NVIDIA GeForce RTX 2080 Ti Modded to Support 22 GB of GDDR6 Memory

[AleksandarK]

Have you ever wondered if your graphics card could pack just a little bit more VRAM than it is advertised to come with? Well, if you have some spare time and some awesome soldering skills, you could find out yourself by placing higher-capacity VRAM chips in place of the standard memory. That is exactly what VIK-on, a hardware modder from Russia, has done with his graphics cards. Before, VIK-on modded the NVIDIA GeForce RTX 2070 to support 16 GB of GDDR6 memory and modded NVIDIA's GeForce RTX 3070 to also support 16 GB of GDDR6 memory. Today, VIK-on has done it again and the modder has tested his skills by proving that it is possible to install 22 GB of GDDR6 memory on the GeForce RTX 2080 Ti graphics card.

The modder has taken a broken NVIDIA GeForce RTX 2080 Ti GPU that was in very bad condition. He had to re-solder the GPU and fix some broken PCB traces. Finally, after that, he tried to install more VRAM than the card came with in the first place. The TU102 SKU is capable of handling up to 48 GB of VRAM, as seen with Quadro RTX 8000 GPU. However, the problem would be firmware support. VIK-on used a strap mod, leading the GPU BIOS to believe that there is much more memory present compared to the stock version, and the card managed to boot. However, some screen flickering was present and it had stability issues, so the mod isn't very successful.



Additionally, one of the biggest problems was that VIK-on couldn't source eleven 16 Gb GDDR6 chips to make the 22 GB GDDR6 VRAM configuration, however, it was proven that with appropriate soldering skills and a bit of luck sourcing components, you could mod your GeForce RTX 2080 Ti graphics card to support 22 GB of VRAM.

View at TechPowerUp Main Site

 

[Fouquin]

You would be surprised at how many GPUs can be "modded" for more VRAM. It's almost always the case that the IMCs can support exceptionally dense RAM packages. Some IMCs even support asymmetrical arrays (I.E. 256-bit with 1.25GB). With some older cards BIOS limits aren't even a thing; like TeraScale cards you can literally just double the VRAM with no other modding required. Nobody does it because it's a waste of time.

 

[QUANTUMPHYSICS]

No thanks. Better to just get the 3090RTX.

Not only is it properly built but it's newer, more efficient and just BETTER.

 

[Space Lynx]

but why

 

[Hemmingstamp]

but why

Fun with burn fingertips is all I can imagine. Joking aside it's good to tinker IMO, something many of the younger generation are missing out on

 

[N3utro]

but why

For science of course

 

[Makaveli]

but why

lol what's what I said.

All of this and its not stable and flickers whats the point.

 

[r9]

epenis enlargement surgery.

 

[zlobby]

Noobs. I always download more RAM when needed.

 

[lexluthermiester]

but why

Why not?

 

[DeathtoGnomes]

waiting for the story about the assembly line to do the modding.

 

[Deleted member 185088]

I wonder why MSI, Asus... etc don't increase memory of their graphics cards, say a 3070 with 16gb, it shouldn't be difficult for them as they have access to the firmware.

 

[Mussels]

I love these dumb idea hacks so much

 

[Panther_Seraphin]

I wonder why MSI, Asus... etc don't increase memory of their graphics cards, say a 3070 with 16gb, it shouldn't be difficult for them as they have access to the firmware.

Money and access to GPUs

A 16Gb 3080 would eat into the sales of the much more profitable 3090. So if someone DID decide to make one you can guarantee Nvidia would suddenly not be able to supply said maker with the GPU cores anymore

 

[Mussels]

I wonder why MSI, Asus... etc don't increase memory of their graphics cards, say a 3070 with 16gb, it shouldn't be difficult for them as they have access to the firmware.

