Thursday, February 15th 2024
22 GB Modded GeForce RTX 2080 Ti Cards Listed on Ebay - $499 per unit
An Ebay Store—customgpu_official—is selling memory modified GeForce RTX 2080 Ti graphics cards. The outfit (located in Palo Alto, California) has a large inventory of MSI GeForce RTX 2080 Ti AERO cards—judging from their listing's photo gallery. Workers in China are reportedly upgrading these (possibly refurbished) units with extra lashings of GDDR6 VRAM—going from the original 11 GB specification up to 22 GB. We have observed smaller scale GeForce RTX 2080 Ti modification projects and a very ambitious user-modified example in the past, but customgpu's latest endeavor targets a growth industry—the item description states: "Why do you need a 22 GB 2080 Ti? Large VRAM is essential to cool AIGC apps such as stable diffusion fine tuning, LLAMA, LLM." At the time of writing three cards are available to purchase, and interested customers have already acquired four memory modded units.
They advertise their upgraded "Turbo Edition" card as a great "budget alternative" to more modern GeForce RTX 3090 and 4090 models—"more information and videos" can be accessed via 2080ti22g.com. The MSI GeForce RTX 2080 Ti AERO 11 GB model is not documented within TPU's GPU database, but its dual-slot custom cooling solution is also sported by the MSI RTX 2080 SUPER AERO 8 GB graphics card. The AERO's blower fan system creates a "mini-wind tunnel, pulling fresh air from inside the case and blowing it out the IO panel, and out of the system." The seller's asking price is $499 per unit—perhaps a little bit steep for used cards (potentially involved in mining activities), but customgpu_official seems to be well versed in repairs. Other Ebay listings show non-upgraded MSI GeForce RTX 2080 Ti AERO cards selling in the region of $300 to $400. Custom GPU Upgrade and Repair's hype video proposes that their modified card offers great value, given that it sells for a third of the cost of a GeForce RTX 3090—their Ebay item description contradicts this claim: "only half price compared with GeForce RTX 3090 with almost the same GPU memory."
Custom GPU Upgrade and Repair: "In today's review, we're diving into the highly acclaimed 2080 Ti 22 GB Turbo Edition graphics card—a powerhouse that's been making waves in the tech community. Let's explore the enhanced capabilities of this card."
The video description continued with: "Upgraded from the standard 11 GB of VRAM, the card now boasts a hefty 22 GB, catering to even the most memory-intensive tasks. It benefits from high-quality Samsung memory modules, and the 300 A core is ripe for overclocking enthusiasts."
Sources:
Ebay Listing, Wccftech, VideoCardz, Tom's Hardware
They advertise their upgraded "Turbo Edition" card as a great "budget alternative" to more modern GeForce RTX 3090 and 4090 models—"more information and videos" can be accessed via 2080ti22g.com. The MSI GeForce RTX 2080 Ti AERO 11 GB model is not documented within TPU's GPU database, but its dual-slot custom cooling solution is also sported by the MSI RTX 2080 SUPER AERO 8 GB graphics card. The AERO's blower fan system creates a "mini-wind tunnel, pulling fresh air from inside the case and blowing it out the IO panel, and out of the system." The seller's asking price is $499 per unit—perhaps a little bit steep for used cards (potentially involved in mining activities), but customgpu_official seems to be well versed in repairs. Other Ebay listings show non-upgraded MSI GeForce RTX 2080 Ti AERO cards selling in the region of $300 to $400. Custom GPU Upgrade and Repair's hype video proposes that their modified card offers great value, given that it sells for a third of the cost of a GeForce RTX 3090—their Ebay item description contradicts this claim: "only half price compared with GeForce RTX 3090 with almost the same GPU memory."
The video description continued with: "Upgraded from the standard 11 GB of VRAM, the card now boasts a hefty 22 GB, catering to even the most memory-intensive tasks. It benefits from high-quality Samsung memory modules, and the 300 A core is ripe for overclocking enthusiasts."
26 Comments on 22 GB Modded GeForce RTX 2080 Ti Cards Listed on Ebay - $499 per unit
nvidia/comments/12xeyvhwww.tomshardware.com/news/3070-16gb-mod
But that begs the question: is this for real, or is it a scum scam ?
Seeins where these card mods are being performed, I also wonder just exactly how many root kits, key loggers and other associated spyware/backdoors they are installing along with those supposed "high-quality Samsung" memory :)
Am I the only one who misses the days of AIB's making their own higher memory edition cards. Off the top of my head I remember those GTX 580 3gb cards back in the day (For those who don't remember, the normal was 1.5gb)!
2080 Ti was indeed a good value chip for a long time though, it makes sense that (a few) people *would* do this, or go for something like this. It's unfortunate the uncertainty that comes with it due to driver signing and all that (others probably know more about it than I do), especially now when you can just point someone to something like a 7800xt (or 3080 if you prefer).
