Hugging Face's logo

Datasets:
iBlessi
/
rtx-5080-llm-power-efficiency

Modalities:
Tabular
Text
Formats:
csv
Size:
< 1K
Tags:
llm
power
energy
efficiency
rtx-5080
gpu
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
Dataset Viewer
Auto-converted to Parquet Duplicate
timestamp
stringlengths
32
32
power.draw
float64
52.4
345
clocks.gr
int64
2.75k
3.29k
clocks.mem
int64
17.6k
17.8k
temperature.gpu
int64
32
51
utilization.gpu
int64
0
93
utilization.memory
int64
1
80
memory.used
int64
1.07k
13.6k
fan.speed
int64
30
30
pstate
stringclasses
2 values
2026-08-20T00:13:41.859447+00:00
53.48
2,745
17,750
34
0
1
1,076
30
P0
2026-08-20T00:13:42.894976+00:00
53.35
2,752
17,750
34
0
1
1,076
30
P0
2026-08-20T00:13:43.936992+00:00
53.28
2,752
17,750
34
0
1
1,076
30
P0
2026-08-20T00:13:44.971131+00:00
54.4
2,797
17,750
34
0
1
1,303
30
P0
2026-08-20T00:13:46.004523+00:00
56.87
2,857
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:47.051571+00:00
62.44
3,112
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:48.097690+00:00
60.15
2,745
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:49.145227+00:00
54.85
2,752
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:50.182400+00:00
54.58
2,745
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:51.215912+00:00
54.77
2,745
17,550
34
9
2
13,341
30
P1
2026-08-20T00:13:52.251346+00:00
54.37
2,782
17,550
34
0
1
13,581
30
P1
2026-08-20T00:13:53.283959+00:00
143.57
3,270
17,550
46
84
68
13,593
30
P1
2026-08-20T00:13:54.324742+00:00
274.99
3,270
17,550
40
25
22
13,593
30
P1
2026-08-20T00:13:55.355793+00:00
233.65
3,270
17,550
48
84
67
13,593
30
P1
2026-08-20T00:13:56.386017+00:00
228.8
3,277
17,550
47
84
68
13,593
30
P1
2026-08-20T00:13:57.423661+00:00
290.28
3,270
17,550
49
84
68
13,593
30
P1
2026-08-20T00:13:58.454407+00:00
123.01
3,285
17,550
36
0
1
13,593
30
P1
2026-08-20T00:13:59.503861+00:00
66.96
3,285
17,550
36
0
1
13,593
30
P1
2026-08-20T00:14:00.538159+00:00
53.11
2,745
17,550
35
0
1
13,593
30
P1
2026-08-20T00:14:01.574872+00:00
52.62
2,752
17,550
34
0
1
13,593
30
P1
2026-08-20T00:14:02.608077+00:00
52.62
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:03.643541+00:00
52.7
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:04.677842+00:00
52.53
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:05.712746+00:00
52.69
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:06.753952+00:00
52.51
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:07.795023+00:00
52.37
2,745
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:08.826476+00:00
52.6
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:09.863011+00:00
52.4
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:10.900173+00:00
52.38
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:14:11.943088+00:00
52.36
2,752
17,550
34
0
1
13,591
30
P1
2026-08-20T00:12:10.692932+00:00
52.97
2,850
17,750
33
4
1
1,105
30
P0
2026-08-20T00:12:11.724268+00:00
53.93
2,782
17,750
32
0
1
1,105
30
P0
2026-08-20T00:12:12.767476+00:00
53.4
2,752
17,750
32
0
1
1,105
30
P0
2026-08-20T00:12:13.821487+00:00
54.68
2,775
17,550
32
0
1
3,252
30
P1
2026-08-20T00:12:14.851469+00:00
116.55
3,202
17,550
45
82
61
3,810
30
P1
2026-08-20T00:12:15.887999+00:00
265.19
3,270
17,550
47
82
61
3,812
30
P1
2026-08-20T00:12:16.920298+00:00
264.37
3,270
17,550
47
82
61
3,811
30
P1
2026-08-20T00:12:17.960454+00:00
207.69
3,285
17,550
35
0
1
3,811
30
P1
2026-08-20T00:12:18.997749+00:00
67.03
3,285
17,550
34
0
1
3,811
30
P1
2026-08-20T00:12:20.026981+00:00
58.48
2,752
17,550
33
0
1
3,811
30
P1
2026-08-20T00:12:21.069074+00:00
53.41
2,745
17,550
33
0
1
3,811
30
P1
2026-08-20T00:12:22.108259+00:00
53.45
2,752
17,550
33
0
1
3,810
30
P1
2026-08-20T00:12:23.150722+00:00
53.12
2,745
17,550
33
0
1
3,810
30
P1
2026-08-20T00:12:24.191248+00:00
53.12
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:25.228705+00:00
53.15
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:26.259091+00:00
53.17
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:27.301839+00:00
53.04
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:28.341039+00:00
53.16
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:29.381779+00:00
53.25
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:30.422898+00:00
53.24
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:12:31.457425+00:00
53.1
2,752
17,550
33
0
1
3,809
30
P1
2026-08-20T00:13:06.074651+00:00
53.05
2,745
17,750
33
0
1
1,076
30
P0
2026-08-20T00:13:07.104660+00:00
52.85
2,752
17,750
33
0
1
1,076
30
P0
2026-08-20T00:13:08.149223+00:00
52.96
2,752
17,750
33
0
1
1,076
30
P0
2026-08-20T00:13:09.180641+00:00
54.96
2,752
17,550
33
24
5
9,453
30
P1
2026-08-20T00:13:10.215673+00:00
55.66
2,752
17,550
33
9
2
9,453
30
P1
2026-08-20T00:13:11.251958+00:00
54.55
2,752
17,550
33
9
2
9,453
30
P1
2026-08-20T00:13:12.295343+00:00
54.59
2,745
17,550
33
9
2
9,453
30
P1
2026-08-20T00:13:13.328335+00:00
54.56
2,752
17,550
33
9
2
10,371
30
P1
2026-08-20T00:13:14.364387+00:00
220.96
3,060
17,550
47
93
80
10,381
30
P1
2026-08-20T00:13:15.406693+00:00
310.29
3,075
17,550
48
90
76
10,381
30
P1
2026-08-20T00:13:16.448613+00:00
308.75
3,075
17,550
44
76
65
10,383
30
P1
2026-08-20T00:13:17.485105+00:00
284.76
3,090
17,550
49
93
79
10,383
30
P1
2026-08-20T00:13:18.518498+00:00
313.82
3,120
17,550
49
93
79
10,383
30
P1
2026-08-20T00:13:19.550600+00:00
281.82
3,247
17,550
50
91
77
10,383
30
P1
2026-08-20T00:13:20.587485+00:00
319.69
3,075
17,550
49
93
79
10,383
30
P1
2026-08-20T00:13:21.626750+00:00
313.9
3,075
17,550
51
93
79
10,383
30
P1
2026-08-20T00:13:22.669903+00:00
84.6
3,285
17,550
36
0
1
10,384
30
P1
2026-08-20T00:13:23.710446+00:00
63.57
2,752
17,550
35
0
1
10,384
30
P1
2026-08-20T00:13:24.750292+00:00
53.66
2,745
17,550
35
0
1
10,384
30
P1
2026-08-20T00:13:25.791635+00:00
53.64
2,745
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:26.835594+00:00
53.53
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:27.877230+00:00
53.55
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:28.917682+00:00
53.59
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:29.949141+00:00
53.41
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:30.979972+00:00
53.54
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:32.024361+00:00
53.48
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:33.071112+00:00
53.25
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:34.115168+00:00
53.35
2,760
17,550
34
1
1
10,384
30
P1
2026-08-20T00:13:35.162604+00:00
52.7
2,760
17,550
34
0
1
10,384
30
P1
2026-08-20T00:13:36.204593+00:00
52.95
2,752
17,550
34
0
1
10,384
30
P1
2026-08-20T00:12:37.421837+00:00
52.78
2,752
17,750
32
0
1
1,072
30
P0
2026-08-20T00:12:38.460039+00:00
52.51
2,752
17,750
32
0
1
1,072
30
P0
2026-08-20T00:12:39.496485+00:00
52.8
2,752
17,750
32
0
1
1,072
30
P0
2026-08-20T00:12:40.546112+00:00
53.92
2,790
17,750
33
1
1
5,473
30
P1
2026-08-20T00:12:41.589897+00:00
61.33
2,812
17,550
33
8
8
5,875
30
P1
2026-08-20T00:12:42.625181+00:00
262.67
3,270
17,550
50
88
75
5,875
30
P1
2026-08-20T00:12:43.660515+00:00
306.7
3,270
17,550
51
88
75
5,877
30
P1
2026-08-20T00:12:44.700258+00:00
306.24
3,277
17,550
50
88
75
5,877
30
P1
2026-08-20T00:12:45.741416+00:00
344.84
3,270
17,550
51
88
74
5,877
30
P1
2026-08-20T00:12:46.773289+00:00
141.89
3,277
17,550
35
0
1
5,877
30
P1
2026-08-20T00:12:47.808719+00:00
67.04
3,277
17,550
35
0
1
5,877
30
P1
2026-08-20T00:12:48.853682+00:00
55.97
2,752
17,550
34
0
1
5,877
30
P1
2026-08-20T00:12:49.884462+00:00
53.23
2,745
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:50.917891+00:00
53.04
2,752
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:51.969964+00:00
53.13
2,752
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:53.000348+00:00
53.04
2,752
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:54.037779+00:00
53.06
2,752
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:55.083932+00:00
53.08
2,752
17,550
33
0
1
5,877
30
P1
2026-08-20T00:12:56.126947+00:00
52.92
2,752
17,550
33
0
1
5,877
30
P1
End of preview. Expand in Data Studio

