Gigabyte Aorus 1080ti - card not initialising #2

[benwaterson]

As per my other thread, I have received a 2nd failed card so thought it would be best to record my progress on this one here so as not to confuse the two.

So, received the card this morning (was an ebay purchase, usual spares or repair)

Card does not look like it has been worked on, no flux or burnt components. It has been taken apart before but has original thermal pads.

Did the usual resistance tests on the voltage rails (all fine). After a quick inspection I plugged it into the test bench and powered it up.

All voltages present = good.

Connected as the primary card it tries to initialise, but just lights the backlight on the monitor after 40secs - indicates either bad ram or bios issue.

Fired up MATS and ran a 10mb test - looks like failed memory banks F1 and F0 - have attached the complete log also.

mats version 400.250. Testing GP102 with 10 MB of memory starting with 0 MB.

Read Error Count: 0
Write Error Count: 432913
Unknown Error Count: 0

=== MEMORY ERRORS BY SUBPARTITION ===
SUBPART READ ERRORS WRITE ERRORS UNKNOWN ERRS
------- ----------- ------------ ------------
FBIOA0 0 0 0
FBIOA1 0 0 0
FBIOB0 0 0 0
FBIOB1 0 0 0
FBIOC0 0 0 0
FBIOC1 0 0 0
FBIOD0 0 0 0
FBIOD1 0 0 0
FBIOE0 0 0 0
FBIOE1 0 0 0
FBIOF0 0 405662 0
FBIOF1 0 27251 0

Failing Bits:
F008 F009 F010 F011 F012 F013 F014 F015 F024 F025 F026 F027 F028 F029 F030 F031
F048 F049 F050 F051 F052 F053 F054 F055

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

I am going to reball one of the "good" ram chips from card #1 and try that on F0.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Out of interest, I have reflowed F0 first - running mats now to see if the errors change

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Interesting result.....

I reflowed F0 (the last chip to the left, near the PCIE slot).

MATS now says all errors on F0, none on F1

FBIOF0 Read errors = 0, Write Errors = 359968

Failing bits 08 to 015

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Even more interesting, ran again, exactly the same errors.

Ran a 3rd time, this time pressing down on the heatsink (i have placed the stock cooler on top of the card, its fairly heavy so is making good contact)

Errors now :

FBIOF0 Read errors = 0, Write errors = 28907

Failing bits 08 to 015.

I think a reball of the current F0 chip will fix this - will report back in due course.

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Even more interesting, ran again, exactly the same errors.

Ran a 3rd time, this time pressing down on the heatsink (i have placed the stock cooler on top of the card, its fairly heavy so is making good contact)

Errors now :

FBIOF0 Read errors = 0, Write errors = 28907

Failing bits 08 to 015.

I think a reball of the current F0 chip will fix this - will report back in due course.

Can you please measure the resistance on that chip prior to reball and mention the data code on the chip ? The more data the more we will be able to re-use parts and also not get sold used parts as new.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Date code on this card is 7VA77 - resistance is 7.18k

However there are 3 lifted pads on the chip, so can't reball this.

Reballed a chip from card #1 but that hasn't resolved anything - am waiting for chips from aliexpress that I ordered a week ago

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Date code on this card is 7VA77 - resistance is 7.18k

However there are 3 lifted pads on the chip, so can't reball this.

Reballed a chip from card #1 but that hasn't resolved anything - am waiting for chips from aliexpress that I ordered a week ago

Even one of the chips that had the higher resistance did not make a difference?

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

The chip I chose from the card #1 had a similar resistance @ 8k, however when fitted to card #2 the FBVDDQ resistance dropped to 10ohms.

So either:

1) Solder bridge under chip
2) A low ohm short on a different power lane on the chip (VDD and VDDQ are both supplied from FBVDDQ)

I will remove the chip and verify at least option 2

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

So looking at another removed chip from card #1

Pad A1 (VDD) to Pad B1 (VSS) = 17k
Pad A2 (VDDQ) to Pad B2 (VSSQ) = 20hms

Can you measure a working chip and let me know what measurements you get?

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

The chip with errors I removed from card #2 (which I am trying to replace) - 7VA77 is:

Pad A1 (VDD) to Pad B1 (VSS) = 17k
Pad A2 (VDDQ) to Pad B2 (VSSQ) = 60hms (which is within range for the resistance of the FBVDDQ rail)

This is the chip with 3 lifted pads (2 power and one data)

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

VDD = Power supply
VDDQ = I/O power supply

Attached schematic image shows they are tied to FBVDDQ

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising #2

You seem to be on to something here

Copying from the other thread and adding the new info:

Two supposedly new GDDR5X Chips Micron code D9VRL date code 8MA77

The resistance of A1 to B1 measures:
20.6 K ohms
19.5 K ohms


The resistance of A2 to B2 measures:
50 ohms
47 ohms

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Perhaps you could run MATS with the memory locations empty to confirm the error counts climbs up. I have seen reports in various discord chats of error counts staying the same even with memory chips removed thus confirming in those cases that the problems are with the GPU built-in memory controller instead of the memory chips (although in your case it sure looks like memory chips based on your findings of the measures of resistance of vMEM (FBVDDQ))

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

I did run one earlier, didn't grab the count - was around 700,000 for a 10mb test i think

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

You seem to be on to something here

Copying from the other thread and adding the new info:

Two supposedly new GDDR5X Chips Micron code D9VRL date code 8MA77

The resistance of A1 to B1 measures:
20.6 K ohms
19.5 K ohms


The resistance of A2 to B2 measures:
50 ohms
47 ohms

Sounds inline with what the resistance should be on FBVDDQ, anything lower than 30ohms on A2 to B2 indicates bad chip - that rail is meant to be 30 to 60 ohms on Pascal cards

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Ram chips will be here on tuesday by the looks of it. Only ordered 2, so hope I don't mess it up

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Ram chips will be here on tuesday by the looks of it. Only ordered 2, so hope I don't mess it up

If you don't mind measuring their resistance before installing it would be awesome

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

Sorry, already soldered in before I read this.

However, I still have 419,000 errors on F0 chip in MATS (all others are fine).

I reflowed the chip after the 1st test but not any better. Beginning to think the pcb has taken a hit and possible microfractures somewhere in the data lines or a lifted pad that's not immediately obvious.

Will have to pull the chip again and double check the data pads. Not easy to trace them though as they go direct to the core.

There is only so many times this can be unsoldered/soldered before the traces are wrecked

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising #2

I am planning my next move to see if I can try anything else before removing the chip again.

The data lines are traceable up until they go under the core (bottom left area), i have tried pressing down on both the ram chip and the core to see if this affects the card's ability to initialise with out memory errors, but no luck.

If I am going to remove the chip then I may check the traces as far as i can by scraping some solder mask and checking continuity (this should find any bad connections under the ram chip)

Open to suggestions.....