[HARDWARE MOD] 980M to Desktop 980 core upgrade

[Khenglish 1154]

Due to a stupid accident by me, I acquired a 980m with a chunk knocked out of the core. Not wanted to scrap a perfectly good top end PCB for parts, I wanted to replace the core. You can see the gouge in the core to the left of the TFC918.01W writing near the left edge of the die.

 

First I had to get the dead core off:

 

 

With no sellers on ebay selling GM204 cores, my only option was to buy a full card off ebay. With no mobile cards under $500,I had to get a desktop card. And with this much effort involved to do the repair, of course I got a 980 instead of a 970.

 

Below is the dead 980 I got off ebay:

 

 

You can see for some reason someone removed a bunch of components between the core and PCI-E slot. I have no idea why anyone would do this. I tried the card and it was error 43. PCB bend seemed to be too little to kill the card, so those missing components had to be it.

 

GPUs can be dead because someone removed or installed a heatsink wrong and broke a corner of the core off, so buying cards for cores on ebay is a gamble. This core is not even scratched:

 

 

Preheating the card prior to high heat to pull the core:

 

 

And core pulled. It survived the pull:

 

 

Next is the 980 core on the left cleaned of solder. On the right is the original 980m core:

 

 

Next I need to reball the 980 core, and lastly put it on the card. I am waiting for the BGA stencil to arrive from China. It still has not cleared US customs:

 

https://tools.usps.com/go/TrackConfirmAction?tLabels=LS022957368CN

 

When that shows up expect the core to be on the card in 1-2 days.

 

So some potential issues with this mod besides me physically messing up:

 

I believe that starting with Maxwell Nvidia started flashing core configuration onto the cores, like intel does with CPUID. I believe this because I found laser cuts on a GK104 for a 680m, but could not find any on two GM204 cores. In addition, Clyde figured out device IDs on the 680m and K5000m. They are set by resistor values on the PCB. The 980m has the same resistor configuration as the 680m for the lowest nibble of the Device ID (0x13D7), but all of the resistors are absent. Filling in these resistors does nothing. Resistors do exist for the 3 and D in the device ID. Flashing a 970m vBIOS on my 980m did not change the device ID or core configuration. If this data is not stored on the PCB through straps or the vBIOS, then it must be stored on the GPU core.

 

So I expect the card with the 980 core to report its device ID as 0x13D0. The first 12 bits pulled from the PCB, and last 4 from the core. 0x13D0 does not exist. I may possibly be able to add it to the .inf, or I may have to change the ID on the board. With the ID's 0 hardset by the core, I can only change the device ID to 0x13C0, matching that of a desktop 980.

 

An additional issue may be that the core may not fully enable. Clyde put a 680 core on a K5000m and never got it to unlock to 1536 CUDA cores. We never figured out why.

 

Lastly, there was very tough glue holding the 980m core on. When removing this glue I scraped some of the memory PCB traces. I checked with a multimeter and these traces are still intact, but if they are significantly damaged this can be problematic for memory stability. I think they are OK though, just exposed.

 

Due to Clyde's lack of success in getting his 680 core to fully unlock I am concerned I might not get 2048. If I don't at least I should still have a very good chip. Desktop chips are better binned than mobile chips (most 980s are over 80% ASIC quality, while most 980ms are below 70%). In addition this 980 is a Galax 980 Hall of Fame, which are supposedly binned out of the 980 chips. Having a 90%+ ASIC would be great to have. The mid 60s chips we get in the 980m suck tons of power.

 

I want to give a special thanks to Mr. Fox. This card was originally his. He sent me one card to mod and one to repair. I repaired the broken one and broke the working one. The broken one is the one I've been modding.

 

Edited July 11, 2016 by Khenglish

[Guest]

Awesome work, Brother @Khenglish. Looking forward to seeing this fly in your P150EM. I'm tagging @johnksss and @Prema because I know both of them are interested in seeing how this shakes out, too.

[Brian 1977]

Damn @Khenglish that's really impressive, good job! Going to promote this to a front page post. I'll update the front page post with your results once you have them but looking good so far. 

[bloodhawk 127]

This is amazing. I might try to do a 980 core swap on my notebook 980 (200W) GPU, if this works. Always happy to swap to a better core. Looking forward to your progress.

Also how did you find the stencil ? Would love to have a link if possible.

[Khenglish 1154]

Ebay for the stencil. GM204 is the same as GK104.

 

The stencil is arriving tomorrow. I don't think I will have time to put the core on until Sunday though.

Edited July 16, 2016 by Khenglish

[TBoneSan 112]

@Khenglish

 

Amazing stuff dude.  This will be absolutely awesome if it works out.. Fingers crossed.

You just redefined 'hardcore' man..

 

 

[Guest]

2 hours ago, TBoneSan said:

@Khenglish

 

Amazing stuff dude.  This will be absolutely awesome if it works out.. Fingers crossed.

You just redefined 'hardcore' man..

 

 

@Khenglish is definitely a rock star with this kind of thing. Very talented... here was the last example he showed us of his amazing talent. 

 

 

[drunkenninja 27]

Wow this is pretty impressive, congratulations, would never expect to find something like this here, but Brian pointed me in this direction after my introduction.

 

I have done a lot of this kind of work in the past, reballed what feels like thousands of chips but is probably more like hundreds, I'm rubbish with v-bios software and matching the components but I am very good at reballing and reseating chips. Anyone into this stuff might know me from over at BGA Mods, I used to mod there for a long time.

 

That chipped die reminds me of the old PS3 GPU's and how careful you had to be removing the IHS cause those machines got so hot 90% of them needed reballing, that heat caked the paste onto the die and 4 RAM chips like superglue.

