A juicy new rumor on NVIDIA's next-gen Ampere GPU (which I revealed back in April 2018) has arrived to kick start the New Year for Team Green, with the Taipei Times reporting that NVIDIA's new Ampere GPU architecture could be up to 50% faster, at half the power consumption of Turing.
NVIDIA will want to fight back big time against AMD's upcoming 'Big Navi' that will come in the form of the currently rumored Navi 21, with NVIDIA hitting the new 7nm node just like AMD is currently on with its Navi-based GPU and the Radeon RX 5500 and RX 5700 series graphics cards. But this news of a 50% hike in performance at the same power consumption is big on its own.
Taipei Times is reporting from a client note by Yuanta Securities Investment Consulting Co that NVIDIA's new Ampere GPU would have a 50% increase in graphics performance while "halving power consumption". NVIDIA has been king of efficiency for a while now and that is on the 16nm, 14nm, and 12nm nodes -- but it looks like it has struck lightning on 7nm if these rumors are true. We could well see the '50% faster performance' in the ray tracing (RT) side of things, and not a general 50% performance improvement.
The report continues that the new Ampere-based graphics cards would see the gaming sector enjoy the benefits of NVIDIA's latest adventures in GPU dominance, with the likes of MSI, GIGABYTE, ASUS, and others seeing revenues boosted through new graphics cards, gaming notebooks, and gaming PCs.
Sales haven't been great on the current Turing-based offerings, but things are expected to change with Ampere. If the rumor of up to 50% more performance at half the power consumption are true... I think we'll be in for some of the most exciting years in GPU history. AMD is launching Big Navi this year, NVIDIA is launching its next-gen Ampere beast, and Intel is re-entering the discrete graphics card game.
To say I'm excited would be an understatement.
I do wonder what NVIDIA could do with the additional GPU die space if the RT side of the card is much more powerful, we would have a gigantic (even more) monolithic GPU that would destroy not just itself (in full RT form), but anything AMD is capable of making right now outside of moving towards the Zen-style chiplet route.