RTX 5080 LLM Power Efficiency: Watts and Joules per Token

Measured board power, tokens per joule, and electricity cost per million tokens for local LLM inference on a retail RTX 5080, with 1Hz telemetry.

What this measures

Board power logged with nvidia-smi at 1Hz while driving fixed-length generations on a retail RTX 5080, converted to tokens per joule and dollars per million tokens.

Efficiency

Model tokens/joule median board W
Llama 3.2 3B 1.20 265
gpt-oss 20B (sparse, MXFP4) 0.665 292
Qwen 2.5 7B 0.53 344 (the peak, not the biggest model)
Qwen 2.5 14B 0.317 312

The sparse 20.9B model is roughly twice as efficient per joule as the dense 14B. Architecture and quantization beat parameter count on the efficiency axis, not only on speed.

The finding almost nobody publishes for consumer cards

This card does not idle where every 24/7 cost model assumes. Across a week of captures it sat at 52-71W at the Windows desktop, pinned in P0 with graphics clocks near 2.9GHz. The cleanest state achievable (wallpaper closed, no model resident) still read 53.7W. Reviews quote single-digit-to-15W idle.

At the May-2026 EIA US residential average of $0.184/kWh, 52-71W around the clock is 456-622 kWh, roughly $85-115/year before a single token is generated.

That inverts the usual budget framing: generation itself costs only $0.04-$0.16 per million tokens, so model choice barely moves a home node's bill. What the box does during the other 23 hours dominates it.

Check your own card with nvidia-smi --query-gpu=power.draw --format=csv before trusting any spec-sheet idle figure, mine included.

Contents

Summary rows plus the raw 1Hz telemetry captures.

Source and license

Canonical page, full method and change log: https://techfuelhq.com/data/rtx-5080-llm-power-efficiency/

Licensed CC BY 4.0. Attribution: "TechFuelHQ" linking to the canonical page.

Corrections welcome. If a row is wrong it gets fixed against the primary source and the correction is logged.
Downloads last month
57
Number of rows:
104
Total file size:
24.5 kB