GIGABYTE Says Next-Gen PCIe Gen5 GPUs Will Move from 8-Pin to New 16-Pin Power Connectors
The press release confirms previous reports regarding PCIe Gen5 graphic cards.
GIGABYTE today issued a press release today announcing their brand new, flagship power supply: the UD1000GM 5.0. As the name suggests, the UD1000GM 5.0 is a 1000W PSU with an 80 Plus Gold efficiency rating and flagship features but almost none of that matters. The exciting bit here is that this is actually a PCIe Gen5 compliant power supply (one of the very first), meaning it can power graphic cards using that PCIe Gen5 spec.
PCIe Gen5 isn’t here yet, at least on mainstream GPUs. That’s why this PSU comes with traditional 8 pin (6+2) power connectors that are used today, but also a new 16-pin PCIe Gen5 power connector as well. The connector is officially called “12VHPWR” and is meant to be used with future graphics cards based on the PCIe Gen5 spec that we will see adopted in AMD and NVIDIA‘s next generation of desktop GPUs.
This 12VHPWR, or 16-pin connector is supposed to replace the existing 8-pin power connectors that are becoming outdated. Out of those 16, four are data pins while the other twelve are main power connectors. Hence, a 16-pin power connector is also a 12+4-pin power connector, just like an 8-pin connector is also a 6+2-pin connector today.
Today’s flagship cards need three 8-pin power connectors with NVIDIA’s Founder’s Edition cards using a 12-pin to 8-pin adaptor to convert their 12-pin to the current industry standard of 8-pin. What’s interesting is that 12-pin connector that NVIDIA used is the exact same as the PCIe Gen5 16-pin power connector, just without the extra 4 data pins. That’s why NVIDIA provides an adapter to adapt it down to PCIe Gen4.
The new 16-pin power connector
Current 8-pin power connectors can provide only 150W of power per connector, which means three of those can pump out up to 450W of power, a limit very close to being touched by current flagship GPUs. In comparison, the new 16-pin power connector can provide up to 600W of power all on its own. That means you need just one connector to power even the beefiest of next-gen GPUs. This, of course, reduces cable clutter and makes things simpler.
The size of the 16-pin power connector is also not bigger than the 8-pin with both largely being the same, so you won’t have to worry about bulky cables or connectors either. GIGABYTE mentions in its press release that next-gen GPUs will either use the 16-pin cable by default or use three 8-pin to one 16-pin cable adapters. That’s why having a PCIe Gen5 16-pin power connector on the power supply means you’d have to only run one straight cable from your GPU to your PSU, no adapters needed.
Official PCI-SIG PCIe Gen5 power specs haven’t been released yet so there’s no telling if GIGABYTE’s 16-pin (12+4) is identical to NVIDIA’s 12-pin that lacks the data pins. NVIDIA never confirmed how much power its 12-pin power connector can pull so we don’t the similarities between that and the standard 12-pin (16-pin but without the 4 data pins) PCIe Gen5 power connector.
PCIe Gen5 means 16-pin power connectors
GIGABYTE’s press release essentially confirms that card based on PCIe Gen4 won’t be able to use this new 16-pin power connector and that if a card wishes to be compliant with the PCIe Gen5 spec then it will have to use the 16-pin power connector as traditional 8-pin connectors won’t cut it anymore. This suggests that if you want to upgrade to one of the next-gen PCIe Gen5 graphic cards, you’d either have to buy a new PSU or rely on the aforementioned adapters to convert 16-pin to triple 8-pin.
GIGABYTE’s UD1000GM 5.0 also comes in a non-PCIe Gen5 flavor, that PSU cogently omits the “5.0” from its name. But, the PCIe Gen5 UD1000GM 5.0 is only the second power supply to be announced supporting the official PCIe Gen5 spec with ASUS‘s Loki being the first. This process will likely accelerate in the coming months as more and more manufacturers will announced their own PCIe Gen5 power supplies, in time for the release of next-gen GPUs from NVIDIA, AMD, and perhaps even Intel.