RTX 2080 Super - crashing, freezing, artifacting

[Dinnercore]

Hello folks, I fear my 2080 super is already on its way out. Did anyone ever save an RTX from the artifacting or is a new card the only chance? Here is the story:

Bought this card with waterblock 9 months ago. It is a 2080 Super iChill from inno3D. Not the first brand I would have chosen but the only available at the time. It worked fine during the past months, played my games very well at 1440p 144hz.
Last week came the day when I decided to take apart my old system and switch a few things out. With that came cleaning the waterloop and rebuilding it. I noticed that there was some minor build-up of a blue residue from the liquid in the block so I thought it would be a good idea to clean it.
In order to do that I had to take it apart (which removed the warranty sticker and voided my warranty...). I cleaned it and remounted the cooler. This is not my first time mounting a waterblock to a GPU and I made sure that everything was in order.

Now after filling the loop and bleeding all bubbles out I started it and it seemed fine at first but after a day of playing some games I started having freezes in CS:GO where the screen would just stop but the game kept running. It happened first after 4 full matches, and then started to happen every 3 minutes. I closed CS:GO and wondered what the issue could be, started Star Citizen and got greeted by lighting artifacts (glowing spheres) almost instantly after starting with a hard crash right after. That is where my heart started to sink, I still remember with dread the whole RTX cards dying after launch. It appears that my card is just now after 9 months showing the signs. Exactly after I voided the warranty. I'm really upset with myself now.

Is there any chance I can get atleast some more use out of this card? Be it baking or hot-air to the memory? Yes I made sure the waterblock is seated correctly, yes I cleaned the die and used fresh Kryonaut, yes the pads were correctly aligned.

What's funny is that I can run some tests like the GPU-Z thingy no problem, even with OC (yes I tested my games at DEFAULT clocks):


(60°C is a lot for a watercooled card, it was like that from stock. I blame the waterblock design)

Something I did notice was some blue residue between the memory chips, like the stuff you see from thermal pads when they break apart except the pads on this unit are grey not blue. Maybe they used a different color on pre-testing? Maybe I got sold an RMA-card?

 

[Regeneration]

Try to RMA the card despite the removed warranty sticker. Claim you just cleaned the card.

You can try to underclock. Clean the card with alcohol and cotton swabs and then reapply thermal paste.

 

[Vya Domus]

Something I did notice was some blue residue between the memory chips

One can put two an two together and conclude that could be from the blue residue which means water might have somehow leaked on the card ?

 

[Deleted member 67555]

Turn off Image sharpening and Low Latency mode

 

[Caring1]

Warranty shouldn't be void just because the sticker is broken.

 

[delshay]

Warranty shouldn't be void just because the sticker is broken.

Does that not depend on what part of the world you live.

 

[OneMoar]

No sane vendor will void your warntee for removing the cooler if they do, don't buy from that vendor

gpu-z test does not put sufficient load on the card to provide any meaningful information

I would tear the card back apart and make sure you got all the funk off the memory chips that could certainly explain your issues

I am also thinking liquid damage maby you had a leak and didn't catch it, in which case the card needs to spend a good hour in a ultrasonic cleaner to have any shot of ever working again

 

[Vayra86]

Does that not depend on what part of the world you live.

There are always two realities: the perspective of customer relationship, common sense and good business, versus the one of hard lining on the rules with the goal to spend as little as possible on your customers.

Always appeal to the former. Great things happen and you get a good look at what sort of company you are truly dealing with.

 

[nguyen]

I would just take the WB out and do everything over again, who knows what you could do wrong cleaning the card the first time. Just make sure you clean every thing nicely, including all thermal paste in between the SMD and any cotton thread from the cotton buds that you used that may get stuck between the SMD.

