Author Topic: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?  (Read 586 times)

0 Members and 1 Guest are viewing this topic.

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Hello guys!

I have a KFA2 1080 Ti EX OC Galaxy (yea I have never heard about that brand before either) which I bought with destroyed caps and got it cheap and I replaced them all and it worked fine for a very few month.

Then it started to lock up in games after an hour or even longer but could go back to desktop and restart the game but lock quicker until cooled down.

Time went on and it got worse until it within a month got so severe that it totally locked up the computer so only a reboot would work. But if it locked up that hard it would not boot again until it cooled down and it was not sure after that either.

After some time it could and booted but if even if I started a youtube video it hanged hard after some seconds, a game it hanged pretty much instantly.

When it hanged is got the "space invader" artifacts on the screen. In 2D it work ok-ish but sometime it get thin short lines here and there and when I push the card together a little at the gpu/memories I could once see some very few pixels disappear and be normal so I 100% expect a memory fault.


Started thus Nvidia MATS and different memory tests but all tests pass!! Although it say it can't test RM upper memory bank or something by what ever the reason which I really want to but don't know how.

I have seen many pictures on these space invader artifacts but not really got the cause of it I think mor them memory problems? And MATS say all is ok so what the heck is going on here? I really, really hope its not the GPU? But everything work pretty much fine otherwise, so its strange.

Any clues?
« Last Edit: September 25, 2021, 05:27:46 pm by William_K »
 

Online wraper

  • Supporter
  • ****
  • Posts: 13798
  • Country: lv
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #1 on: September 25, 2021, 05:31:50 pm »
Did you run MATS with proper parameters so it tests 100% of the memory? What I have downloaded from Russian forums usually had a bat file with parameters to test only a small fraction of the total memory.
 

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #2 on: September 26, 2021, 09:21:20 am »
Yes I ran the specific MATS and followed this from the start for the initial install things:

I could choose to test 100% or the full amount in MB of the 6Gb memory but that would have take hours based on how slow even 100 Meg was, which was really surprising based on how fast this memory really is, so I don't know if that is normal.

It also mentioned in the video that even 5MB should be enough for testing, but I don't know and it don't seem so for me.
« Last Edit: September 26, 2021, 09:26:10 am by William_K »
 

Online wraper

  • Supporter
  • ****
  • Posts: 13798
  • Country: lv
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #3 on: September 26, 2021, 10:27:25 am »
It also mentioned in the video that even 5MB should be enough for testing, but I don't know and it don't seem so for me.
It's enough only if you have a completely dead chip, otherwise it's not.
 

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #4 on: September 26, 2021, 01:01:32 pm »
Don't get how you ment there? If the gpu is dead (which I doubt in this case) then you can't test any memory on the board?

It say:

"Warning: RM unavailable - won't test GP10x upper memory".

Read that had to do with if you tested a secondary card if fex you had grapics in the cpu, which I don't so I see its strange.
« Last Edit: September 26, 2021, 02:03:13 pm by William_K »
 

Online wraper

  • Supporter
  • ****
  • Posts: 13798
  • Country: lv
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #5 on: September 26, 2021, 01:32:28 pm »
Don't get how you ment there? If the gpu is dead (which I doubt in this case) then you can't test any memory on the board.
Why do you assume faulty gpu when we talk about testing RAM? RAM is not a chip or what? Detecting completely dead RAM chip or if one of the data lines to RAM failed. Depending on which memory bank failed card often still displays image. It will show image even with missing RAM chips. Not to say you can run MATS on a card which does not show image at all, and then simply check report.txt file for results. Fast partial test is unlikely to detect RAM which is just unstable or has only a few bits failed.
« Last Edit: September 26, 2021, 01:36:56 pm by wraper »
 

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #6 on: September 26, 2021, 02:06:26 pm »
Thought if fex any memory controller was faulty on the gpu, and well I assumed the memory was involved if there is graphical artifact on the screen?
 

