is normal?
BTW i described that GPU have 1 termopad without contact to heatsink, they found no issue.
111.jpg
P_20220314_173101.jpg
3333.jpg


Now GPU is going back to shop, then it will get back to me. I have 2 ways only now - spend more money (as the GPU wasnt expensive at all!) and fix termopads, or undervolt it.
Future games for sure gonna require more from GPU, if 2018 and 2020 titles are wasting that card, guess what it gonna be with Stalker 2 or GTA 6. I'm really dissapointed that such expensive GPU is done so poor. If manfucaturer have so loose client support, i will definetly not gonna their product again. Thats first time for graphics cards in my history
even MSI did the job done after 2nd RMA, but they even replaced GTX 1080 to 1080Ti - thats a big reward for troubles with service.
 

Attachments

  • 2222.jpg
    2222.jpg
    294.2 KB · Views: 0
No way can that be normal.. my Zotac 3080 would hit 92, and I thought that was hot. Recently upgraded to a EVGA 3080ti and memory will top out around 80-85.
 
Odigo said:
No way can that be normal.. my Zotac 3080 would hit 92, and I thought that was hot. Recently upgraded to a EVGA 3080ti and memory will top out around 80-85.
Click to expand...
Thats exacly right, its impossible that 122 degree on memory after 4 minutes and 23 seconds in furmark got this temps and its normal, but shop declares they "furmarked" it for 2 days and didnt found any problem. ASUS service said "No problems found, Test OK"
1.png
2.png
3.png

Fact of invalid thermalpad is totally ignored also.
Thats letter i added to GPU once it was send to RMA:
5.png

I say again - beware of ASUS warranty support. Had 3 troubles with their mobos in past 13 years, but i tought its over, staff and some internal rules has changed, but no, its exacly same, post PRL people and rules.
 
Last edited:
Why is your fan only at 64% when the GPU is that hot? is the fan on auto or did you manually set it?

Those temps (hot spot and junction) look about right with a core temp that high.
 
I don't understand why people don't just crack it open and identify the seating issue themselves instead of relying on the manufacturer which they do a poor job most of the time. If you are building your own system, you should be able to do this.
 
  • Like
Reactions: heslo
like this
Digital Viper-X- said:
Why is your fan only at 64% when the GPU is that hot? is the fan on auto or did you manually set it?

Those temps (hot spot and junction) look about right with a core temp that high.
Click to expand...
Well, i had custom fan curve set for whole time, not sure if is turn it back to default on THIS test, but i'm sure i did checked that also on default curve, and difference was 1 degree on core and RPM wasnt much higher (its quiet BIOS). It looks like there is problem with conductivity of components and heatsink, inside the case i got micro tornado in terms of air movement.
 
Rvenger said:
I don't understand why people don't just crack it open and identify the seating issue themselves instead of relying on the manufacturer which they do a poor job most of the time. If you are building your own system, you should be able to do this.
Click to expand...
That actually was first thing i wanted to do, BUT thermopads in Poland, with conductivity above 12W/mk is VERY, VERY expensive + you got to pay for shipping. Its very large cost :
6400zł - GPU
~550zł - Thermopads (not full set, i stopped adding them to basked once i reached 450zł, and it was half of set, didnt added shipping)
= 7100zł for GPU
Resell value is 4500zł.
And major reason of not opening it - i want to sell that crap. Any sign of opening the GPU (even different thermalpads, and there was one to replace) is lowering reselling price really hard. I had buyer, for 4700zł, but meantime he bought NEW card in store for that price.
 
vlad_8011 said:
Well, i had custom fan curve set for whole time, not sure if is turn it back to default on THIS test, but i'm sure i did checked that also on default curve, and difference was 1 degree on core and RPM wasnt much higher (its quiet BIOS). It looks like there is problem with conductivity of components and heatsink, inside the case i got micro tornado in terms of air movement.
Click to expand...

