Gigabyte Aorus 1080ti - card not initialising

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

VDD (positive) to VSS (GND), most of the GDDR5X chip pins are for power, so 1A to 1B perhaps (see image)

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

I am now systematically removing the ram chips one by one to:

1) Find which are good still (if any!) to re-use
2) See if the low ohms on the FBVDDQ rail is gone once all removed

Removed 3 so far and all low ohms on the chips, 1 is 14ohm which is still too low (should be 30-60ohms)

So my assumption appears valid - with all ram chips removed the FBVDDQ resistance is now almost exactly 60ohms, so no low ohms partial short within the core itself. Of the 11 chips removed, 6 have a reasonable resistance (kiloohm range, nominal value seems to be around 14k), 1 is around 800ohms, 4 are in the 2ohm range so deemed dead.

The last 7 removed were all better resistance, so I believe there was a short under on of them (solder ball perhaps) as the FBVDDQ rail didn't return to 60hm till after the final 7 were removed

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

A card with no ram at all installed is able to run MODS/MATS blind but will obviously fail on all memory tests (purpose of that being to see if the core still functions), the decision now is whether to continue and fix the low ohm 5v rail (by removing further DRMOS chips) or assess the new faulty card that should be with me tomorrow and see the extent of the damage on that first.

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

So my assumption appears valid - with all ram chips removed the FBVDDQ resistance is now almost exactly 60ohms, so no low ohms partial short within the core itself. Of the 11 chips removed, 6 have a reasonable resistance (kiloohm range, nominal value seems to be around 14k), 1 is around 800ohms, 4 are in the 2ohm range so deemed dead.

The last 7 removed were all better resistance, so I believe there was a short under on of them (solder ball perhaps) as the FBVDDQ rail didn't return to 60hm till after the final 7 were removed

I have the same card with the same low resistance on vmem. I have not yet removed the memory but your test sure makes it look like it is worth to try! Thanks !!

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

It would be interesting to find out if there is a specific bad date code on the memory chips that went bad or a specific location they were on the pcb.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

It was the upper/middle banks by the looks of it.

I have received the latest card, will open a new thread for that one.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

Thread for card #2 here https://www.badcaps.net/forum/showthread.php?t=105416

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

As a data point I have two supposedly new GDDR5X Chips here Micron code D9VRL

Their resistance of 1A to 1B measures:

20.6 K ohms date code 8MA77 (KJF4)
19.5 K ohms date code 8MA77 (K7C9)

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

According to "CSN-11: Product Marks/Product and Packaging Labels Introduction" from micron on page 6


Date codes are alphanumeric characters that indicate the year and workweek the parts
were marked, in even-numbered workweeks. The first characteris the last numberin the
year, and the second (alpha) character is the workweek.

A = 2 E = 10 I = 18 M = 26 Q = 34 U = 42 Y = 50
B = 4 F = 12 J = 20 N = 28 R = 36 V = 44 Z = 52
C = 6 G = 14 K = 22 0 = 30 S = 38 W = 46
D = 8 H = 16 L = 24 P = 32 T = 40 X = 48

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

This means my "new chips" were made in 2018 in the 26 week of the year.

[zefrench]

Re: Gigabyte Aorus 1080ti - card not initialising

Of note, it is mentioned that in 2018 Micron had many bad batch of GDDR6 memory (different then GDDR5x used in the 1080 ti). Supposedly anything older than date code 8Y is bad. I have no other information at this point.

Of course this is the different memory type GDDR6 that is probably unrelated. D9WCW BGA code.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

Date code on this card's ram is 8JA77

[steviemac_uk]

Re: Gigabyte Aorus 1080ti - card not initialising

definitely should consider this..

[techy1122]

Re: Gigabyte Aorus 1080ti - card not initialising

Did you get anywhere with this at the end?

I have a very similar fault on Inno3d 1060 3GB version, all memories come up with errors. Replaced all memory modules with known good ones, but still he same the memory voltages stays on 1.35V all the time /not going to 1.45v on start-up like any other card would/ and screen turns on after 40 seconds like when you have one bank faulty.

All other voltages are normal and present. /Could not try to flash the bios as I could not find it anywhere so far/.

The other interesting thing the GPU core voltages goes to 0.9v and drops to 0.67v on the GPU side.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

This card was relegated to a parts card due to all the ram chips being shorted (not worthwhile to buy 11 new chips)

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

Did you get anywhere with this at the end?

I have a very similar fault on Inno3d 1060 3GB version, all memories come up with errors. Replaced all memory modules with known good ones, but still he same the memory voltages stays on 1.35V all the time /not going to 1.45v on start-up like any other card would/ and screen turns on after 40 seconds like when you have one bank faulty.

All other voltages are normal and present. /Could not try to flash the bios as I could not find it anywhere so far/.

The other interesting thing the GPU core voltages goes to 0.9v and drops to 0.67v on the GPU side.

What I will add is that I have seen multiple times now a MATS result erroring on all banks even though its only 1 to 3 chips that are erroring - are your errors read or write errors? Do you have the MATS log?

Have you tried other BIOS files from here (after backing up BIOS on card first)? https://www.techpowerup.com/vgabios/...emSize=&since=

Also, what resistance are you reading on VMEM (FBVDDQ)?

[techy1122]

Re: Gigabyte Aorus 1080ti - card not initialising

Hello, got it sorted st the end. Turned out, after all that work it was a dry joint underneath the gpu. Reflowed and got the board working again. The vram voltage has gone up to 1.37v. Passed memory test and furmark stress test too. Just ordered a new stencil to reball it later.

I had exactly the same log as yours but I can not find it.

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

Ok, that's great.

I have picked up 1080ti board with missing GPU core for cheap, so will attempt to reball this core and install in that.

No doubt it will have issues, but worth a try for £45 for the board (ram chips that are on it worth more than that).

Will need to move a chip from B0 to D0 to match the chip config.

Will document my progress...

[benwaterson]

Re: Gigabyte Aorus 1080ti - card not initialising

So my assumption appears valid - with all ram chips removed the FBVDDQ resistance is now almost exactly 60ohms, so no low ohms partial short within the core itself. Of the 11 chips removed, 6 have a reasonable resistance (kiloohm range, nominal value seems to be around 14k), 1 is around 800ohms, 4 are in the 2ohm range so deemed dead.

The last 7 removed were all better resistance, so I believe there was a short under on of them (solder ball perhaps) as the FBVDDQ rail didn't return to 60hm till after the final 7 were removed

More up to date info regarding testing of these https://youtu.be/IBW5alE6K0w?t=345

[computeka]

Re: Gigabyte Aorus 1080ti - card not initialising

Ok, that's great.

I have picked up 1080ti board with missing GPU core for cheap, so will attempt to reball this core and install in that.

No doubt it will have issues, but worth a try for £45 for the board (ram chips that are on it worth more than that).

Will need to move a chip from B0 to D0 to match the chip config.

Will document my progress...

How did you end up with your problem?