Online Labrat101

  • Supporter
  • ****
  • Posts: 565
  • Country: 00
  • Renovating Old Test Equipment & Calibration .
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #7 on: September 26, 2021, 03:12:56 pm »
 If this card was used for mining. And was pulled because it failed.  You are wasting your time . As the mining process can run the cards at max loads and put abnormal stress on the GPU and the memory..
"   All Started With A BIG Bang!! .  .   & Magic Smoke  "
 

Online wraper

  • Supporter
  • ****
  • Posts: 13798
  • Country: lv
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #8 on: September 26, 2021, 03:33:45 pm »
Thought if fex any memory controller was faulty on the gpu, and well I assumed the memory was involved if there is graphical artifact on the screen?
Not sure if you properly understood what I wrote.
Is it a RAM failure? Likely but not necessarily. As I already wrote, you cannot test RAM instability with just a short partial test. Short test is only suitable for finding the most obvious faults, like a whole data line not working or a complete RAM IC failure. Not suitable for cases when card actually runs but is not stable.
 

Offline testpoint1

  • Frequent Contributor
  • **
  • Posts: 280
  • Country: us
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #9 on: September 28, 2021, 02:13:28 am »
pls at least test "test 94", following command:
.//mods gputest.js -oqa -old_gold -test 94 -fan_speed 50 -dramclk_percent 100 -ignore_fatal_errors -run_on_error -matsinfo
 

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #10 on: September 29, 2021, 09:02:52 am »
Ok I will run that

Thought if fex any memory controller was faulty on the gpu, and well I assumed the memory was involved if there is graphical artifact on the screen?
Not sure if you properly understood what I wrote.
Is it a RAM failure? Likely but not necessarily. As I already wrote, you cannot test RAM instability with just a short partial test. Short test is only suitable for finding the most obvious faults, like a whole data line not working or a complete RAM IC failure. Not suitable for cases when card actually runs but is not stable.

Ok I get it. Short tests only for major faults, and I should perhaps then just be pretty much lucky to find it. It pretty much always show artifacts, once in a Linux Mint dist it didn't but at a reboot it did, and so on, But not 3D. Will test the entire memory during work and see if its even done when I get home 8 h later just to see sort of speak. And the card doesn't seem like it has been very hot during a long period or so.

If this card was used for mining. And was pulled because it failed.  You are wasting your time . As the mining process can run the cards at max loads and put abnormal stress on the GPU and the memory..

True, but the strange part then is that it ran just perfectly fine during 1.5 - 2 month or what it was and then slowly degraded during a month or so, and I doubt (?) a mining card that has been decommissioned because it broke due to long time overload would still work that long afterwards? But sure I can be wrong for all what I know but I hope I'm not.

The only thing that was not ok on it when I got it was that the electrolytics had been torn off "by kids" as the seller said Idk about that, and they had been tried to be glued in place again with hot glue. :palm: They were easy to replace with new ones though and the card worked just fine afterwards and I was very happy.


« Last Edit: September 29, 2021, 09:07:12 am by William_K »
 

Online Labrat101

  • Supporter
  • ****
  • Posts: 565
  • Country: 00
  • Renovating Old Test Equipment & Calibration .
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #11 on: September 29, 2021, 10:01:26 am »
The caps were ripped off by kids and hot glue back.  :palm:
These boards are multi layer there is a chance that even though you think you replaced them there is a good chance that the iner traces were damaged.  And I guess you assumed they even stuck the right values back in the right places.   :palm: .
Maybe why it worked for a short time.  Cap voltages may be wrong.  .
. If kids have messed with this card it has been for sure some what miss handled . Plus static charge .  Bent twisted,  dropped,  etc.
 :popcorn:
« Last Edit: September 29, 2021, 12:40:27 pm by Labrat101 »
"   All Started With A BIG Bang!! .  .   & Magic Smoke  "
 

Offline William_K

  • Contributor
  • Posts: 13
  • Country: se
Re: 1080 Ti KFA2, kinda "common" artifacts but don't match up to what I see?
« Reply #12 on: September 29, 2021, 10:24:09 am »
You guess a lot, of course I didn't guess, gees. I know what value, voltage, ESR, height and so on they should be and ordered the right ones before I got the card and those on there was the original and also had the right value and so on, so I'm 100% confident I put the right ones on there.

All the Vcore layers are connected at this point and I of course checked the vias for any damage but those 4 that was ripped off (not all of them) so had the legs not been ripped out from the board but stayed on 2 of them. The other two the legs had been ripped out from the board but the holes looked fine since the vias them self had not been ripped out so there was still solder in them so I'm confident they are all alright.

----

Alright, have done a full memtest now (I think, took 4 hours) and it doesn't seem good with not a single memory chip working as it seem.
« Last Edit: September 29, 2021, 09:17:10 pm by William_K »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf