GTX 1080 Ti vs RTX 20 Series: Is 11GB of VRAM Still Worth It?

The GTX 1080 Ti refuses to die gracefully. Years after its prime, people still look at its 11GB of VRAM and wonder whether it's secretly the smarter buy than a newer RTX 2060 or 2070. That was what this debate came down to: should you skip the newer architecture, DLSS, ray tracing, and longer driver runway, and grab the old Pascal monster because modern apps and games keep getting hungrier? In a Proxmox-flavored setup the question gets even messier, because the original poster wasn't only thinking about games. They wanted to run multiple VMs simultaneously, which makes VRAM feel like the one spec that actually matters.

The 1080 Ti argument is still emotionally strong

There's a reason the 1080 Ti still gets respect. It was king for a long time, and that reputation didn't come from nostalgia alone. Eleven gigabytes of VRAM still looks good next to a 6GB RTX 2060, especially if you're thinking about multiple guests, heavier workloads, or avoiding the feeling that your GPU ran out of room before it ran out of muscle. One commenter gave the classic answer: if you truly need the extra VRAM, the 1080 Ti can still be fine. The word doing the work there is "need," which rules out wanting it or feeling safer with it.

For non-gaming VMs, the extra horsepower of a 20-series card may not be the whole story. One user said the extra oomph of newer cards probably wasn't worth much for various non-gaming VMs, and that AMD might be a great VRAM option if buying. That lands because virtualization changes the buying math. A gaming chart won't tell you much if your real goal is carving resources across workloads, and raw frames per second stop being the center of the universe. Capacity, driver behavior, passthrough weirdness, and how the card behaves under Linux suddenly matter more than benchmark bragging rights.

RTX has the features that actually aged better

The opposing case was blunt: modern features may matter more than the 1080 Ti's extra VRAM. DLSS is the big one. One commenter argued that DLSS helps produce good image quality while lowering render load, and that RTX 20-series cards also bring newer shader support. The 1080 Ti can lean on FSR or Intel-style upscaling, but it doesn't get proper DLSS. Ray tracing on early RTX cards never became the killer feature people hoped for, especially on lower-end models, but DLSS absolutely changed the value story for Turing cards.

That matters if the card will do any gaming or modern graphical work. A 1080 Ti with 11GB can still choke if the game expects architectural features it doesn't have, or if the settings that fill that VRAM also demand more modern rendering tricks. One commenter called that a dependency circle: you buy the older card for VRAM, then lower graphics settings until the workload no longer needs all that VRAM anyway. Think of VRAM as desk space: great to have, and useless if the tools on the desk are too old for the job.

Driver support is the dealbreaker people overlook

The least emotional and most damning argument was driver support. Several people pointed out that the GTX 10-series is now on the wrong side of NVIDIA's long-term support line, while Turing cards still appear in current releases. One commenter said you'd have to curate driver versions because NVIDIA no longer supports the 1080 Ti in the same way, and installing the latest drivers may not work. People underestimate that kind of problem until it ruins a weekend.

In a Proxmox environment, driver friction is a big deal. Nobody wants a GPU that only works if you freeze the right driver branch, dodge a kernel update, and pray that passthrough doesn't get spicy. Older NVIDIA cards can be fun, but "fun" in infrastructure often means "you now own a tiny compatibility museum." If you're using the GPU casually, maybe that's acceptable. If the machine is supposed to host dependable workloads, driver aging may matter more than those extra gigabytes ever will.

vGPU dreams make everything weird

The thread also drifted into sharing one GPU across multiple VMs. Someone asked how that works, because they thought simultaneous GPU use across VMs wasn't supported. The answer that came back was "vGPU unlock," one of those phrases that sounds like a cheat code and behaves like a warning label. Others quickly added nuance: on older cards like the 1080 Ti it may work through a workaround, but it can be hit-or-miss. One commenter warned it may require a custom GPU BIOS and could even risk bricking the card.

At that point the 1080 Ti starts to look less like a bargain and more like a project. If the goal is multiple VMs sharing GPU resources, a consumer Pascal card with hacks may not be the clean path, and someone suggested a Pascal Quadro instead for that kind of use. Another noted that multiple LXCs can share the same GPU without unlocking, but the original scenario was about VMs and not just containers. LXCs and VMs don't play by the same GPU-sharing rules, and mixing them up is how people buy the wrong card for the right idea.

The AMD crowd had a point

A surprisingly strong third lane opened up: skip both the 1080 Ti and the low-end RTX 20-series and look at Radeon. One commenter said that if the goal is more VRAM for a better price, just buy a Radeon card. Another suggested AMD cards with 16GB, and someone else brought up a 7900 GRE as a sweet spot with 20GB of VRAM, strong performance, and reasonable power draw. The original poster even agreed that AMD drivers on Linux feel refreshingly painless because they work out of the box.

AMD still isn't automatically the best answer for every VM or passthrough use case. NVIDIA keeps advantages in CUDA-heavy workflows, some media stacks, and certain software ecosystems. But if the core pain is VRAM per dollar and Linux driver sanity, AMD deserves a serious look. Reading between the lines, the 1080 Ti versus RTX 20-series framing may be too narrow, and sometimes the right reply to "old NVIDIA or newer NVIDIA?" is "why are we only shopping NVIDIA?"

The right answer depends on the job

The most honest comment in the whole discussion was also the shortest: it depends on what you're using it for. For gaming, a decent RTX 20-series card, especially a 2070 Super or 2080-class card at the right price, is likely the smarter long-term pick because DLSS and driver support matter. Avoid the original 6GB RTX 2060 if VRAM is the worry. For non-gaming VM workloads where raw VRAM matters more than features, the 1080 Ti can still make sense, as long as you're comfortable with older driver branches and potential weirdness.

For multiple VMs sharing one GPU, be extra careful. A consumer 1080 Ti plus vGPU unlock may sound like homelab wizardry, but it's a long way from buying something designed for that role. For Linux-first, VRAM-heavy setups, AMD may be the saner path. The 1080 Ti is still a legend, and legends don't get driver updates forever. It can still be a good buy at the right price, for the right workload, with eyes open, as long as you don't mistake 11GB of VRAM for a time machine.