For shits and giggles, can you run your test again with 100% fan speed?(With furmark like in that first pic)
Tornado of air doesn't matter if there is no direct airflow over the fins and components
 
vlad_8011 said:
That actually was first thing i wanted to do, BUT thermopads in Poland, with conductivity above 12W/mk is VERY, VERY expensive + you got to pay for shipping. Its very large cost :
6400zł - GPU
~550zł - Thermopads (not full set, i stopped adding them to basked once i reached 450zł, and it was half of set, didnt added shipping)
= 7100zł for GPU
Resell value is 4500zł.
And major reason of not opening it - i want to sell that crap. I had buyer, for 4700zł, but meantime he bought NEW card in store for that price.
Click to expand...
Its likely a buckled or missed thermal pad. Also, the manufacturers go the lazy route and put thermal pad strips across the VRAM. You can carefully remove them and cut them in to individual sizes and you will have extra thermalpad for anywhere that is missed. I am betting its a buckled thermal pad thats offsetting the heatsink.
 
Digital Viper-X- said:
For shits and giggles, can you run your test again with 100% fan speed?(With furmark like in that first pic)
Tornado of air doesn't matter if there is no direct airflow over the fins and components
Click to expand...
First - GPU didn't returned to me, second - what it would change? I didn't bought silent card to use 100% fan speed, as I use speakers, not headset. Third - when card was new, problem wasn't existing.
Besides that on default fan speed of quiet bios had almost same temps.
 
Rvenger said:
I don't understand why people don't just crack it open and identify the seating issue themselves instead of relying on the manufacturer which they do a poor job most of the time. If you are building your own system, you should be able to do this.
Click to expand...
If it's out of warranty yes.. I wouldn't touch mine if new with a warranty (assuming I was in a different RMA situation than OP)
 
vlad_8011 said:
First - GPU didn't returned to me, second - what it would change? I didn't bought silent card to use 100% fan speed, as I use speakers, not headset. Third - when card was new, problem wasn't existing.
Besides that on default fan speed of quiet bios had almost same temps.
Click to expand...

The fact that the fans were running at 64% when the core was at 84C is what it would change :) that sounds wrong all together. I wonder if somewhere in your config the fan curve is messed up.
 
I can help you with this, but first off ASUS RMA service has been sucking for over a decade :D

The card itself has memory heatsinks that are separate from the GPU and relies on only one side of the memory heatsinks being in contact with the GPU heatsink/heatpipes. So the trick is for you to find/stack some thick thermal pads so the other side of the memory heatsink is in contact with the five exposed heat-pipes. This gave me an instant 10c+ drop in memory temperature from 100-105c down to 90-95c on any given day. Here is a picture attached, I didn't bother removing the two white stickers before applying thermal putty (stacked pads work too I just have a bunch of putty on hand).


3080TUF.png
 
legcramp said:
I can help you with this, but first off ASUS RMA service has been sucking for over a decade :D

The card itself has memory heatsinks that are separate from the GPU and relies on only one side of the memory heatsinks being in contact with the GPU heatsink/heatpipes. So the trick is for you to find/stack some thick thermal pads so the other side of the memory heatsink is in contact with the five exposed heat-pipes. This gave me an instant 10c+ drop in memory temperature from 100-105c down to 90-95c on any given day. Here is a picture attached, I didn't bother removing the two white stickers before applying thermal putty (stacked pads work too I just have a bunch of putty on hand).


View attachment 466384
Click to expand...
Yeah, I saw that trick on Reddit. Also saw copper sheets trick, but that all require to mod card, thus loosing warranty = lower reselling value.

Never had to mod any hardware in such low time from buy. I had GTX 1080Ti gaming X trio, which required only to repaster core after 2 YEARS.

Maybe that's a huge overstate, but i think ASUS is doing this on purpose. They want your GPU die faster, thus they didn't apply that one thermalpad. Casual users don't even use Afterburner, they don't care if something overheat or not, they don't monitor temps untill they hear blast of fans. Fans don't react to VRAM temps, but to core temps and power consumption. Imagine what gonna happen after, let's say 2 years. GDDR6X is fresh, it's not been tested on longevity, micron says 125 degree is permanent damage for those chips, but were they testing 110 degree for year or longer?

I'm surprised, that any AIB partner decided to produce anything while Nvidia gave them info month before production, they (partners) could wait a little bit longer and prepare proper cooling solution. If it's not on purpose, than what? Wasn't they releasing second revision of 3080 (LHR)? That was great opportunity to lay down second thermopad there.

