Thermal Pad Failure

Message boards : Number crunching : Thermal Pad Failure
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile RueiKe Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 460
Credit: 275,937,024
RAC: 339,812
Taiwan
Message 1984245 - Posted: 9 Mar 2019, 4:58:52 UTC

My main quad Vega64 machine has been unstable for a while and one card finally failed. I removed it and found that the thermal pads seeped liquid that covered a significant part of the card including the contacts to the PCIe slots. I used the thermal pads provided by EK when I purchased the waterblocks about 1.5 years ago.

I should probably change Thermal Pads for all 4 cards and maybe even other EK Waterblock installations. Any recommendation on higher quality thermal pads?

Also, The 3 remaining cards are still flaky with one error in dmesg "Invalid PCI ROM header signature" so maybe other cards and the MB are affected. I will seek a good approach to clean the PCBs.
ID: 1984245 · Report as offensive
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 183
Credit: 16,981,787
RAC: 47,966
Antarctica
Message 1984247 - Posted: 9 Mar 2019, 5:16:36 UTC - in response to Message 1984245.  

thermal grizzly. but atm their site has been hack i think.
ID: 1984247 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2751
Credit: 470,361,807
RAC: 847,332
Canada
Message 1984251 - Posted: 9 Mar 2019, 6:00:39 UTC

I believe Fujipoly Extreme is one of the high end thermal pads. And worth more than gold, lol.
ID: 1984251 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 8302
Credit: 718,782,831
RAC: 1,742,847
United States
Message 1984252 - Posted: 9 Mar 2019, 7:16:13 UTC - in response to Message 1984251.  
Last modified: 9 Mar 2019, 7:19:31 UTC

The FujiPoly pads I bought from Amazon were about $16. Enough to cover the gpu die. Would need a larger size or another package to do the VRMs and memory dies.

https://www.amazon.com/gp/product/B00MQ0ITXW/ref=oh_aui_search_asin_title?ie=UTF8&psc=1
https://www.amazon.com/gp/product/B00MQ0COHE/ref=oh_aui_search_asin_title?ie=UTF8&psc=1

I've seen the seepage of the silicone oils from the OEM thermal pads on just about every gpu card I've owned. I don't believe they are of the same quality as the FujiPoly pads. The ones on the repaired Intel motherboard have been in place for about a year and no sign of seepage on them. But I don't think that is long enough. I see seepage on gpu cards after about 3 years.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1984252 · Report as offensive
Profile RueiKe Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 460
Credit: 275,937,024
RAC: 339,812
Taiwan
Message 1984255 - Posted: 9 Mar 2019, 8:20:03 UTC
Last modified: 9 Mar 2019, 8:38:08 UTC

Thanks for the thermal pad recommendations. I will have to see what I can get here in Taiwan. The FujiPoly looks most interesting, but Grizzly may be easier to acquire. I did get feedback on IG that the liquid is likely Silicone oil, which is non-conductive and should not cause a problem, but this card was pulled out of the system because it would cause the system to crash within seconds if it were enabled. I am still having problems on the system with the remaining 3 cards. dmesg shows the following error at boot:
amdgpu 0000:47:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff

And compute intermittently hangs on all of the cards (not at the same time). So I am concerned that the fluid went down into the PCIe slot. I plan to swap out the MB as soon as I get new thermal pads on all of the cards. I tried to clean the residue off with isopropyl alcohol, but the stain on the board did not reduce. Anyone ever see the MB cleaning video by der8aur where he uses the dishwasher??
ID: 1984255 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2751
Credit: 470,361,807
RAC: 847,332
Canada
Message 1984257 - Posted: 9 Mar 2019, 8:31:45 UTC

Ether areole (engine starting fluid) works pretty darn good for cleaning a lot of things and evaporates almost instantly. A WD40 nozzle fits most cans so you have the hose for more accurate spray.

BUT use OUTSIDE. It is very explosive!
ID: 1984257 · Report as offensive
Profile Wiggo "Democratic Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 16316
Credit: 210,265,453
RAC: 144,762
Australia
Message 1984261 - Posted: 9 Mar 2019, 10:11:12 UTC

I use ethyl alcohol myself.

Cheers.
ID: 1984261 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1497
Credit: 319,007,099
RAC: 163,703
United States
Message 1984266 - Posted: 9 Mar 2019, 11:07:42 UTC - in response to Message 1984261.  

I use ethyl alcohol myself.

Cheers.


Ahhh! But do you use it IN yourself?
ID: 1984266 · Report as offensive
Profile Wiggo "Democratic Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 16316
Credit: 210,265,453
RAC: 144,762
Australia
Message 1984267 - Posted: 9 Mar 2019, 11:10:13 UTC - in response to Message 1984266.  

I use ethyl alcohol myself.

Cheers.
Ahhh! But do you use it IN yourself?
Yes I do. :-D

Cheers.
ID: 1984267 · Report as offensive
Kevin Olley

Send message
Joined: 3 Aug 99
Posts: 786
Credit: 183,729,363
RAC: 407,743
United Kingdom
Message 1984286 - Posted: 9 Mar 2019, 15:45:01 UTC - in response to Message 1984255.  

Anyone ever see the MB cleaning video by der8aur where he uses the dishwasher??