I have cleaned and remount WB many times any still did something wrong half the time LOL, like forgot to peel the plastic cover on the thermal pads, or the old thermal paste can sometimes be electrically capacitive so it might cause some problem with stuck between the SMD. At one time I tried putting Conductonaut on the Shunt Resistors to increase the power limit, only for my 2080 Ti to shutter like mad but caused no damage to the card. I actually bought 2 2080 Ti at launch, put WB on them and they are still running fine (1 for me 1 for my friend build).

 

[Elysium]

The card works just fine for 9 months, then you decide to clean up your loop and suddenly it's not working just fine anymore? You did something wrong, go back over your process and figure it out.

 

[delshay]

There are always two realities: the perspective of customer relationship, common sense and good business, versus the one of hard lining on the rules with the goal to spend as little as possible on your customers.

Always appeal to the former. Great things happen and you get a good look at what sort of company you are truly dealing with.

Let's cut things short & not the long winded version, "it applies" in certain country's.

 

[nguyen]

Let's cut things short & not the long winded version, "it applies" in certain country's.

People in countries that got it good abuse the hell outta it though, often I have seen people straight up return GPU that are not "golden sample" until they get one, what a bunch of privileged...

 

[moproblems99]

I noticed that there was some minor build-up of a blue residue from the liquid in the block so I thought it would be a good idea to clean it.

Something I did notice was some blue residue between the memory chips

This is what would concern me. If this was water, it could be bad. Maybe a good cleaning will help you in that regard.

 

[Chomiq]

Warranty shouldn't be void just because the sticker is broken.

In US warranty void stickers are pretty much illegal, in EU - not so much.

 

[Dinnercore]

Thank you all for the replies. You all brought up some good points and I took the card apart again, checked everything and re-mounted. No use the card can't do 3D anymore.

One can put two an two together and conclude that could be from the blue residue which means water might have somehow leaked on the card ?

The blue stuff on the pcb was thermal-pad residue. It was soft, squishy like silikon, texture like rubber and went back into shape after squishing it. 100% sure there was no leak, I know what spills and leaks look like and this was not the cause.

I would tear the card back apart and make sure you got all the funk off the memory chips that could certainly explain your issues

I am also thinking liquid damage maby you had a leak and didn't catch it, in which case the card needs to spend a good hour in a ultrasonic cleaner to have any shot of ever working again

I cleaned it with alcohol and made sure no residue is left. There was no spill or leak. I remounted again and made sure everything is in order.

I would just take the WB out and do everything over again, who knows what you could do wrong cleaning the card the first time. Just make sure you clean every thing nicely, including all thermal paste in between the SMD and any cotton thread from the cotton buds that you used that may get stuck between the SMD.

I have cleaned and remount WB many times any still did something wrong half the time LOL, like forgot to peel the plastic cover on the thermal pads, or the old thermal paste can sometimes be electrically capacitive so it might cause some problem with stuck between the SMD. At one time I tried putting Conductonaut on the Shunt Resistors to increase the power limit, only for my 2080 Ti to shutter like mad but caused no damage to the card. I actually bought 2 2080 Ti at launch, put WB on them and they are still running fine (1 for me 1 for my friend build).

I was the first to question myself, I did watercooling for dual-pcb cards in the past that required upwards of 20 individual thermal pads. This is not the first time I'm doing this and I have a mental procedure to avoid making mistakes like leaving the foil on a pad. In this case however the pads are the original ones that came with the card, no peeling required.

The card works just fine for 9 months, then you decide to clean up your loop and suddenly it's not working just fine anymore? You did something wrong, go back over your process and figure it out.

I thought the same but can't find any fault with my paste application or cleaning and re-mounting. Like I mentioned above I checked it again and its good. Also this is the first time the card was being tested under full load, since the past 9 months it was in a system with an X5690 CPU and not the 10900X I use now. So maybe my card was faulty from the beginning but it just now shows.

This is what would concern me. If this was water, it could be bad. Maybe a good cleaning will help you in that regard.

The pcb is clean, I wiped everything with alcohol and made sure its good. No leaks or spills.