I really hate how I was threated by ASUS and nothing positive comes in my mind. After I receive card I will definitely gonna record video of ambient temp, stress test in games and RMA it again - with prove that something is wrong, as right now shop and ASUS didn't proved that GPU is ok. Still wonder about adding consumer Ombudsman (using translator here, not sure about name in English) into messages.
 
vlad_8011 said:
I really hate how I was threated by ASUS and nothing positive comes in my mind.
Click to expand...
Forgive my ignorance, but all your email are to a MocnyKomputer.pl, that's not ASUS. It seems to me you are actually complaining about the computer shop you bought it from giving you bad customer service; not actually ASUS. Am I crazy here? :cautious:
 
You're right, i forgot add reply from ASUS consumer service... They also state this temps are fine.

4.png
 
So... the memory temperatures are high, but does that mean the card was crashing and unable to play games at its stock configuration?

If the card worked, but was hot, I don't see why that's an RMA-able offense as it is working as designed. If it's not stable, OK - that's valid.

The other question here that I can't tell from the potato picture of the current rig, but is it receiving enough airflow to prevent it from marinating in your case?
 
Is the memory temperature shown in gpu-z is the memory case temperature? Assuming it is and I'm not 100% sure if the asus card uses Micron GDDR6X memory (same logic applies), the memory operating temperature spec is 105°C max as state by Mircon. Certainly to get below that temperature, the asus engineer would have to presumably follows Micron thermal design notes as stated here or their internal design. There are other factors like ambient temperature (related to room temperature, PC case temperature and fan speed) that would need to work in tandem to ensure that the asus card can keep the memory case temperature below the datasheet spec even there is a need to ramp the fan up to the max 100%. I would imagine if asus rma dept tested your card and under their test condition (they need to tell you this) and found that the memory is operating within Micron spec, they probably will reply that they detect no problem at all. I think you need to request their testing result and conditions used, at least knowing what they meant by "detect no problem".
 
pututu said:
Is the memory temperature shown in gpu-z is the memory case temperature? Assuming it is and I'm not 100% sure if the asus card uses Micron GDDR6X memory (same logic applies), the memory operating temperature spec is 105°C max as state by Mircon. Certainly to get below that temperature, the asus engineer would have to presumably follows Micron thermal design notes as stated here or their internal design. There are other factors like ambient temperature (related to room temperature, PC case temperature and fan speed) that would need to work in tandem to ensure that the asus card can keep the memory case temperature below the datasheet spec even there is a need to ramp the fan up to the max 100%. I would imagine if asus rma dept tested your card and under their test condition (they need to tell you this) and found that the memory is operating within Micron spec, they probably will reply that they detect no problem at all. I think you need to request their testing result and conditions used, at least knowing what they meant by "detect no problem".
Click to expand...
Thats the problem. I have no direct contact about my RMA with their service, i can contact through shop.
The case i use is Lian Li O11 Air Mini, 2x 120mm na bottom (Noctua S12A PWM Black), 2 x 140mm on front (Noctua A14 PWM black), 1x Noctua S12A on back and 1 x NF12 black PWM on top. CPU cooler is Scythe Fuma 2 with 2 x Noctua S12A, with lowered PPT to 110W, corve optimize on allcore -15, throttling set to 75 degree.
Moreover i made custom panel for case, with big trapeze shape cutout
278983297_476322954278974_5938695354173353136_n.jpg

Case is standing next to allways opened windows, ambient is around 19-20 degree, never more as i like to keep it cold.
 
Rather than speculating (and I haven't dealt with asus rma yet, fingers crossed), try to obtain a copy of the asus temperature report and their testing condition.
 
Original RPM on this temperature is around 80, but i'll test iut specially for you on 100.

BTW this card never reached 100% rpm, even for first week when i was putting it on max load for testing (didnt even used afterburner then).

Edit : I think I found some way to limit overheating, together with issue - driver boosting algorithm. It looks like I give that GPU too much airflow - it was boosting as long as cooler wasn't limit - when it was hot all around it probably started to throttle. Just need to check and limit max boosting clock I guess, but that's driver issue - I didn't had this as card was new (used older driver back then). May it be possible?

