This census collects the current on each of the six 12 V pins of a 16-pin graphics card connector, read by owners whose hardware measures it. Rows are grouped by cable type, and a group publishes once it holds five readings taken at 6 A total or more. It opened on September 29, 2026 with zero rows.
- Dataset ID
techfuelhq-pin-current-census- Version
- 0.1.0
- Format
- CSV, HTML
- Published
- Page updated
A 16-pin graphics card connector carries its whole 12 V load on six pins. Intel’s design guide rates each of those pins at 9.2 A and the connector at up to 600 W, and at 600 W the six pins share 50 A, so an even split puts 8.3 A on each. That leaves under one amp of room per pin.
Nothing in the connector makes the split even. It depends on how well each pin seats, and until recently an owner had no way to see it. Some hardware now measures every pin. ASUS puts a sensor chip on its ROG Astral cards, Thermal Grizzly sells an inline meter that logs each pin, and owners post their six numbers in forum threads and issue trackers, one card at a time. A search in September 2026 turned up no dataset that collects those readings and sets them beside the cable they came from.
This page opens that census. One row is one connector on one card, read once under one load. The data, the field definitions, the validator and the report form live in the open repository: github.com/iBlessi/techfuelhq-census (CC BY 4.0).
If you are here because a connector looks or smells wrong, shut the machine down first. The symptoms of a failing supply are on the PSU failure symptoms guide.
How results are reported
The census publishes distributions. For each cable type it gives the number of readings, the median ratio of the busiest pin to the average pin, the lowest and highest ratio, and how many readings had a pin at 9.2 A or more. The sample size sits beside every figure.
It publishes no verdict on a card, a cable or a power supply. A reading describes one connector as it was seated on one day, and the same cable pulled out and pushed back in can read differently the next, which is why every row records how many months the cable has been in the card.
What one row records
A row is one connector on one card, read once under one load. Of its 31 fields, you fill in 13, the reading fills in 13, 3 are worked out from the row and the maintainer sets 2.
All 31 fields of a row
| Field | Filled by | What it records |
|---|---|---|
card_ | You | Card brand. As printed on the box: ASUS, MSI, Gigabyte. |
card_ | You | Card model. The model line: ROG Astral RTX 5080 OC. |
gpu | You | GPU. The chip, written like RTX 5080 or RX 9070 XT. |
sensor | You | What measured it |
capture | The reading | Capture |
cable_ | You | Cable. What carries power from the supply to the card. |
inline_ | You | Anything between cable and card |
psu_ | You | Power supply brand |
psu_ | You | Power supply model. With its wattage if that is part of the name: RM1000x. |
psu_ | You, optional | Power supply watts |
months_ | You | Months this cable has been in this card. Whole months since it was last plugged in. A cable reseated last week is 0. |
load_ | You | What the card was doing |
load_ | You, optional | Which one. The game or test by name. |
samples | The reading | Samples counted |
duration_ | The reading | Seconds covered |
total_ | The reading | Total current, mean (A) |
pin1_ | The reading | Pin 1, mean (A) |
pin2_ | The reading | Pin 2, mean (A) |
pin3_ | The reading | Pin 3, mean (A) |
pin4_ | The reading | Pin 4, mean (A) |
pin5_ | The reading | Pin 5, mean (A) |
pin6_ | The reading | Pin 6, mean (A) |
peak_ | The reading | Highest single pin seen (A) |
peak_ | The reading | Highest total seen (A) |
min_ | The reading, optional | Lowest pin voltage seen (V) |
imbalance | Worked out from the row | Busiest pin over the average pin |
max_ | Worked out from the row | Busiest pin's share of the total (%) |
band | Worked out from the row | Load band |
notes | You, optional | Notes |
submitted_ | The maintainer | The day the report was sent |
source_ | The maintainer | The issue the row came from |
Cable types
| Value | Meaning |
|---|---|
native-16pin | The supply maker's cable, 16-pin at both ends |
native-8pin-psu-side | The supply maker's cable, 8-pin sockets at the supply, 16-pin at the card |
boxed-adapter | The adapter that came in the card's box |
third-party | A cable or adapter from another company |
Load bands
The band comes from the total current in the row.
| Value | Meaning |
|---|---|
idle | Under 6 A total |
moderate | 6 to 25 A total |
high | 25 A total and above |
The 6 A floor follows astral-hwmon’s reference measurements. Its notes record that the sensor reports current in 20 mA steps, that the pin shares scatter at low current for that reason, and its own share figures use samples at 6 A total and above. I took the same floor.