The card now can't even hold 10 seconds of 3D. I tried with OC, stock and with an 'underclock', Valley crashes within seconds now:

Every wave is a run that crashed. This card is dead and I don't know why. There was no heat issue, I treated everything like a raw egg since this is actually a new and valuable card. I was far less careful with my watercooled Vega64 back in the day, which still works fine. And that is a 400W card not a 200W.
Since it ran for multiple hours before and I ran 2 hours of Valley after assembly yesterday it definitly got worse.

Second card to die on me in my lifetime, first was my GTX660 but after 5 years of use.

As for the warranty, I will try to contact the manufacturer and the retailer I got it from but I have little hope. Sticker broken means RIP warranty in Germany. That goes for most brands big or small, only MSI and EVGA allow for cleaning and changing thermal interfaces.

 

[Vayra86]

In US warranty void stickers are pretty much illegal, in EU - not so much.

Will not hold up in court most likely, which was the reasoning behind my earlier comment. You never get it easily, so ask nicely and appeal to proper business practices.

Or take the mentioned short version and lose by default, to each his own eh @delshay

 

[nguyen]

@Dinnercore Have you tried putting in a negative core clock offset ? My friend's Strix 2070 Super was actually not stable at its stock boost clock (2070mhz) so I had to put -50mhz and it works perfectly. You mentioned undervolting but that could also mean overclocking + undervolting (higher clock at lower voltages).
Do the same for VRAM also, try -200mhz first.
If the card worked before with a slower CPU doesn't means it is stable at that clock, a faster CPU will stress the GPU more (higher FPS) thus exposing the instability the GPU may have...

 

[Dinnercore]

You mentioned undervolting but that could also mean overclocking + undervolting (higher clock at lower voltages).

I said underclocking, I left it at stock powerlimit and applied negative clock offset. I tried -77 core and -266 memory it still fails within 10 seconds each time. Temp says 44°C max.

 

[nguyen]

Well from your last image, the GPU temp reach 56C in matter of seconds despite being watercooled means either you mount the WB wrong or the pump is dead (or not connected)...

 

[OneMoar]

Well from your last image, the GPU temp reach 56C in matter of seconds despite being watercooled means either you mount the WB wrong or the pump is dead (or not connected)...

Uh no that is perfectly normal for a 2080 full-clock Idle temps in the 60's are not unheard of, and that is not warm enough to be causing a problem anyway so yea ... nope

dead card is dead sadly
F

 

[Poul-erik]

From denmark.




If you have a slot more for a graphics card on your motherboard, you can try it in this, it might only be x8,

maybe it will work, and then it's your motherboard that's wrong, (I think)

 

[xtreemchaos]

RMA it, thay should be ok with the W/seal breached, thay tend to look at it as Q/C morethan a antitamper seal nowa days.

 

[nguyen]

Uh no that is perfectly normal for a 2080 full-clock Idle temps in the 60's are not unheard of, and that is not warm enough to be causing a problem anyway so yea ... nope

dead card is dead sadly
F

He has a watercooled 2080 Super, the idle temp in his image is 25C
I have a 2080 Ti watercooled and the temp rise to a maximum of 44C with maximum load in 20 minutes, not in seconds like his. Seems like the WB is not making proper contact...
Look closely at the sudden rise in GPU temp, no watercooled GPU would react that way

 

[OneMoar]

He has a watercooled 2080 Super, the idle temp in his image is 25C
I have a 2080 Ti watercooled and the temp rise to a maximum of 44C with maximum load in 20 minutes, not in seconds like his. Definitely something wrong with the pump...

doesn't fking matter the card is artifacting at 60c

its a low-binned Inno3d card I am not at all surprised its dead

 

[nguyen]

doesn't fking matter the card is artifacting at 60c

There are other components that overheat too you know. Either way the mounting of the waterblock is definitely wrong, so a remount is in order...and yeah if the VRM is fried then there is nothing to salvage...