Screenshot_20220422-224908249 (1).jpg

But that's still is only core clock temp, memory clock is on spec and it's overheating.
 
Last edited:
vlad_8011 said:
Original RPM on this temperature is around 80, but i'll test iut specially for you on 100.

BTW this card never reached 100% rpm, even for first week when i was putting it on max load for testing (didnt even used afterburner then).

Edit : I think I found some way to limit overheating, together with issue - driver boosting algorithm. It looks like I give that GPU too much airflow - it was boosting as long as cooler wasn't limit - when it was hot all around it probably started to throttle. Just need to check and limit max boosting clock I guess, but that's driver issue - I didn't had this as card was new (used older driver back then). May it be possible?

View attachment 466520
But that's still is only core clock temp, memory clock is on spec and it's overheating.
Click to expand...
I would do a driver clean as well, reset any settings on gpu tweak or after burner,.
Fan speed is based on temperature, not on load. So even if you are on 100% load, but temps are low, the fans will not spin up fast.
 
Schro said:
Are games crashing and/or is the system unstable?
Click to expand...
Even if they arent, 112C mem is way too hot. Not only are you losing performance due to thermal throttling, it will likely lead to memory degradation/failure (long term). Fwiw, furmark is a worst case scenario situation. I bet the memory only hits ~100c while gaming "normally" and this should really be what you look at, not a furmark stress test. The TUF is known for having poor mem cooling though, so this isn't that abnormal.

I don't know how/why thermal pads are so expensive in Poland, but that's really your only solution to this since Asus is telling you to pound sand. All you need is to get some good thermal pads (Gelid extreme or ultimate would be my suggestion), and add an additional pad like legcramp suggested. This would definitely shave off at least 10C. Then if needed, just undervolt the card.
 
xenium said:
Even if they arent, 112C mem is way too hot. Not only are you losing performance due to thermal throttling, it will likely lead to memory degradation/failure (long term). Fwiw, furmark is a worst case scenario situation. I bet the memory only hits ~100c while gaming "normally" and this should really be what you look at, not a furmark stress test. The TUF is known for having poor mem cooling though, so this isn't that abnormal.

I don't know how/why thermal pads are so expensive in Poland, but that's really your only solution to this since Asus is telling you to pound sand. All you need is to get some good thermal pads (Gelid extreme or ultimate would be my suggestion), and add an additional pad like legcramp suggested. This would definitely shave off at least 10C. Then if needed, just undervolt the card.
Click to expand...
I'm just trying to understand if he has a true issue with his card not working as designed (i.e. not stable while gaming), as that would warrant an RMA and indicate poor service without receiving a resolution. But if OP (who is unlikely to be a video card design engineer) is only have an issue with the measured temperature of the memory while the product as a whole is working, I would see no fault in how the RMA has been handled.
 
Just for reference I'm adding my 3080 TUF temps (non LHR though) while gaming

GPU temp max 64°
Mem temp max 88°
GPU hotspot max 76.5°

Max fanspeed was 76% so your fan speed does seem low especially given your temps.
 
Denpepe said:
Just for reference I'm adding my 3080 TUF temps (non LHR though) while gaming

GPU temp max 64°
Mem temp max 88°
GPU hotspot max 76.5°

Max fanspeed was 76% so your fan speed does seem low especially given your temps.
Click to expand...

Yeah, my TUF 3080 10GB v2 LHR has similar numbers. I don't remember the fan %, but it's around 2000rpm while gaming. (Flight Sim in 4K)

He's running Furmark with fan at 1500rpm.
 
I had an Asus GTX 570 that I had to RMA at one point. It was such a bad experience that for years I would buy extended warranty if given a chance rather than deal with that kind of hassle again.
Though in retrospect the only real hassle was trying to decipher the website in order to contact Asus. Everything else I have come to learn is pretty standard stuff.
 
Replace that missing pad.
Run the fan at a sensible speed.
Problem solved while keeping performance up.

Relying on throttling to keep temps down on ram is daft.
 
