Sunday, November 1st 2026

Modded RTX 2080 Ti 22 GB Shows What Extra VRAM Can Still Do for Turing

PC Games Hardware has put a modded GeForce RTX 2080 Ti 22 GB through their test gauntlet, and the extra memory seemingly pays off once games run out of headroom. At 4K with DLSS Performance and FSR 3.1 Frame Generation, the card averaged 52.5 FPS in Flight Simulator 2024 against 30.2 FPS on a regular 11 GB model, roughly 74% faster. Spider-Man 2 jumped from 39.4 to 64.0 FPS, and Indiana Jones and the Great Circle crashed to the desktop on the 11 GB card. At 1440p with DLSS Quality, the gap was smaller, though those first two games still ran about 20% slower on 11 GB. Of course, PCGH chose nine memory-hungry games on purpose, and most titles run fine on 11 GB. Keep in mind that these are just one outlet's numbers, and other results may vary.

These cards have shown up everywhere from Alibaba to eBay listings at $499, and this one came from a reader. All eleven memory chips are swapped for 2 GB Samsung parts. No firmware changes are needed, since relocating a single strap resistor tells the GPU about the extra capacity, and NVIDIA's regular drivers pick it up. Its weak blower-style cooler and 250 W limit held it back, so PCGH tuned both cards to identical 1.9 GHz, 15 GT/s, and 280 W settings. At around €500, the 22 GB card costs quite a bit less than a 16 GB RTX 5060 Ti, which goes for €770 in Germany. PCGH still calls it a gamble, citing eight-year-old hardware and soldering work that can fail weeks later, and advises checking the seller's reputation and whether any warranty goes beyond a 30-day return window.
Sources: PCGH, via Videocardz
Add your own comment

25 Comments on Modded RTX 2080 Ti 22 GB Shows What Extra VRAM Can Still Do for Turing

#1
shuffy_losys
There's a few takeaways I can make of this, but none of them are really that useful to the present day, even ignoring the RAM situation.
Posted on Reply
#2
Dr. Dro
And despite it all, this modified GPU should still slot in below the Titan RTX (which has a complete TU102, 384-bit bus with 24 GB installed) by a sizable margin. Indeed, I haven't seen the lucky folks who own one complain to this day.

Unfortunately they are still reliably selling for $900-950 on eBay as of now, with no signs of decline because of Nvidia's continued driver support for Turing and the fact that these are still fairly powerful, 24 GB GPUs.
Posted on Reply
#3
lexluthermiester
shuffy_losysThere's a few takeaways I can make of this, but none of them are really that useful to the present day, even ignoring the RAM situation.
The 2080ti is still a solid card today.
Posted on Reply
#4
N/A
Well, that’s not quite right. 22 GB is 50% more. So, for 750, it might not be worth it; personally, I’d still go for a 5070 with 12 GB or a 3080 with 20 GB for 900, but only if my 2080 Ti is no longer up to the task.
Posted on Reply
#5
Nostras
Honestly these numbers are a good example of a card just not having enough horsepower to properly utilize the extra VRAM and why absolute numbers matter next to relative ones. Might be interesting for local AI models, but for practicality I'd argue the extra VRAM is not worth it. If you're really that paranoid about VRAM just get a 6800(XT) or 6900XT second hand. These will be either significantly cheaper or similarly priced. Or anything 6700XT-6900XT - 7700XT-7900 GRE.
Posted on Reply
#6
_roman_
Germany has a lot of technicians. When you want such a card ask for a VRAM swap yourself with your graphic card.

So often people claim on the nvidia side, this is not my opinion, VRAM does not matter, the card will struggle regardless of the VRAM. I remember this a lot with teh nvidia 3070 and similar cards.

I checked the german source quickly. Low end card for WHQD. The site menitons 500€ street prioce which seems for myself a bad bargain, especially because i bought a radeon 9070 OC card recently for 560€, a few years ago a radeon 7800xt for below 600€ what i remember. I am certain that nvidia 2xxx something card is far weaker as a radeon 9070 OC. I have 2 years warranty which started around 05 / 2026.

