System restarts under heavy GPU load? - Overclock.net - An Overclocking Community

Forum Jump: 

System restarts under heavy GPU load?

Reply
 
Thread Tools
post #1 of 13 (permalink) Old 07-15-2019, 08:28 PM - Thread Starter
New to Overclock.net
 
1096bimu's Avatar
 
Join Date: May 2015
Posts: 23
Rep: 0
System restarts under heavy GPU load?

Hi, I recently got an MSI 2080ti Sea Hawk EK X, it worked fine but temps were quite high, I could get it to like 35 degrees over liquid temperature so I opened it up for repaste with liquid metal.
The first time I did this temps didn't improve at all, if anything it got a bit worse. And I had problems of the computer just resets with no warning and no error after the restart, except for Windows saying it was an unexpected shutdown. My theory was that I didn't apply enough liquid metal, and some parts of the die was not actually getting contact, they over heated, triggered some temperature sensor and caused system reset.

So I opened it up again, added a lot more liquid metal. Of course I know this is conductive so I basically built a wall around the die with non-conductive thermal paste, completely covering all the resistors around the die. I also added additional littler washers for more mounting pressure and pressed the block harder into the die. This time temps were much much better, I'm seeing like 10 degrees delta even at 400w, and no resets during gaming for at least 2 hours.

However I noticed that I get instant system reset, the same as before, if I run Userbench? as soon as the graphics test comes up it's instant system reset? I then tried furmark and it too caused system reset in a few minutes, where as it didn't a couple of hours before? I'm pretty sure I did ran furmark for like 5 minutes before and it didn't reset. Also the game I was playing was Total War TK, it could pull 300w just like furmark, and again it didn't reset the system for two hours.

I'm going to try the game again now, maybe something changed inside the graphics card that's starting to cause problems where as it didn't before. Or could it be a software issue? Cuz I mean the Userbench crash is just instant and guaranteed even from a cold machine. Except I could run it no problems with -1000mhz clocks and 50% power, so it absolutely has to do with power?

BTW I'm running a Seasonic 600w 80+ Titanium fanless, it does have airflow in the case I made sure of that. It's a bit low compared to the tradition but I mean the graphics tests that are giving me problems don't stress the CPU that much, if anything you'd expect the games that also stress the CPU to be more problematic on the PSU right?

Last edited by 1096bimu; 07-15-2019 at 08:39 PM.
1096bimu is offline  
Sponsored Links
Advertisement
 
post #2 of 13 (permalink) Old 07-15-2019, 08:42 PM
Overclocker
 
JackCY's Avatar
 
Join Date: Jun 2014
Posts: 9,758
Rep: 330 (Unique: 237)
With the autoboost of GPUs it may be boosting way too high on cold temps.
Other than that you messed it up with LM or have some dud running hot on water cooling.

You do need plenty power for these cards and depends on rest of your system. 600W... probably doable but it depends.
If it runs fine with negative clocks offset then tada you got a dud that's boosting too high under water.
JackCY is offline  
post #3 of 13 (permalink) Old 07-15-2019, 08:44 PM - Thread Starter
New to Overclock.net
 
1096bimu's Avatar
 
Join Date: May 2015
Posts: 23
Rep: 0
Is the symptom of unstable clocks a sudden reset of the entire system?
I just tried Furmark again, it's perfectly stable at 100% power. If I shorted something with LM I don't think power should matter since arcing distance would be determined by voltage not current?
I know the PSU is a weird choice, I kind of have a fetish for this particular one, but I mean it's rated for 600w continuous from the 12V rail, instantaneous load tolerance should be significantly higher

Last edited by 1096bimu; 07-15-2019 at 08:49 PM.
1096bimu is offline  
Sponsored Links
Advertisement
 
post #4 of 13 (permalink) Old 07-16-2019, 03:30 AM
New to Overclock.net
 
Zfast4y0u's Avatar
 
Join Date: Dec 2017
Posts: 294
Rep: 1 (Unique: 1)
to me it seams ur block dosent sit correctly on the gpu and causes parts of core or something else to overheat. that can be dangerous as well, cause cores inside the chip could instantly die. maybe check did they put correct thermal pad thicknesses on vram chips and vrm. also, dont bother with LM on water, i doubt u will see any improvements, better try with paste first to establish does block sit correctly or not, also u can see how paste spread.
Zfast4y0u is offline  
post #5 of 13 (permalink) Old 07-16-2019, 06:17 AM - Thread Starter
New to Overclock.net
 
1096bimu's Avatar
 
Join Date: May 2015
Posts: 23
Rep: 0
That's what I thought too, problem is the thermal pads on the VRAMs are already super thin, there's nothing I can do about those except to remove them and put thermal paste instead?
The VRM pads are thick but they're really far away from the die area, I think the flexibility of the PCB should be enough to tolerate it.
But then I think this is the only sensible explanation since it didn't do this before I repasted the card, it had tones of thermal paste in it and that definitely guaranteed contact, except thermals was just **** for a water cooled card.
Damn I'm never buying another pre-installed block again, even though it's EK branded it's significantly lower quality than actual EK blocks, and the styling makes me sick, only bought it because it was cheap.
1096bimu is offline  
post #6 of 13 (permalink) Old 07-16-2019, 10:21 AM
New to Overclock.net
 