How to take the reading
- Put the card under a steady load for at least a minute: a game you play, a benchmark loop, or a render.
- Read the six pins while the load runs. A log is better than a glance, because it holds every sample.
- Note which cable carries the power, what sits between the cable and the card, and how many months the cable has been plugged in.
| What you have | How to read it |
|---|---|
| A ROG Astral card on Windows | GPU Tweak III shows six pin currents under Power Detector+. Type the six numbers into the builder below. |
| A ROG Astral card watched by 12VHPWR Guard | The guard writes a flight file to its logs folder only when it limits the card or shuts the machine down. The file holds the minute before. It opens in the builder, and its row is kept apart from ordinary readings. |
| A ROG Astral card on Linux | A session file from astral-hwmon’s log-session.py opens in the builder. |
| A WireView Pro II or another meter | Type the six numbers it shows. |
A log is summarised over the part of it that was under load. If 20 or more samples sit at 25 A total or above, the row uses those; otherwise it uses every sample at 6 A or above, or the whole log when fewer than 20 samples reach that. Peaks are kept apart. The highest single reading anywhere in the log is recorded beside the means.
Send a report
What you paste is read on this page, inside your browser. None of it is sent anywhere until you open the report on GitHub yourself.
Sending needs a GitHub account, because the report is filed as an issue. No account? Copy the report and email it to contact@techfuelhq.com.
What has been published
Current count: 0 reports. No group has rows yet, so nothing is published.
A cable type publishes once it holds five readings taken at 6 A total or more. Below that it is listed as collecting, with its count and no figures.
Limits of this data
- Reports come from people who chose to send them. An owner who saw a worrying number is more likely to report than one who saw nothing. The figures describe the rows in the dataset.
- Only owners of measuring hardware can report. That is mostly ROG Astral cards, so the census says little about other cards until inline meters add rows.
- The readings are as good as the sensor. ASUS’s guide to Power Detector+ gives no accuracy figure for it.
- A reading at or above 9.2 A is counted and reported as a count. What it means for that connector is between its owner and the maker of the card and cable.
Use the data
The dataset is one file. submissions.csv holds one row per reading, in the field order above, under Creative Commons Attribution 4.0. Attribute it as “TechFuelHQ open census” with a link to this page.
To size a supply for a card like these, the PSU wattage calculator works from the parts in the build. The coil whine census and the GPU sag census are the sibling datasets.
Change log
- 0.1.0 (2026-09-29): fields, load bands, report builder, validator and intake form. The census opened with 0 rows.
Frequently asked questions
How many amps should each pin of a 12V-2x6 connector carry?
Which graphics cards can show the current on each pin?
What does H++ on a graphics card power connector mean?
Does a reading taken at the desktop count?
What license is the data under?
Sources and corrections
- Last updated
- Methodology
- See our methodology for research and review standards. It draws on 7 cited sources, listed below, each checked against the original page on the date above. The census opened with zero rows on 2026-09-29. The report builder on this page and the repository that receives reports run the same code, which is tested against a capture published with astral-hwmon: it reproduces that capture’s stated 322 samples, 48.26 A peak total, 8.56 A on the busiest pin and 11.944 V minimum. The connector rating and the H++ marking are quoted from Intel’s design guide. Which cards measure each pin is taken from ASUS and from the tools’ own documentation.
- Sources
- Intel: ATX Version 3 Multi Rail Desktop Platform Power Supply Design Guide, revision 2.1a (November 1, 2023), PCI Express Add-in Card Connectors accessed 2026-09-29
- ASUS ROG: How GPU Tweak's Power Detector+ alerts you to abnormal current on your ROG Astral graphics card (April 20, 2025) accessed 2026-09-29
- VideoCardz: ASUS ROG Astral RTX 50 gets 12V-2x6 per-pin monitoring on Linux (August 18, 2026) accessed 2026-09-29
- GitHub: ksokolowski/astral-hwmon, reference measurements of 2026-08-14 accessed 2026-09-29
- GitHub: humza-khalid/12vhpwr-guard accessed 2026-09-29
- Thermal Grizzly: WireView Pro II accessed 2026-09-29
- GitHub: iBlessi/techfuelhq-census (definitions, validator, intake form, CSV) accessed 2026-09-29
- Update log
- 2026-09-29 — Page updated.
- Corrections
- Spotted an error or a stale number? Email contact@techfuelhq.com. Confirmed corrections are added to the update log above.