 

If you guys have any specific questions feel free to post them and PM me, I may not have all the answers but it would be a pleasure to help out here. I have done most things. Actually exchanging different version GPU's and PROCESSORS was of interest for me a while back but I never had the software knowledge you guys have so is very interesting. I'm pretty old skool so don't use an IR unit but instead a self modded T-8280 griddle and a Aoyue 968 Wand.

 

One thing I will say is try to avoid ovens and such, this is really delicate work and at relow temps I always nudge the chip with a chopstick just to confirm reflow state, with an oven you can't do that, even though tou could in theory confirm the reflow temps with a couple, TC  monitors (always use one either side of the chip). 

Edited July 17, 2016 by drunkenninaj
Added Equipment

[Khenglish 1154]

Yeah oven use is definitely not ideal, but I do this stuff as a hobby, not a business, so I can't justify the cost of a decent bga station on a grad student stipend.

 

Usually I don't have a hard time with figuring out if I melted all the solder or not. When adding a core to a pcb I can look from the side and see the core drop slightly when all solder balls become molten. Removing a core is much more tedious, but is still not as bad as reballing.

 

What I usually spend the most time on is reballing. On the first try on a core I find usually around a dozen solder balls randomly around the core did not wet to the solder pads. I then clean all flux off, remount the stencil and try again. On this 2nd try I usually finally have all balls on, but not all are centered properly on their solder pads. I then remove the stencil, clean all flux off, try again with fresh flux, and even then I still usually have a few balls that are not centered. If I give them a little nudge when molten they will finally center on the pad. This whole process takes hours, and I would really like to shorten it. I find my heat gun will often heat stencils too unevenly making the stencil warp so much that solder balls can slide under the stencil, so this means long heating cycles in the toaster. Plus the uneven heating by a heat gun really only makes 63/37 use viable, as unleaded solder's melting point is so close to temps that kill the core package.

Edited July 18, 2016 by Khenglish

[Khenglish 1154]

Core finally reballed. If the mount is poor I will be very very angry...

[Brian 1977]

Fingers crossed...

[Guest]

14 minutes ago, Brian said:

Fingers crossed...

Mine, too.

[Khenglish 1154]

Card cooling. New brain installed.

[Brian 1977]

[Khenglish 1154]

IT WORKS!

 

2048 SHADERS!

 

980 DESKTOP DEVICE ID!

 

MODDING VBIOS NOW TO MAKE IT RENDER

Edited July 18, 2016 by Khenglish

[Guest]

13 minutes ago, Khenglish said:

IT WORKS!

 

2048 SHADERS!

 

980 DESKTOP DEVICE ID!

 

MODDING VBIOS NOW TO MAKE IT RENDER

WOOT!!!! Congrats, Brother @Khenglish

[Khenglish 1154]

OK so modding the device ID to 0x13C0 in vBIOS did not help. The card won't even report clocks now.

 

Waiting for Prema.

[Khenglish 1154]

So it actually works with the 980m vBIOS. I tried modding too soon. I just needed to reinstall the driver.

 

I only ran a very lightweight render test because right now the card is only running on 2 phases. I'm pulling the phase driver from my 980m now to get the 3rd phase back up.

[Brian 1977]

37 minutes ago, Khenglish said:

IT WORKS!

 

2048 SHADERS!

 

980 DESKTOP DEVICE ID!

 

MODDING VBIOS NOW TO MAKE IT RENDER

 

12 minutes ago, Khenglish said:

So it actually works with the 980m vBIOS. I tried modding too soon. I just needed to reinstall the driver.

 

I only ran a very lightweight render test because right now the card is only running on 2 phases. I'm pulling the phase driver from my 980m now to get the 3rd phase back up.

 

 

Congrats man, I knew you could pull it off. Super curious to see if it OC's better than the 980m after you get those power phases added in. 

[Dr. AMK 82]

Congratulations for your unusual achievement, something we do not see a lot, proving the possibility of the GPU core processor transfer successfully between the desktop and mobile versions, this achievement may  motivate others to invent new and useful things as you did.

Waiting for the OC, testing and benchmark results after @Prema Bios. Good Luck

 

[Khenglish 1154]

3rd phase is up!

 

Here's 1300/6800:

 

http://www.3dmark.com/3dm11/11422402

 

I'm going to push much harder, just starting conservative for now in case I blow a fuse or something. The card now draws 20% more power, which isn't bad for a 33% bigger core.

[drunkenninja 27]

15 minutes ago, Khenglish said:

3rd phase is up!

 

Here's 1300/6800:

 

http://www.3dmark.com/3dm11/11422402

 

I'm going to push much harder, just starting conservative for now in case I blow a fuse or something. The card now draws 20% more power, which isn't bad for a 33% bigger core.

 

Great Job, this wasn't an easy task, you can feel proud...

 

You will likely get bitten by the bug now and start hunting for chips to swap 

[Khenglish 1154]

Here's 1360/6850 firestrike:

 

http://www.3dmark.com/3dm/13379728

[Khenglish 1154]

OK pushing voltage now. Going over 1.1625V shuts down my 330W PSU. I need to change out the connector for a better one and it should be able to do 1.2V

 

I broke 20k gpu score 3dm11 with 1416/6950:

 

http://www.3dmark.com/3dm11/11422494

 

1414/6955 firestrike 15165 GPU score:

 

http://www.3dmark.com/3dm/13380739

Edited July 18, 2016 by Khenglish

[Guest]

AWESOME!

 

 

Already looking great with 980M vBIOS...will have to cook up a real Franken-Mod to match the Monster you created there Doc...