I stand/stood by that card being a good deal for a VERY long time though. They became pretty cheap (not counting when effected by the crypto boom; also not everyone knew/kept track of that if you were looking for a single card) pretty quickly after the 3000 series launched (~$400), and to this day I'll point to the idea you could overclock one (~20+% from FE) to around the level of a 4070, which is the best-kept secret of an nVIDIA card in years, especially since it supports DLSS. Many people probably didn't consider them because of launch price/tier, etc, but the value was there on the used market. It was also a funny card given the 3000 series didn't OC well (bc Samsung), neither did the 6800 (because of a host of reasons), and if you even were to consider a 7700xt you would probably buy a 7800xt...So it's quite a gem if you're 'that guy' and/or bought it at the right time, because the playable settings/DLSS made it a pretty nice upper, now lower mid-range card esp wrt at least 'playable' settings. It is/was hanging on pretty well, those things considered.
IMO, and obviously just that, it meshed it perf pretty well with 11GB. Are we at a point where you should be looking at (at least) 12GB and probably 16GB? Sure, but performance on those cards is also higher, especially in absolute terms.
The thing about a 2080 Ti is that the things that really stress ram it doesn't really have the hardware capability to use anyway (obviously there are instances of 4k/1440p DLSS though), so in that respect I think it's fine where it is. If the same-ish performing card were created today, but had, for instance, better RT capability, I would hope it would have greater than 11GB, or even 12GB.
Oh, that's right, they did. The 4070 series, and I did complain about it because it does have technology to better use that higher ram and keep playable settings.
Sometimes I feel like articles exist to see if I'm consistent: for me to say this doesn't need more RAM. It's not as simple as that, as it's a lot easier to implement a 128-bit clam shell or 256-bit design now with the performance capability of what's possible on 4/5nm than to demand a fair price for 22/24GB back when this came out on '12' (actually 16nm), and/or even the full 384-bit chip/12GB (as it was pretty massive; 754mm2) back then.
It's so important to remember that a 1GB chip (now, or rather when I recently checked) costs something like $3 on the open market. It's not expensive, and you shouldn't have to step up to a card that costs twice the price, $600 more (wrt 4070 vs 4080 at launch) for $12 of ram. I'm sure 2GB (16Gb) chips are slightly more expensive, but you get my point. Not trying to fight, simply state it's not a major ask from AMD/nVIDIA to incorporate a decent amount (that won't limit the actual GPU's potential) on a card, especially not now, especially for the asking prices (and nvidia's margins) of some cards. Things should have proportional perf/ram of where the baseline scales at the high-end, right now (imo) that is that of a stock 4080. Even that card could use a little more ram (I think the absolute [OC] perf of 7900xt nailed it pretty well with 20GB; very similar to relative absolute perf of 2080Ti with 11GB), but at that point we're dicing hairs that aren't nearly as important in the near-term.
I'd much prefer 3070 Ti with 16 GB, this modification makes way more sense (also significantly faster in games compared to 2080 Ti if we're talking 16 GB 3070 Ti).
It truly did rob most of the fun from this hobby, which I trace back to twisting pots to adjust PSU rails and off-setting 5v to overclock BH5 3.3v ram without higher available settings in BIOS (the DFI 'toasters'); EASY pot-based GPU volt-mods that never broke any card I used (driver fallback and OCP/OVP were always a thing). I have a VapoChill in my closet.
While AMD did eventually follow (to some extent), it was not very quickly or to as egregious extent IMHO. That said, voltage/PL's do exist, CPUs do have boost caps, etc etc etc etc.
Most of the fun is lost these days. Some due to everything being binned so concisely and hardware locks to offset what anybody can now do with a simple software setting through programmable VRM chips and what-not, but some due to things like this...Which nVIDIA most-certainly started and has made a lot of money because of doing.
It's important to remember ATi was (and AMD largely is) a engineering company first. nVIDIA is a MARKETING company first. When you realize this, why things are the way they are becomes a lot more clear.
Huang wanted to be (like) Steve Jobs. He wanted nVIDIA to be (like) Apple. He largely succeeded. I largely don't support either company because better HARDWARE solutions exist, or exist earlier, often for less money and more personal freedom (I don't remember an iAnything ever having a SD card slot or replacable battery), without contrived plans to fleece customers or portend software is a way to sell unrequired new hardware (which dates at least back to buying PhysX and running it on CUDA, if not earlier).
Lots of people simply don't remember/care how it used to be, and that's what makes me the most sad, because it used to be SUCH a fun hobby, especially for aspiring engineers or hobbyist tinkerers trying to save a bit of dosh. There was a lot of knowledge shared that made the scene better. nVIDIA won the battle of attrition versus AIBs (and end-users), and it pisses me off so few people remember and/or care.
So if I ever appear bitter, that's why! It's nothing against anyone personally, and I appreciate everyones' opinions (even when we don't agree, ntm I'll freely admit my speculation is often anywhere from slightly off to totally wrong). It doesn't mean that Apple/nVIDIA can't or don't make good products nor they don't have talented employees (they most certainly do), or can't POPULARIZE/invent a new useful technology (they most-certainly have). It's the how and why...and it all comes back to taking as much money from you and putting it into their pocket as often they can, and doing this by controlling their ecosystems very tightly.
