Datasets:

iBlessi
/

rtx-5080-llm-throughput

capture_date_utc stringdate 2026-08-12 00:00:00 2026-08-12 00:00:00	method_version stringclasses 1 value	gpu stringclasses 1 value	driver float64 611 611	ollama_version stringclasses 1 value	model stringclasses 4 values	model_digest12 stringclasses 4 values	parameter_size stringclasses 4 values	quantization stringclasses 2 values	num_ctx int64 4.1k 16.4k	prompt_tokens int64 471 1.67k	reps int64 3 3	median_decode_tok_s float64 94.5 313	median_prefill_tok_s float64 4.86k 19.4k	median_load_ms float64 123 299	vram_resident_mib int64 5.5k 15.7k
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	gpt-oss:20b	17052f91a42e	20.9B	MXFP4	4,096	503	3	190.14	9,078.28	297.7	14,067
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	gpt-oss:20b	17052f91a42e	20.9B	MXFP4	4,096	1,658	3	187.27	10,464.66	279.5	14,060
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	gpt-oss:20b	17052f91a42e	20.9B	MXFP4	16,384	503	3	188.63	9,286.61	299.2	14,355
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	gpt-oss:20b	17052f91a42e	20.9B	MXFP4	16,384	1,658	3	187.73	10,554.32	280.2	14,355
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:14b	7cdf5a0187d5	14.8B	Q4_K_M	4,096	488	3	96.65	5,202.23	182.5	10,841
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:14b	7cdf5a0187d5	14.8B	Q4_K_M	4,096	1,674	3	94.48	4,876.41	125.8	10,855
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:14b	7cdf5a0187d5	14.8B	Q4_K_M	16,384	488	3	97.02	5,225.96	131.3	13,157
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:14b	7cdf5a0187d5	14.8B	Q4_K_M	16,384	1,674	3	94.58	4,864.67	128.9	13,159
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	llama3.2:3b	a80c4f17acd5	3.2B	Q4_K_M	4,096	471	3	311.96	18,196.57	162.8	15,682
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	llama3.2:3b	a80c4f17acd5	3.2B	Q4_K_M	4,096	1,657	3	298.21	17,681.08	156.2	15,682
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	llama3.2:3b	a80c4f17acd5	3.2B	Q4_K_M	16,384	471	3	312.89	17,885.62	162.9	5,495
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	llama3.2:3b	a80c4f17acd5	3.2B	Q4_K_M	16,384	1,657	3	298.59	19,397.36	158	5,501
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:7b	845dbda0ea48	7.6B	Q4_K_M	4,096	488	3	180.02	10,152.28	129.8	6,190
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:7b	845dbda0ea48	7.6B	Q4_K_M	4,096	1,674	3	176.79	9,672.39	123.4	6,190
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:7b	845dbda0ea48	7.6B	Q4_K_M	16,384	488	3	179.66	10,133.1	127.9	7,022
2026-08-12	llm-throughput-ollama-v1	NVIDIA GeForce RTX 5080	610.88	0.32.1	qwen2.5:7b	845dbda0ea48	7.6B	Q4_K_M	16,384	1,674	3	176.61	9,984.37	127.7	7,034

RTX 5080 Local LLM Throughput — Measured, Not Estimated

Measured decode and prefill tokens/second, load times, and VRAM residency for local LLMs on a single retail NVIDIA GeForce RTX 5080 (16 GB): gpt-oss:20b (MXFP4), Qwen 2.5 14B and 7B (Q4_K_M), and Llama 3.2 3B (Q4_K_M), at 4k and 16k context with ~500 and ~1,650-token prompts.

Method

Every figure is the median of 3 runs using ollama's native eval_count/eval_duration counters — never wall-clock division.
Every repetition prepends a unique nonce so each run pays a full prefill; without this, warm runs hit ollama's prompt cache and report ~60,000 tok/s cache-hit rates that measure nothing.
Generation pinned: temperature 0, seed 42, 256-token target output.
Model identity pinned by digest per row (a tag like qwen2.5:14b can silently move upstream).
Bench environment recorded per row: RTX 5080, driver 610.88, ollama 0.32.1, Windows 11, Ryzen 7 7800X3D.

Headline results

Model	Quant	Decode tok/s	Resident VRAM
Llama 3.2 3B	Q4_K_M	298–313	5.5 GiB (alone)
gpt-oss 20B	MXFP4	187–190	14.0–14.4 GiB
Qwen 2.5 7B	Q4_K_M	177–180	6.2–7.0 GiB
Qwen 2.5 14B	Q4_K_M	94–97	10.8–13.2 GiB

The 20.9B MXFP4 model decodes 2x faster than the dense 14B at Q4_K_M and matches the dense 7B — architecture and quantization beat parameter count. Decode speed holds flat from 4k to 16k context at these prompt sizes.

Caveats

Single retail unit, single driver, single runtime — these are this card's numbers. The VRAM column is whole-GPU nvidia-smi at measurement time and can include a previously-benched model still resident (rows measured alone are the lower readings).

Canonical source

Method detail, change log, and living updates: https://techfuelhq.com/data/rtx-5080-llm-throughput/

License

CC BY 4.0 — attribute as "TechFuelHQ RTX 5080 LLM Throughput Dataset" linking to the canonical page.

Downloads last month: 25

Number of rows:

Total file size:

7.3 kB