Weird piece of silicon... Seems like the "bad bin" extends to the power draw, or the larger die parasitically increases the power usage...
Where is this from?
These cards are 110% gated by software power limits, no reason to overbuild the board.
That was my first conclusion too, this reeks of the 1060's they released on GP104. Worst possible bin but still stable and even worse perf/watt while being a much larger die. That also says a lot about OC potential. Not exactly a winner.
NVIDIA GP104, 1708 MHz, 1280 Cores, 80 TMUs, 48 ROPs, 6144 MB GDDR5, 2002 MHz, 192 bit
www.techpowerup.com
Back then the backstory was 'too much inventory'. What is it this time? Looks like business as usual to me
It makes sense too for a cost effective performance card; use all those larger dies as well as you can, and those weak bins have accumulated now.
Said this before the supers launched, they should just launch cards with 5 at the end of the name... 2060S=> 2065, 2070S=> 2075, 2080S=> ??
That or they should just have done a full refresh of the lineup as much as people would have hated it...
They kinda did, really, but they needed segmentation in the names between RTX and GTX too and thén still make the updated Turings stand out. It would have been a number-crazy maze even more so. I understand why they picked an "S", it also differentiates Turing RTX further like the 16 series also helps it. This is almost a carbon copy of Kepler Refresh but with an RT lineup alongside. And I think history repeats in that sense too; we got Maxwell afterwards and the initial goal was to have that gen already on 16nm. Now the next one in line is Ampere. I really do hope they won't do another Maxwell with that
Luckily AMD already moved to 7nm so its a VERY hard sell not to do the same now.