IT is a technical interesting approach, but money wise a bad option. PCGH, Igorslab and some others motivated me to buy around 06 / 2026 a radeon 9070. They stated in 11/2025 till 07/2026 that radeon graphic card prices will go up.
Note. PCGH was referenced here as the source for this test. A low quality german website which has offensive moderators. This time also used thew wrong units. I expect more than morals from a tech website. On the other hand they still sell low quality paper hardware and gaming magazines, even today.

-- I think it its a 22GiB GPU, not a 24GiB gpu, but whatever
Posted on Reply
#7
lexluthermiester
N/A22 GB is 50% more.
No, 22GB is 100% more. The 2080ti comes with 11GB.
Posted on Reply
#8
N/A
At around €500, the 22 GB card costs quite a bit less than a 16 GB RTX 5060 Ti
This doens't hold true. I mean, 22 GB costs 770 euros. And that changes everything, especially after the recent unlock for double the computing power in FP16.
Posted on Reply
#9
RootinTootinPootin
That's why I kept my Titan RTX, its still a monster in gaming in 1440p, can still tank higher settings and that 24gb vram is an abundance..
Posted on Reply
#10
Dr. Dro
NostrasHonestly these numbers are a good example of a card just not having enough horsepower to properly utilize the extra VRAM and why absolute numbers matter next to relative ones. Might be interesting for local AI models, but for practicality I'd argue the extra VRAM is not worth it. If you're really that paranoid about VRAM just get a 6800(XT) or 6900XT second hand. These will be either significantly cheaper or similarly priced. Or anything 6700XT-6900XT - 7700XT-7900 GRE.
Strongly recommend against buying anything RDNA 2 for use on Windows, especially with compute in mind. The hardware isn't up to the task, as they fully lack hardware matrix multiplication and native support for most numeric data formats used for deep learning workloads. On Linux they can be somewhat workable as long as you don't know or care about what "efficiency" or "performance" mean, but on Windows the situation is dire. Not even to go into the whole driver support thing. There is a very specific reason why they are building 22 GB 2080 Ti's instead of 32 GB 6900 XT's.
Posted on Reply
#11
Lew Zealand
Same as having a 22GB 5060, 4060 Ti or 3070. In other words, overkill. 11GB already complements a 2080 Ti nicely, save that VRAM for a card that can use it better: 3080/Ti. Or a 4070/Ti/Super but maybe those use higher density chips?
Posted on Reply
#12
robergems
I love my 2080ti. but dumb design. mine had only the blower fan and no vent on the end. so I made one. cut out the rectangular slot. then took metal mesh and with a flat head soldering iron to plastic weld melted a metal mesh rectangle from the inside. otoh I dont play demanding games anyway.

got mine 2 years $220 from a video editor. in this shitty market with their greed, err high inflation, im keeping all my pcs and not upgrading anytime soon. not cause I cant, because I dont want to support the greed.

hilarious and pathetic the "sale"s you see from tech sites. "save $25 dollars on this bundle deal only $4000 for a mb/ram/cpu." gfy
Posted on Reply
#13
lexluthermiester
Lew ZealandSame as having a 22GB 5060, 4060 Ti or 3070. In other words, overkill. 11GB already complements a 2080 Ti nicely, save that VRAM for a card that can use it better: 3080/Ti. Or a 4070/Ti/Super but maybe those use higher density chips?
While you might have a point, there are technical reasons why a 2080ti was selected. People can do whatever they want with their own property.
Posted on Reply
#14
Legacy-ZA
It's always better to have more VRAM than less; I care little for the excuses people find in trying to "justify" the lack thereof. I already find that in some scenarios, my 16GB isn't enough. I despise that green goblin that deprives us of a good amount of VRAM on our cards.

Back when vendors were allowed to add more to their models, I always opted in to buy the larger-capacity VRAM card, and that decision never ever disappointed.
Posted on Reply
#16
Nathaaab
NostrasIf you're really that paranoid about VRAM just get a 6800(XT) or 6900XT second hand. These will be either significantly cheaper or similarly priced. Or anything 6700XT-6900XT - 7700XT-7900 GRE.
Meh. Not really. The issue is AMD's VRAM compression is significantly worse. 16GB on a RX6800 was always full for me in certain VR scenarios. On my current RTX 4070 TiS, the 16GBs are never full in the same or worse scenarios (since I upped internal VR res even more).