No, but I have used warm water with washing up liquid and a toothbrush on a number of occasions with no problems, the main thing is to make sure it is rinsed off with plenty of clean water afterwards and is dried properly after before it is powered up.

Another liquid that is useful for cleaning is lighter type petrol, useful for some grease residues and the type of glues used on sticky labels etc.
Kevin

ID: 1984286 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 8302
Credit: 718,782,831
RAC: 1,742,847
United States
Message 1984288 - Posted: 9 Mar 2019, 16:05:03 UTC - in response to Message 1984255.  

Silicone oil is not conductive. It is insulating. If on contacts, will not allow proper mating of the contact surfaces unless there is enough mechanical pressure to break through the film. It also has a tendency to spread.

It is common in electronics manufacturing and repair to put circuit boards through a low pressure wash or just handwash with the equivalent of Simple Green cleaner and degreaser. Then you drip dry the PCB for ten minutes and then put it into an industrial oven set at 145°F. and bake the PCB for 6 hours. I used to do that all the time when I was in the repair shop.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1984288 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7493
Credit: 2,439,035
RAC: 325
Italy
Message 1984293 - Posted: 9 Mar 2019, 16:28:23 UTC

When I worked in SGS (now STMicroelectronics) in Agrate I saw technicians testing PCBs using a hair dryer (called phon in Italy referring to the Favonius hot wind, or Phoen). They heated every component until the system crashed and then they substituted it.
Tullio
ID: 1984293 · Report as offensive
Profile RueiKe Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 460
Credit: 275,937,024
RAC: 339,812
Taiwan
Message 1984539 - Posted: 11 Mar 2019, 11:41:47 UTC

Appreciate Wiggo's ethyl alcohol suggestion. It made it easier to stick the card into the dishwasher! I washed with plain water at 50C for about 70min, then baked per Keith's suggestion. I could not find FujiPoly locally, so I had to order from Amazon US. It will be a week before I can test it out. Not sure if all of this will revive the card, since it would crash the system in minutes when installed, but it is an interesting experiment!
ID: 1984539 · Report as offensive
account
Volunteer tester

Send message
Joined: 31 Dec 99
Posts: 1773
Credit: 6,295,856
RAC: 2,976
Saint Pierre and Miquelon
Message 1984623 - Posted: 11 Mar 2019, 23:13:50 UTC

The new graphite based thermal pads are interesting.
Linus Tech Tips did a video on one such product...
The LAST Thermal Paste you'll ever need??.
Larger sheets of a similar product, made by Panasonic, are also available...
5.5"x12" sheet
They do look to be finicky to apply, but are reusable.
ID: 1984623 · Report as offensive
Profile RueiKe Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 460
Credit: 275,937,024
RAC: 339,812
Taiwan
Message 1985469 - Posted: 16 Mar 2019, 12:28:24 UTC - in response to Message 1984623.  

The new graphite based thermal pads are interesting.
Linus Tech Tips did a video on one such product...
The LAST Thermal Paste you'll ever need??.
Larger sheets of a similar product, made by Panasonic, are also available...
5.5"x12" sheet
They do look to be finicky to apply, but are reusable.

Just watched the video. Looks good, but it is electrically conductive, so it may not be appropriate for VRMs. Also, not sure if there are any metal components in it. I looked at other versions on Amazon and they looked metalic. I would be concerned with metal in the contact with back side of the die if I wasn't sure that it did not contain copper or gold due to their high diffusivity in Si.
ID: 1985469 · Report as offensive
Profile RueiKe Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 460
Credit: 275,937,024
RAC: 339,812
Taiwan
Message 1985470 - Posted: 16 Mar 2019, 12:32:25 UTC

Update: After letting it sit for a week, the contaminant continued to seep out of the card, so I ran it through the dishwasher again and another bake. I reassembled with FujiPoly and put it back into the system. No problems booting and now running SETI for the last few hours with no issues. One caution is that some of the inductors started to show a little rust. I suggest coating them with a small amount of oil after the clean to prevent this.
ID: 1985470 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2751
Credit: 470,361,807
RAC: 847,332
Canada
Message 1985473 - Posted: 16 Mar 2019, 13:07:08 UTC - in response to Message 1985469.  

Also those sheets are 0.2mm compared to 1.0mm typically used for GPUs/VRMs.
With no compression fit you would run into air gaps.
ID: 1985473 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 8302
Credit: 718,782,831
RAC: 1,742,847
United States
Message 1985524 - Posted: 16 Mar 2019, 19:30:05 UTC - in response to Message 1985469.  

The new graphite based thermal pads are interesting.
Linus Tech Tips did a video on one such product...
The LAST Thermal Paste you'll ever need??.
Larger sheets of a similar product, made by Panasonic, are also available...
5.5"x12" sheet
They do look to be finicky to apply, but are reusable.

Just watched the video. Looks good, but it is electrically conductive, so it may not be appropriate for VRMs. Also, not sure if there are any metal components in it. I looked at other versions on Amazon and they looked metalic. I would be concerned with metal in the contact with back side of the die if I wasn't sure that it did not contain copper or gold due to their high diffusivity in Si.

I got the impression the material was graphene based. So yes electrically conductive.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1985524 · Report as offensive

Message boards : Number crunching : Thermal Pad Failure


 
©2019 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.