Zfast4y0u's Avatar
 
Join Date: Dec 2017
Posts: 294
Rep: 1 (Unique: 1)
Quote: Originally Posted by 1096bimu View Post
That's what I thought too, problem is the thermal pads on the VRAMs are already super thin, there's nothing I can do about those except to remove them and put thermal paste instead?
The VRM pads are thick but they're really far away from the die area, I think the flexibility of the PCB should be enough to tolerate it.
But then I think this is the only sensible explanation since it didn't do this before I repasted the card, it had tones of thermal paste in it and that definitely guaranteed contact, except thermals was just **** for a water cooled card.
Damn I'm never buying another pre-installed block again, even though it's EK branded it's significantly lower quality than actual EK blocks, and the styling makes me sick, only bought it because it was cheap.
seams pad thicknesses are okay, usually vrm pads are 1mm thick and vram 0,5mm. do not put paste on vram/vrm, it will be a mess.

do temps spike up like 20c on load as soon u put stress on card, or they rise slowly? from my experience when there is too little TIM/contact between die and water block, temps tend to spike quite a bit even when opening browser only, if there is proper contact and enough paste to cover whole die evenly, there should be no temp spikes.
Zfast4y0u is offline  
post #7 of 13 (permalink) Old 07-16-2019, 03:05 PM - Thread Starter
New to Overclock.net
 
1096bimu's Avatar
 
Join Date: May 2015
Posts: 23
Rep: 0
I know what you mean, it spikes before I repasted it, that was why I wanted to repaste it.

Problem is this card has a bad design, it’s actually impossible for the die to make contact with the block when there is no TIM, because the standoffs are a bit too tall. What the stock card does is they just dump lots of paste in there so you beige the gap with paste, but paste has poor conductivity so you get high temps and high temperature delta between the core and the liquid.

I opened up the card again, messed with the TIM again, seemed to have solved the problem but it seems to get worse since a few hours later now, I’m getting restarts again. The problem is with repaste, displays temperatures are actually really good, 50 degrees at 400w, yet it still restarts.
1096bimu is offline  
post #8 of 13 (permalink) Old 07-16-2019, 03:42 PM
New to Overclock.net
 
Zfast4y0u's Avatar
 
Join Date: Dec 2017
Posts: 294
Rep: 1 (Unique: 1)
if you can, rma that thing.
Zfast4y0u is offline  
post #9 of 13 (permalink) Old 07-16-2019, 04:00 PM - Thread Starter
New to Overclock.net
 
1096bimu's Avatar
 
Join Date: May 2015
Posts: 23
Rep: 0
I can't, taking off the block voids the warranty.

However I'm now starting to suspect it's not the graphics card at all, it's the PSU. Even with the 80+ Titanium rating it's going to be doing 70w of heating at full load, I suspect I simply don't have enough airflow for that. I found that warming the system makes it worse, this is why it seems to work fine every time I put it back together, and also immediately after it restarts, it's much easier to trigger another restart. Where as if I turn it off and leave it for a while, it's stable again.

If there really was a piece of the chip not making contact with the block, it's a tiny piece of silicone and shouldn't be affected by the rest of the system, it'll heat up and trigger protection immediately. It would make more sense if it was the PSU's 12v rails getting too hot and it simply cutting my power. I'm going to put a giant fan on the PSU and see if that solve it.

Yea that kind of defeats the purpose of a "fanless" PSU...

Last edited by 1096bimu; 07-16-2019 at 04:07 PM.
1096bimu is offline  
post #10 of 13 (permalink) Old 07-16-2019, 04:36 PM
New to Overclock.net
 
Zfast4y0u's Avatar
 
Join Date: Dec 2017
Posts: 294
Rep: 1 (Unique: 1)
Quote: Originally Posted by 1096bimu View Post
I can't, taking off the block voids the warranty.

However I'm now starting to suspect it's not the graphics card at all, it's the PSU. Even with the 80+ Titanium rating it's going to be doing 70w of heating at full load, I suspect I simply don't have enough airflow for that. I found that warming the system makes it worse, this is why it seems to work fine every time I put it back together, and also immediately after it restarts, it's much easier to trigger another restart. Where as if I turn it off and leave it for a while, it's stable again.

If there really was a piece of the chip not making contact with the block, it's a tiny piece of silicone and shouldn't be affected by the rest of the system, it'll heat up and trigger protection immediately. It would make more sense if it was the PSU's 12v rails getting too hot and it simply cutting my power. I'm going to put a giant fan on the PSU and see if that solve it.

Yea that kind of defeats the purpose of a "fanless" PSU...
i know right, i have corsair rm1000x that has fan build into it, but it never spins up, psu shroud was running hot to touch even when pc is just idle. was annoying heck out of me, took it apart, ripped that **** fan out of psu, put mine that i can control and im done with it. psu is ice cold now with constant cooling over internals in it, like it should be in any proper made psu out there, and no u cant hear fan spinning, i hear just HDD noise ^^ and its rated 1000w on 12v rail @50c.

wanted to tell u check psu if u can, swap it with other one, but thought u had it covered already.
Zfast4y0u is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off