2x NVIDIA GeForce RTX 3090
Compare 2x NVIDIA GeForce RTX 3090 against another machine in the Arena
Reference: GeForce RTX 30 series on Wikipedia
- Class
- Multi-GPU
- Memory
- 48 GB GDDR6X
- Runs (Q4_K_M)
- —
- Bandwidth
- 936 GB/s
- TDP
- 700 W
- Released
- 2020-09-24
- Price (US)
- $2548 used as of 2026-08
Sourcing and disambiguation notes
Bandwidth is per card (936.2 GB/s, NVIDIA spec): a layer-split model reads from one card at a time, so decode speed tracks a single card's bandwidth, not the sum. TDP summed at stock 350W/card; the source rig this hardware id is benchmarked against ran both cards at a 115% power limit (up to ~805W combined) to chase throughput. Price is 2x the Jan-2026 ~$800 used-eBay figure from the source. 2026-08 update: recalculated as 2x the current $1,274 single-card used price (resaleprices.com), superseding the Jan-2026 $1,600 estimate previously recorded. No actual two-card pair listing was found (one anecdotal $650 eBay story describes a lucky mispriced Dell OEM pair, not a market rate) -- this figure is CALCULATED, not an independently sourced pair price. New stays null since the single-card new price is also null (discontinued).
Specification sources
- https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/ — vendor, 2020-09-24
Measurements
Decode is token generation — the speed you feel while an answer streams. Prefill is prompt processing — the wait before it starts. Why bandwidth predicts decode speed.
No records yet for this hardware. Know of a published benchmark on it? Submit the link.
What verified, single-source and estimated mean, and the same rows with every filter and sort in the benchmarks explorer.
Speed over time
What it can run
Fit is arithmetic, not a measurement — how it is computed. Against 48 GB; models that fit are listed largest first.
No modeled quant fits in 48 GB.