GeForce RTX 4080 Price & Used Prices: Which Local LLMs Can It Run?
Price alert
We'll email you when the eBay (US) price reaches your target.
NVIDIA's RTX 4080 (2022) is a 16GB gaming card that fits 12B-class even at 8-bit, about 40 tokens/s. Its 12VHPWR plug has melted; seat it fully.
eBay
GeForce RTX 4080 price history
eBay
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Full GeForce RTX 4080 specifications
- Memory16 GB · GDDR6X
- Bandwidth717 GB/s
- FP16 / BF16195 TFLOPS
- FP8 / INT8389.9 / 389.9 TOPS
- Bus width
- 256-bit
- Architecture
- Ada Lovelace (AD103)
- Cores
- 9,728
- Tensor cores
- 304
- Ecosystem
- CUDA
- Interface
- PCIe 4.0 x16
- NVLink
- Not supported
- Power
- 320 W
- Power connector
- 16-pin 12VHPWR
- Type
- Consumer
- Launch date
- Nov 16, 2022
- MSRP
- $1,199
Best GeForce RTX 4080 deals on eBay
Hand-picked deals from trusted eBay sellers. Shop with confidence.
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
UsedZOTAC RTX 4080 16GB AMP AIRO 16GB Graphics Card ZT-D40810B-10P (Cosmetic Damage)Top Rated$1,020Buy on eBay
UsedLenovo NVIDIA GeForce RTX 4080 GDDR6X 16GB Dedicated Graphics Card GPUTop Rated$1,050$1,575-33%Buy on eBay
UsedMSI GeForce RTX 4080 Ventus 3X OC 16GB GDDR6X Graphics Card V511-009RTop Rated$1,050Buy on eBay
UsedLenovo Nvidia RTX 4080 16GB Graphics Card GPU PC UsedTop Rated$1,090Buy on eBay
UsedPNY XLR8 Gaming Verto EPIC-X Triple Fan GeForce RTX 4080 16GB - GPU - ExcellentTop Rated$1,090Buy on eBay
UsedAsus GeForce RTX 4080 TUF Gaming 16GB GDDR6X Graphics Card 221214235102531$1,150Buy on eBay
What LLMs can a single GeForce RTX 4080 run?
One RTX 4080 runs up to 12B at 4-bit at 70 to 90-plus tokens/s. Above 27B it needs 2-bit, and Qwen3.6 35B A3B is the one to run, at about 149.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 E4B8B | 3.8 GB | Runs | 118.3 | 128K |
 Qwen3.5 9B9.7B | 4.1 GB | Runs | 109.9 | 256K |
| Gemma 4 12B12B | 4.7 GB | Runs | 96.4 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 126.3 | 256K |
| Qwen3.6 27B27.8B | 11.8 GB | Runs | 42.8 | 64K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 50.7 | 93K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 137.7 | 75K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 41.3 | 41K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 148.9 | 181K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 5 GB | Runs | 94 | 128K |
| Qwen3.5 9B9.7B | 5.7 GB | Runs | 83.2 | 256K |
| Gemma 4 12B12B | 7.1 GB | Runs | 67.8 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Needs RAM offload | 37.7 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 16.5 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Needs RAM offload | 36 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Needs RAM offload | 50.7 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 8.2 GB | Runs | 60.7 | 128K |
| Qwen3.5 9B9.7B | 9.5 GB | Runs | 52.8 | 196K |
| Gemma 4 12B12B | 12.7 GB | Runs | 40.1 | 183K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Needs RAM offload | 25.5 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 29 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Needs RAM offload | 22.5 | 198K |
| Gemma 4 31B30.7B | 32.6 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Needs RAM offload | 33.7 | 256K |
RunsFits fully in VRAM; speed is estimated from VRAM bandwidth.
Needs RAM offloadMoE only: expert weights sit in system RAM, so speed is estimated from 70 GB/s RAM bandwidth.
Too bigWon’t fit on one card at this quantization, even with RAM offload.
