Single rail and multiple PCIe modular connections


First time build: Planning for an eventual four GPUs I’ve over-engineered a bit with the PSU: a Supernova P2 1600. (Well might not be that overboard at the end of the day if I really get four x080 Ti’s in there :wink:

Based on the many forum threads, I plan to power my 2080 Ti(s) with separate cables.

Question: The Supernova is a single rail PSU. Is there any difference between the “VGA” (PCIe) connectors 1-4 and 5-9 which are designated differently in the manual? (See below). My naive approach is just to treat them all as 8-pin power sources… but is there any magic in those 5-9 connectors that specifically say they are for “6pin plus 8(6+2)”. Seems like they are just begging me to plug one of those dual cables in which are included in the box.

Thanks for the feedback.

VGA1-4 will end as a single 6+2 pin connector, so basically a single 8-pin connector.

VGA5-9 will end as a 6+2 pin connector with and additional 6-pin connector piggybacked off of the 6+2.

Just treat all the 6+2 pin connectors from 1-9 as 8-pin connectors and leave all the piggybacked 6-pin connectors from 5-9 hanging.

1 Like

Thanks. Yup, I gathered I could just leave the extra pins hanging. (Or I could just buy extra 6+2 -> 6+2 single cables to get rid of the extra 6 pins all together.)

I am just curious why EVGA would go through the trouble of actually designating 5-9 separately when the actual power is the same for all nine.

Anyway, thanks. I just wanted to make sure. Appreciate the quick feedback.

I loathe piggybacked cables, but I understand some people prefer to use a single 6+2+6 cable for cards that use 6+6 or 8+6 connectors. I certainly wouldn’t use them for beefcake cards like a 2080 Ti, but for a 2060 they’d most likely be fine.

EVGA probably just did it to be accurate since the cable layout is different.

yeah. Once I’ve gone through a few income cycles and I’m ready to acquire more GPUs I’ll happily splurge for a few more cables.

I guess the 3080 Tis are on deck for the fall. Will be interesting to see where they land for price point.

This is all part of a “dream build” for AI/ML. (Not really a gamer.) But Cuda cores are sexy as hell!

Thanks again.

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.