It's not a 1 to 1 comparison but a significant difference, the only explanation I have is that VRAM on AMD is less efficiently used and uses worse compression.
Posted on Reply
#17
ZoneDymo
Dr. DroAnd despite it all, this modified GPU should still slot in below the Titan RTX (which has a complete TU102, 384-bit bus with 24 GB installed) by a sizable margin. Indeed, I haven't seen the lucky folks who own one complain to this day.

Unfortunately they are still reliably selling for $900-950 on eBay as of now, with no signs of decline because of Nvidia's continued driver support for Turing and the fact that these are still fairly powerful, 24 GB GPUs.
I find this comment baffling.

Despite it all? despite what? despite them managing to force more ram on a gpu than Nvidia did and it showing improvements in tasks....it would still slot below a stronger card that came out of the factory with even more ram and costing way more money.....ok ermm shocking news?

"I havent seen the lucky folks who own one (what does luck have to do with it? was it a lottery to get one? could you not just buy it?) complain to this day"
What would they complain about? that their old card is not doing super well anymore? atleast not as good as other newer cards? I would think they would just buy a new card....

"Unfortunately they are still reliably selling for $900-950 on eBay as of now, with no signs of decline because of Nvidia's continued driver support for Turing and the fact that these are still fairly powerful, 24 GB GPUs."

K...relevance? nobody was talking about that card in this article...its about how a product was handicapped from the factory.
Posted on Reply
#18
JustBenching
ZoneDymoI find this comment baffling.

Despite it all? despite what? despite them managing to force more ram on a gpu than Nvidia did and it showing improvements in tasks....it would still slot below a stronger card that came out of the factory with even more ram and costing way more money.....ok ermm shocking news?

"I havent seen the lucky folks who own one (what does luck have to do with it? was it a lottery to get one? could you not just buy it?) complain to this day"
What would they complain about? that their old card is not doing super well anymore? atleast not as good as other newer cards? I would think they would just buy a new card....

"Unfortunately they are still reliably selling for $900-950 on eBay as of now, with no signs of decline because of Nvidia's continued driver support for Turing and the fact that these are still fairly powerful, 24 GB GPUs."

K...relevance? nobody was talking about that card in this article...its about how a product was handicapped from the factory.
The 2080ti was handicapped from the factory? Wow...okay.
Posted on Reply
#19
FierceRed
All this just makes me lament what they did to my boy - the RTX 3080 10GB - all the more. The most legendary piece of silicon since the 1080 Ti. It was clearly cut down from what it should have been and that's a position confirmed by facts years later.

Release the 3080 Schafer Cut!
Gemini said:

Was the 3080/Ti VRAM amount "nerfed" on purpose?
Yes, it was a deliberate, profit-driven segmentation strategy rather than a technical limitation.
When the Ampere architecture (RTX 30-series) launched, Nvidia intentionally kept VRAM allocations low on its mid-to-high-end tier cards to protect the profit margins of their ultra-premium tier. [1]
  • Protecting the RTX 3090: The flagship RTX 3090 launched with a massive 24GB of VRAM and a steep $1,499 price tag. If Nvidia had given the $699 RTX 3080 a comfortable 16GB or 20GB of VRAM right out of the gate, many professional creators, machine-learning researchers, and enthusiasts would have bought the cheaper card instead. Keeping the 3080 at 10GB forced those who required high memory capacities into buying the much more expensive 3090. [1, 2, 3, 4]
  • The Evidence of Pivots: Nvidia originally planned to give these cards more VRAM, but adjusted engineering targets based on supply limits and market conditions. Leaked and unreleased physical engineering samples—such as an RTX 3080 Ti 20GB variant—have frequently surfaced in the hands of collectors, proving that higher-capacity boards were designed and fabricated before Nvidia decided to slash the retail specs down to 12GB. [1, 2, 3]
  • The Micron G6X Shortage: Another major justification was the scarcity of the ultra-fast GDDR6X memory modules co-developed with Micron. At the time, 2GB high-density GDDR6X chips were not yet in mass production. To achieve 20GB or 24GB, Nvidia had to put 24 individual 1GB chips on both sides of the circuit board (which they did for the 3090). Doing this on a lower-priced $699 card would have been too complex and expensive. [1]
Nvidia structured the 3080 and 3080 Ti memory capacities precisely to squeeze maximum yield out of Micron's early batch of GDDR6X chips while firmly trapping high-capacity professional buyers behind the $1,500 barrier of the 3090. [1, 2, 3]