This used to happen all the time, and then 1GB GDDR5 cards became 2GB GDDR3 with shit performance, and people got mad that performance was all over the place between the 'same' model cards
I guess Nv and AMD put a stop to it, with all the shit variants hurting the products image

 

[evernessince]

This used to happen all the time, and then 1GB GDDR5 cards became 2GB GDDR3 with shit performance, and people got mad that performance was all over the place between the 'same' model cards
I guess Nv and AMD put a stop to it, with all the shit variants hurting the products image

1030 DDR3 edition

"excuse me sir but I exist"

 

[Icon Charlie]

When I was more involved in the Pacific Rim than I am now, I knew that Nvidia was castrating their own cards for sectional profit. It is far worse not than back then but it still indeed happen.

IMHO A very good GPU's that was created is the 1050ti. There is a great deal they could have done with this bugger as that little GPU had so much potential. So to my Chagrin in hearing that Nvidia ramping up the 1050ti once again kind of made me smile. I would have love to see innovation on this GPU as well as others however...

Greed kills innovation. Greed kills just about everything in the end.

 

[mechtech]

Mad soldering skills right there

 

[Blueberries]

I'm sure the 1% of owners that would benefit from this would love to void their warranty.

 

[watzupken]

I am not sure what is the point of doing this. Sure you have the potential of having more VRAM, but at the expense of killing it. Its a darn waste when people are struggling to get a good graphic card nowadays, and people just "throwing away" old GPUs just to prove that it can accomodate more VRAM. I know its mentioned that is broken in the first place, but the modder seems to be able to fix it to the point it runs. So its still a waste.

 

[Valantar]

I wonder why MSI, Asus... etc don't increase memory of their graphics cards, say a 3070 with 16gb, it shouldn't be difficult for them as they have access to the firmware.

Nvidia wouldn't allow them to. Nvidia determines the vast majority of specs for their various GPU lineups, including the amount of VRAM.

I am not sure what is the point of doing this. Sure you have the potential of having more VRAM, but at the expense of killing it. Its a darn waste when people are struggling to get a good graphic card nowadays, and people just "throwing away" old GPUs just to prove that it can accomodate more VRAM. I know its mentioned that is broken in the first place, but the modder seems to be able to fix it to the point it runs. So its still a waste.

Did you miss the part where the GPU was physically damaged and not working before this mod? So not only did they mod it, they repaired a broken GPU.

As for "the point" - the point is to see if it can be done. Nothing more, nothing less. And that's fine - it's pretty interesting to see, after all!

 

[Ruru]

I've always been wondering that could it be possible to put that missing memory chip and hack the bios for 1080 Ti/2080 Ti to be a 12GB 384-bit card

but why

You may as well ask why people do extreme overclocking. Everything computer-related doesn't need to be practical. It's a hobby just like any other hobby.

Money and access to GPUs

A 16Gb 3080 would eat into the sales of the much more profitable 3090. So if someone DID decide to make one you can guarantee Nvidia would suddenly not be able to supply said maker with the GPU cores anymore

16GB 3080 wouldn't make sense as 16GB would result just for a 256-bit memory bus unless a mixed density solution is supported.

 

[Mussels]

You wouldnt be able to up the bit depth of the memory controller or enable missing GPU Cores, all you can really do is double the density of the chips since they're pin compatible
It's like swapping two 8GB sticks for two 16GB sticks in a desktop vs trying to add a fifth slot to a board

It's not practical, but it IS amazing to see the potential for higher VRAM variants of cards if manufacturers were allowed to by Nvidia/AMD, and saw a need.

3080Ti being an 12GB 3090 is a perfect example of this going the other direction (and what NV could have launched from day 1)

 

[Ruru]

You wouldnt be able to up the bit depth of the memory controller or enable missing GPU Cores, all you can really do is double the density of the chips since they're pin compatible
It's like swapping two 8GB sticks for two 16GB sticks in a desktop vs trying to add a fifth slot to a board

It's not practical, but it IS amazing to see the potential for higher VRAM variants of cards if manufacturers were allowed to by Nvidia/AMD, and saw a need.

3080Ti being an 12GB 3090 is a perfect example of this going the other direction (and what NV could have launched from day 1)

I'm just wondering that is the missing 32-bit physically cut down from the GPU die, or is it just not active because of the missing memory chip.

I mean if anyone hasn't tried doing it, how do we know if the controller is fully enabled, but since there is only 11 chips, the bus is 352-bit because of that.