I come from an era where we fought against that, and ATi (and old AMD) used to represent the possibilities of hardware, and what AIBs/you could do with it being YOUR right.
That's largely gone now, and it makes me sad because there are less people truly informed wrt how things work. This is because of less access to engineers by the press, replaced by marketing (ntm nVIDIA largely lies and obfuscates where they can [get away with it]) and everything nVIDIA makes is a 'black box' that is 'hardware dependant' (also largely a lie); less dudes out there practicing with a soldiering iron or messing around with binary because of that, which in-turn makes the next-generation of enthusiasts even MORE uneducated and willing to accept their bs.
This is why you see me praise those ripping apart DLSS/FSR to add it to games; as well as other 'injection'-type programs. That's all we really have left from that golden age.
That, and, well, a bit of stuff like this (it's just too bad many times it is done in bad-faith through bios/driver bs).
Used 3090s are cheaper as well. So if you're into something that's "there's no such thing as enough VRAM" (= not video games) you'd be more pleased with a 3090. Especially thanks to its higher VRAM bandwidth.
That type of thinking is why my 2080 Ti (and those used for other builds) was/were cheap. That's cool and everything (and I get you're talking usefulness of 8->16gb versus 11->22GB or 16 vs 11GB).
I respect your conceivable use for it and/or the size/power consumption differences.
This...this is what I mean, though. This is not the hardware community that I remember. Not a slight against you (or your preference, which I respect)...just times changing. Different eras.
The 16GB 4070 Ti Super sucks because they gutted the core too much. 4070 Ti 12GB could use 16GB, just as that card could/does. As much? No, but both more than 12GB. But as many can't grasp, this is on purpose to keep the margins on 4080 (that can do 'all the things' at an acceptable perf level [including RT *right now*) and to make many future products with similar raster performance and more ram look better (like a [5060?] 16GB 6144sp card with higher clocks or a conceivable 18GB 5070 with 4080 raster). 4070 Ti 12GB with 24GB *would* be an interesting product, imho. Not just now, but in a year, two, three from now...AFTER those next cards release and their limitations more apparent. The thing is, nVIDIA does not want that comparison so that you would notice the inching forward of raster versus their cost-savings (smaller process) and margains.
Like I said, it's cool to like what you like. Just understand nVIDIA does things like this very purposely in exactly the *type* of ways I'm talking about. An AD104 user (of any type) is more likely to upgrade to a 5070 than if nVIDIA had given them and/or designed them with more RAM (16/24GB). Likewise, a 4070TiS user may upgrade to a 5070 because it's raster perf will be good-enough, unlike 4070 Ti S, for which I have said several times 'adequete for the high-end' is around that of a stock 4080 or overclocked 7900xt (which I recommend not just for the extra RAM that pairs well with it's absolute performance, but the cheapest with good-enough raster longevity). RT is a moving target, in-which future nVIDIA generations will relegate 4080 (and it's adequate RT) to being unable to play at resolutions it is capable of raster. In-turn, those users will upgrade even if they don't need more raster or RAM.
As I've also said, sometimes these things don't show up on the charts because sometimes game are picked and chosen for certain reasons (which I don't always agree with) and/or cards become 'too old' to keep testing. But are they really too old if they perform the same as one or two or three tiers down of products nVIDIA is still selling, with now simply more ram (and features limited to new generation cards in order to market them and earn higher margin)?
I'm not here to change your mind of what performs better and/or is a better option, just here to help you understand the only way these things change is if users are more vocal about it. Otherwise, your cards age worse (in and more and more ways nVIDIA can drum up, like RT) when they would otherwise age just fine (ex: if RT was on a fixed ratio across generations). These are cheap, and fixable issues. nVIDIA doesn't change them because doing things this way allows them to make a literal killing on high-end margains, trick people with cards that are mis-matched but cheap for them to make, and also then sell them very cheap (to make) cards later with adequate performance (but perhaps missing enough RAM for their lifetime) more OFTEN to consumers. Margin? Volume? They said "why not both?" by doing what they do.
This means additional 11 GB will only be useful in non-gaming scenarios where, believe me or not, almost similarly priced used 3090 destroys this 2080 Ti either way.
Yet additional 8 GB will improve the gaming performance of 3070 Ti which is faster than 2080 Ti by itself. NB: 2080 Ti runs out of speed before running of VRAM in gaming if we don't count extreme cases such as VRAM hogging games or 8K textures mods.
So yeah, modded 3070 Ti is both more useful for gaming and much more useful for professional usage than the OG 3070 Ti, whereas the modded 2080 Ti is only better in professional usage, gaming experience stays vastly untouched. This is why I stated doubling the VRAM is more welcome in 3070 Ti rather than in 2080 Ti which has a different set of bottlenecks.
(They have a history of going to great lengths to prevent their own previous generation products from competing with their primary market.)
Also, SLI is completely useless for gaming now. No new game takes advantage of it to the point that two 2080 Tis is better than one.