Schro said:
Are games crashing and/or is the system unstable?
Click to expand...
No, not yet
Digital Viper-X- said:
I would do a driver clean as well, reset any settings on gpu tweak or after burner,.
Fan speed is based on temperature, not on load. So even if you are on 100% load, but temps are low, the fans will not spin up fast.
Click to expand...
I had clean driver install, allways using DDU. Fan is based on core temperature and power draw - if you dont believe me, check what happen if you turn "allways prefer max performance" power setting in NV control panel. Fans will spin even on 30 degrees. On nvidia forums this is explained.
Nenu said:
Replace that missing pad.
Run the fan at a sensible speed.
Problem solved while keeping performance up.

Relying on throttling to keep temps down on ram is daft.
Click to expand...
I want to sell this GPU. If i open it, it will loose 1/2 of value. Besides that, non of review showed need for 100% fan spin, its waay too loud, and i rather undervolt it than replace fans after 6 months (if i wont sell it).
 
vlad_8011 said:
I want to sell this GPU. If i open it, it will loose 1/2 of value. Besides that, non of review showed need for 100% fan spin, its waay too loud, and i rather undervolt it than replace fans after 6 months (if i wont sell it).
Click to expand...
Just say you repasted it and keep the price up (at sensible level though compared to new).
Many people will be thankful its not an issue.

Why are you selling it?
 
vlad_8011 said:
No, not yet

I had clean driver install, allways using DDU. Fan is based on core temperature and power draw - if you dont believe me, check what happen if you turn "allways prefer max performance" power setting in NV control panel. Fans will spin even on 30 degrees. On nvidia forums this is explained.

I want to sell this GPU. If i open it, it will loose 1/2 of value. Besides that, non of review showed need for 100% fan spin, its waay too loud, and i rather undervolt it than replace fans after 6 months (if i wont sell it).
Click to expand...

Asus has a 0 fan setting, this probably disables it. I have GPUs running at 200+w with 0% fan sometimes because they're running at 20-30c. :) Power has no bearing on it, it is based on temp.
 
Nenu said:
Just say you repasted it and keep the price up (at sensible level though compared to new).
Many people will be thankful its not an issue.

Why are you selling it?
Click to expand...
I don't want any GPU with GDDR6X and power draw more than 300W.

Seeking for 6800XT again, I really regret I swapped to RTX this time, I did it just for test, and now I feel stuck. XFX Merc or ASUS TUF was my main question actually before RMA, but I decided to give it a try on ASUS service (3080) to fix it before sell. My decision was made 2 months ago. I had enough problems with NV, and this is my last PC build, don't want it to get overheat because construction problems.
 
I dont think in this case ASUS has any fault.
First, you say yourself, that gaming is fine, no issues.
Second, the shop and ASUS tested your card and found no issue.
Third, your fan curve is not appropiate.

I think your anger is misdirected.
If the heatsink bothers you, buy some damn pads and fix it.
You want some, give me your address and Ill mail you some. Ive got extras to spare.
 
According to the datasheet for the Micron memory modues that this card most likely has, the operating temperature range is supposed to be 0C to 105C. So, if it's really hitting 112, that's outside of the manufacturer's specified parameters for the board's components. You'd think Asus's RMA department would have been briefed on this, particularly since hot-running memory (now that the driver reports it) is a common complaint with 30 series cards.

I don't understand RMAing it over this, though, unless Poland has really terrible consumer protection, where the warranty really is void if you remove the heatsink and check the TIMs. Just Amazon some fresh thermal pads, install them carefully, and run an aggressive fan curve, and you should be good. Maybe also do what Legcramp said, and add some extra pads between the memory heat spreader and the actual heatsink. If silence is that important to you, you probably shouldn't have bought a card this thirsty, but you might be a good candidate for water cooling. A water jacket would definitely fix the hot-running memory.

Also, protip: You can avoid damaging the warranty sticker on most grahics cards by heating it gently with a hair dryer and then using fine tweezers to remove it. Put it somewhere safe, and then put it back when you're done, and the RMA department will probably never notice you messed with it. At least, that works with the little round ones you see on the screws. I've seen a few EVGA cards with a bigger one, and those are a lot harder.
 
Status
Not open for further replies.
Back
Top