4 ms·

Those are some astounding numbers if the average power draw has not significantly changed. I'm looking forward to the 4070 or 4070 Ti.

by TillE 4y ago

gtirloni 4y ago
Agreed, the xx70 models have consistently been a great price/performance option.
ZeroCool2u 4y ago
NVIDIA has effectively changed the naming scheme for their cards this year. Now there are two 4080 cards, with wildly different specifications. Basically, one of them is actually a 4080 and the other is what they used to call a 4070. This is reflected in their name and price here[1]. [1]: https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080/ https://www.nvidia.com/en-us/geforce/graphics-cards/40-serie...
- schnitzelstoat 4y ago
  > Now there are two 4080 cards, with wildly different specifications. This is going to be awful when buying pre-built machines. They'll just say it has a 4080 and stick the crappy one in.
paulmd 4y ago
People let a bunch of the AMD twitter bros work them into a froth over power efficiency. Everyone knew it was basically a 2-node leap (Samsung 8nm is a 10+ node, it's about on par with TSMC 10nm, so there's 2 generations of improvement there), the bullshit from the twitter gang was utterly unfounded since day 1. Everyone who wasn't in the tank was telling you god's own truth: NVIDIA is going to leap two nodes, it was always gonna be a huge leap, bigger than Pascal. They were drowned out by the AMD hypetrain crowd as usual who desperately wanted a flop and were trying to will it into existence. It does go to show just how far NVIDIA is ahead on architecture... they could fight RDNA2 to a standstill with basically a whole node of disadvantage, and they did the same thing with Turing vs RDNA1 where they were using TSMC 12FFN (which is literally 16FF with the "your mom-sized" reticle - there is no optical shrink in 12FFN) and managed to fight 7nm cards to a draw on both performance and efficiency.

Those are some astounding numbers if the average power draw has not significantly changed. I'm looking forward to the 4070 or 4070 Ti.

by TillE 4y ago

Those are some astounding numbers if the average power draw has not significantly changed. I'm looking forward to the 4070 or 4070 Ti.

gtirloni 4y ago

Agreed, the xx70 models have consistently been a great price/performance option.

ZeroCool2u 4y ago

NVIDIA has effectively changed the naming scheme for their cards this year. Now there are two 4080 cards, with wildly different specifications. Basically, one of them is actually a 4080 and the other is what they used to call a 4070. This is reflected in their name and price here[1]. [1]: https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080/ https://www.nvidia.com/en-us/geforce/graphics-cards/40-serie...

schnitzelstoat 4y ago
> Now there are two 4080 cards, with wildly different specifications. This is going to be awful when buying pre-built machines. They'll just say it has a 4080 and stick the crappy one in.

paulmd 4y ago

People let a bunch of the AMD twitter bros work them into a froth over power efficiency. Everyone knew it was basically a 2-node leap (Samsung 8nm is a 10+ node, it's about on par with TSMC 10nm, so there's 2 generations of improvement there), the bullshit from the twitter gang was utterly unfounded since day 1. Everyone who wasn't in the tank was telling you god's own truth: NVIDIA is going to leap two nodes, it was always gonna be a huge leap, bigger than Pascal. They were drowned out by the AMD hypetrain crowd as usual who desperately wanted a flop and were trying to will it into existence. It does go to show just how far NVIDIA is ahead on architecture... they could fight RDNA2 to a standstill with basically a whole node of disadvantage, and they did the same thing with Turing vs RDNA1 where they were using TSMC 12FFN (which is literally 16FF with the "your mom-sized" reticle - there is no optical shrink in 12FFN) and managed to fight 7nm cards to a draw on both performance and efficiency.