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- You want 12B models at 8-bit
Gemma 4 12B at 8-bit (12.7 GB) runs at about 40 tokens/s, versus about 68 at 4-bit
- Slightly faster than the 5070 Ti on long prompts
It has 10% more BF16 compute, and its bandwidth deficit only shows during generation
- You want FP8 without worrying about software versions
This generation supports FP8 starting from CUDA 11.8, unlike the 50 series, which needs CUDA 12.8 or later
Not a good fit
- You want FP4
Only the 50 series has it
- Your case is small
The Founders Edition is three slots thick and 304 mm long, one slot thicker than the 5080 Founders Edition
- Power draw matters to you
320W, 20W more than the faster 5070 Ti
Before you buy
4080 or 4080 SUPER
The 4080 SUPER, released in January 2024, has 5% more cores, the same 16GB and slightly higher bandwidth, so the two are only a few percent apart when running models. Listings often mix them up; just check the model name, since there's no difference in which models fit.
The 12VHPWR connector
Both the 4080 and 4090 use the first-generation 16-pin 12VHPWR connector. The 4090 had cases of melting connectors, which NVIDIA's investigation attributed mainly to plugs not being fully inserted. So plug the adapter into the card first, make sure it's fully seated, and only then connect the power supply cables. For how to plug it in and what to check when inspecting a card, see the RTX 4090 page.
Founders Edition size and power supply
The Founders Edition is three slots thick and 304 mm long, the same size as the 4090 Founders Edition, and NVIDIA recommends a 750W power supply for the system. If you want to add a second card later, it's hard to fit two three-slot cards side by side on a typical motherboard.
Compared with similar GPUs
GeForce RTX 4080 vs GeForce RTX 5070 Ti, GeForce RTX 3090, and GeForce RTX 4080 SUPER 32GB: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | GeForce RTX 4080 | GeForce RTX 5070 Ti | GeForce RTX 3090 | GeForce RTX 4080 SUPER 32GB |
|---|---|---|---|---|
| Memory | 16 GB | 16 GBSame | 24 GB50% more | 32 GB100% more |
| Bandwidth | 717 GB/s | 896 GB/s25% higher | 936 GB/s31% higher | 736 GB/sSame |
| BF16 / FP16 compute | 195 TFLOPS | 176 TFLOPS10% lower | 142 TFLOPS27% lower | 209 TFLOPS7% higher |
| INT8 compute | 390 TOPS | 352 TOPS10% lower | 285 TOPS27% lower | 418 TOPS7% higher |
| Qwen3.8 27B speed (2-bit) | 50.7 t/s | 62.3 t/s23% faster | 64.8 t/s28% faster | 52 t/sSame |
| Largest dense model (4-bit) | 23B | 23BSame | 37B61% more | 51B122% more |
| Power | 320 W | 300 W6% lower | 350 W9% higher | 320 WSame |
| GPU interconnect | Not supported | Not supported | NVLink 112.5 GB/s | Not supported |
| Interface | PCIe 4.0 x16 | PCIe 5.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 |
| Released | 2022-11 | 2025-02 | 2020-09 | 2024-01 |
| eBay median |
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
GeForce RTX 4080 rental price per hour
On-demand rental of the GeForce RTX 4080 starts at $0.27 per hour (RunPod), about $66/month at 8 hours a day. Buying one (eBay median $1,399) pays off in about 4.5 years, 2.3 years or 9 months at 4, 8 or 24 hours a day.
| Option | Price | 4 h/day | 8 h/day | 24 h/day | View |
|---|---|---|---|---|---|
 RunPodRentUpdated 46 min. ago | $0.27/hr | $33/mo | $66/mo | $197/mo | Rent |
 Vast.aiRentUpdated 46 min. ago | $0.43/hr | $52/mo | $105/mo | $315/mo | Rent |
 eBayUpdated 6 hr. ago | $1,399 | Payback ~4.5 yrsPower $7/mo | Payback ~2.3 yrsPower $14/mo | Payback ~9 moPower $42/mo | Buy |
Break-even uses the US residential average ($0.18/kWh) and 320 W rated power, excluding resale value and the rest of the build.
Rental links are referral links: we may receive a commission or credit if you sign up through them. As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Amazon.jp
Xianyu