I’ll put this part of the configuration via the aquasuite in front of the rest of the settings before I go into the other points later, because the following is unique and really new among all the 12V2X6 solutions. It also shows what kind of changes have to be made just to protect the high material assets of customers, because a “standardization mafia” like the PCI SIG has failed so badly and continues to do so.
Maximum load per pin and for all pins combined under continuous load (update)
The standard specifies 9.5 amps per pin as the maximum for a single contact. This theoretically adds up to a little over 680 watts peak. However, there is a significant reduction here, and you cannot simply multiply the 9.5 amps per pin by six to derive a real continuous load. These 9.5 amps are an ideal value that only applies if a single contact is perfectly aligned, clean, and thermally insulated. In practice, however, it is not just one pin that heats up, but all contacts influence each other. The more pins carry current at the same time, the more the contact resistance and local temperatures rise, reducing the load capacity of each individual contact.
For this reason, a so-called derating is applied in the standards and also in the PCI-SIG specification. This means that if all six power pins are loaded simultaneously, only about 7 to 8 amps can be expected per contact. This results in a total continuous load of approximately 55 amps, which corresponds to around 600 watts at 12 volts. This value is also officially specified as the permissible continuous power of the 12V-2×6. Now we also know where the trigger value of 7.5 A for balancing comes from.
The 9.5 amps are therefore only considered a theoretical maximum under laboratory conditions and, of course, as a switching point for monitoring, while the 55 amps total load represent the realistic limit for a fully occupied, evenly loaded connector. Even small manufacturing tolerances, slightly misaligned contacts, or minimal contact resistance can alter the temperature balance to such an extent that higher currents quickly become critical.
Escalation logic in the aquasuite
I will now describe the alarm cascade as it is structured with all its logic in the interface and how it behaved in my test. The AMPINEL permanently evaluates the six 12V lines, logs limit violations per wire, total and peak current as well as residual currents and increases the intervention intensity in stages if necessary. The actions per stage can be freely combined, the evaluation and triggering of output actions takes place every second, a short alarm delay in seconds prevents false triggers for short peaks.
Levels 0 to 2 – information and early warning phase
At the lowest level, up to 7.5 A, nothing happens except logging if desired. From level 1, I recommend pure signaling on the device, i.e. visual indications via LED patterns and, if necessary, a short acoustic alarm when one or more pins exceed 9.5 A. This allows you to see immediately if individual lines are approaching the limit value without affecting game operation. Optionally, a Windows pop-up can be activated in parallel and a tacho pulse can be reported to a mainboard fan connection, which can be switched off in the event of an alarm so that you can also recognize the event in the UEFI or in monitoring tools. This tacho signal is connected via S2, the pin assignment is documented in the instructions.
Stage 3 – Controlled disarming in the OS
This is the first stage in which the AMPINEL actively intervenes in the operating sequence. Applications with a high GPU load can be automatically terminated via the aquasuite background service.
In concrete terms, this means
-
When games or 3D rendering software are running, they are specifically closed.
-
The AMPINEL thus prevents further power peaks that could lead to an escalation.
-
It is important to note that the system itself remains intact and does not shut down.
This creates a thermal reserve at the plug contacts and reduces the electrical load on the individual cores without immediately pulling the emergency brake. This is expressly intended and serves as an intermediate step between a pure warning and hard intervention. Especially with new cards and power supply units, which can already show unbalanced loads under standard conditions, this stage is a decisive safety gain: the card is relieved, logs are written and a conscious decision can then be made as to whether to continue or allow a deeper escalation.
Stages 4 and 5 – Controlled shutdown
If the unbalanced load persists or the contact or electronics temperature continues to rise, I escalate to a controlled shutdown. The AMPINEL can simulate the mainboard’s power button via signal output S1, short for a normal shutdown pulse or long for a forced shutdown. Both variants can be assigned to individual stages in the user interface; the S1 interface and its purpose are clearly described.
Stages 6 and 7 – Hard interventions on the card
In extreme cases, the sense signal to the graphics card can be deactivated first and then the 12V supply to the card can be completely disconnected. Completely switching off the graphics card power supply is intended as an extreme action and serves as a fire protection measure if balancing and gentle interventions are no longer sufficient.
New card, new hardware and the first alarm
A new power supply unit, a brand new graphics card, the original 12V 2×6 cable snapped in until it clicks – and yet the AMPINEL immediately sounds the alarm. Four lines above the 9.5 ampere mark, a peak value of 10.68 amperes, 54.03 amperes total current and 4.69 amperes differential current clearly document how skewed the load distribution was. And all this at a hotspot temperature of just 42.75 °C, well before any real thermal problems should have set in. Without AMPINEL, a brand-new system would have been operating directly beyond specification – a prime example of how little trust this connector deserves.
In retrospect, it is precisely moments like this that expose those “media professionals” who tried desperately to blame the problem on the users and labeled me a senile clickbaiter. The story of the supposedly “too stupidly plugged in” plug and a misguided reviewer was quickly written and could be chanted wonderfully, instead of seriously addressing the design weaknesses of the 12V power whip. Instead of technical analysis, there was blame, instead of criticism of the plug design, there was the convenient claim that the users had simply not pressed correctly.
Today the logbook shows (not for the first time!) in black and white that even with new hardware, original cable and correctly locked plug, critical unbalanced loads can occur immediately. It is therefore not a “user error”, but a design problem that can only be mitigated by external protection technology such as the AMPINEL. Cynically, Jensen’s current whip has not only created an accessory industry, but also a whole wave of excuses, which go up in smoke just as quickly as the plugs themselves.






































333 Antworten
Kommentar
Lade neue Kommentare
Moderator
Veteran
Urgestein
Urgestein
1
Urgestein
Urgestein
Urgestein
Veteran
1
Mitglied
Urgestein
1
Urgestein
Neuling
1
Urgestein
1
Veteran
Alle Kommentare lesen unter igor´sLAB Community →