Skip to main contentThe Local Bench

GeForce RTX 5090 32 GB

NVIDIA · Discrete GPU · as of 2026-10-12

Specs sourced · published reference runs · catalogue speeds mostly estimated
Discrete graphics card

  • Data checked: 2026-10-12. Review interval shown below. Prices need review after 30 days; specifications and benchmark collections after 90 days. Dates describe the recorded check, not a live retailer quote.

    • Verified Checked against the primary source (for example the model’s config.json on Hugging Face).
    • Sourced Taken from a cited source; not yet independently checked.
    • Estimate Calculated with the published formulas; not a measurement.
    • Sample Placeholder data for layout; not a real figure.
    • Not yet measured No public, citable score found for this benchmark yet. We never fill the gap with a guess.

    Verified metadata is not a measured performance result. Only cited runs matching model, device, quant and recorded context can replace speed estimates. Estimates assume one stream and F16 KV cache, with a ±25% uncertainty band. A published run applies only to its recorded runtime and conditions; missing settings are unknown.

    • llama-2-7b · rtx-5090-32g
      llama.cpp (CUDA) · Q4_0 · tg128
      Recorded 2026-10-12 · historical source record · repeat count unknownSourced
      Llama 2 7B Q4_0, llama-bench tg128, full GPU offload, flash attention off, single GPU. Community scoreboard entry; commit and submitter in the source table. Run source

    Evidence labels · Freshness policy · Calculation methodology

32GB
memory, 31.5 GB usableSourced
1,792GB/s
memory bandwidthSourced
$7,199.99
price, USDSourced
8of 10
catalogue models fit at 8k

32 GB and 1.8 TB/s: the fastest single consumer card for decode.

Decode-speed ceiling for a model is 1,792 GB/s ÷ GB read per token. For a 16.6 GB file that is 108 tok/s; real runtimes reach roughly 60% of it. Why bandwidth is the limit.

See every model on this device

Specifications

Memory
32 GBSourcedVendor spec page, checked 2026-10-12. Source
Bandwidth
1,792 GB/sSourcedNVIDIA's GeForce spec pages list bus width but not bandwidth; figure from the cited Wikipedia spec table (bus width cross-checked on nvidia.com, 2026-10-12). Source
FP16 dense compute
209 TFLOPSSampleSample FP16 dense compute; pending vendor-sheet citation.
Host RAM (offload)
64 GBSampleAssumed host RAM for offload.
Vendor TDP / board power
575 WSourcedTotal Graphics Power, nvidia.com. Source
Load power at wall
650 WSampleSample wall power under load.
Price, USD
$7,199.99SourcedLowest new in-stock listing on Newegg US (PNY ARGB OC), far above the US$1,999 launch MSRP. Checked 2026-10-12. Source
Price, CAD
CA$9,036SourcedLowest new in-stock listing on Newegg Canada (ASUS TUF, marketplace seller). Checked 2026-10-12. Source

Models that run

Default quant, single stream, 8k context. Fastest first; estimates are ±25% until measured.

Model
Quant
Memory
Verdict
Decode, tok/s
Max context
gpt-oss-20bMXFP412.9 GBFits469.6Estimate128k
Qwen3 30B-A3BQ4_K_M19.9 GBFits382.4Estimate32k
Llama 3.1 8B InstructQ4_K_M6.6 GBFits179.4Estimate128k
Qwen3 8BQ4_K_M6.8 GBFits172.4Estimate32k
Qwen3 14BQ4_K_M10.9 GBFits104.0Estimate32k
Mistral Small 3.2 24BQ4_K_M16.2 GBFits68.6Estimate64k
Gemma 3 27B ITQ4_K_M18.2 GBFits60.9Estimate128k
Qwen3 32BQ4_K_M22.5 GBFits49.1Estimate32k
gpt-oss-120bMXFP464.3 GBNeeds offload21.6Estimate—
Llama 3.3 70B InstructQ4_K_M45.8 GBNeeds offload2.1Estimate—

Published reference runs

Third-party llama.cpp measurements on this device, with conditions and source. Llama 2 7B Q4_0 runs are the community calibration test; they are not catalogue models.

Model
Quant
Date
Runtime · test
Decode, tok/s
Conditions
llama-2-7bQ4_02026-10-12llama.cpp (CUDA) · tg128290.0SourcedLlama 2 7B Q4_0, llama-bench tg128, full GPU offload, flash attention off, single GPU. Community scoreboard entry; commit and submitter in the source table. Source

Usable memory: discrete VRAM − 0.5 GB reserve; unified RAM × 0.75; fits when total ≤ 90% of usable (concept brief §7.3). Markdown · JSON