Source
Posted on Reply
#20
Dr. Dro
ZoneDymoI find this comment baffling.

Despite it all? despite what? despite them managing to force more ram on a gpu than Nvidia did and it showing improvements in tasks....it would still slot below a stronger card that came out of the factory with even more ram and costing way more money.....ok ermm shocking news?

"I havent seen the lucky folks who own one (what does luck have to do with it? was it a lottery to get one? could you not just buy it?) complain to this day"
What would they complain about? that their old card is not doing super well anymore? atleast not as good as other newer cards? I would think they would just buy a new card....

"Unfortunately they are still reliably selling for $900-950 on eBay as of now, with no signs of decline because of Nvidia's continued driver support for Turing and the fact that these are still fairly powerful, 24 GB GPUs."

K...relevance? nobody was talking about that card in this article...its about how a product was handicapped from the factory.
Baffling why? It's a perfect showcase as to why memory has always been at a premium. You are aware that the 2080 Ti has the same TU102 core, right? The key differences being 4352 vs 4608 cores, and 11 (352-bit, single-sided) vs 24 (full 384-bit, clamshell). 22 GB 2080 Ti is simply doing the clamshell thing and retaining the specifications - that if these modified cards aren't going a step further and using double density GDDR6 ICs that were not available 8 years ago. If Nvidia had released it with 22 GB back then, it would have been considered unnecessary and contributive of its high price.

It`s not even the first time Nvidia does that, 1080 Ti with 11 GB, Titan X/Xp with only one side, reserving clamshell for the P5000/6000 models
Posted on Reply
#21
MikOnTopGG
22GB seems like an interesting amount for a mod. How many VRAM slots are there on this one?
Posted on Reply
#22
VoidRunner96
melkor_unlimited8GB ought to be enough for everybody.
THIS>..... its what valve is stressing also. Nvidia and dev got some gamers thinking weird.

Nvidia & AMD came with all this fancy fidelity stuff to aid in make things shiny and detailed and the solution for high res graphics etc is throw more vram at it.. It almost feel like when intel was on the Ghz rampage with hyper-treading etc... almost every new cpu release bump the speed by alot, then they claim we can't keep increasing as is..

Is the solution really more vram, of course not, because we see even with lots of vram, and poor optimize game still is crap.

The industry need a settle point (maybe 8 or 12) for a few years (maybe 5yrs or 10), then you bump it up... after-all, high quality games are taking like 3+ years now to make.

Anything above 12GB is offered to business
Posted on Reply
#23
Dr. Dro
MikOnTopGG22GB seems like an interesting amount for a mod. How many VRAM slots are there on this one?
11. 352-bit.
Posted on Reply
#24
lexluthermiester
MikOnTopGG22GB seems like an interesting amount for a mod. How many VRAM slots are there on this one?
It has to do with the number of RAM chips on the card. 11 chips at 1GB per chip. What the modders did in this case is mount 2GB chips that have the same operating specs as the originals. This effectively doubles the VRAM from 11GB to 22GB. It's a very cool mod!
Posted on Reply
#25
MikOnTopGG
lexluthermiesterIt has to do with the number of RAM chips on the card. 11 chips at 1GB per chip. What the modders did in this case is mount 2GB chips that have the same operating specs as the originals. This effectively doubles the VRAM from 11GB to 22GB. It's a very cool mod!
Yeah, okay I see.
Posted on Reply
Add your own comment
Nov 21st, 2026 22:04 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts