For $500-$600 (maybe even $700) next gen?
They will give us just 30-40TF (tops), and you're going to like it.
And yes, in the right hands, talented developers make these consoles shine.
Maybe more like 60 - 70 TF but yeah it's ultimately always going to come down to the talent.
I expect the PS6 to be around 50 tflops, and the leap in graphics to be smaller than between any other generation.
Maybe closer to 60 - 70 TF but I think that will come from a mix of GPU and HBM-PIM (Processing-In-Memory) technology, not just GPU.
Then again I'm adamant on chiplets, Big/Little CPU clusters (with lots of "little" CPU cores closer to memory for data locality), PIM memory and PNM technologies, full fat active interposers with wholly integrated data and decompression systems built-in, and HBM memory being the standard for 10th-gen in one way or another.
Problem is we need new types of gameplay and mechanics with all this extra horsepower. I'm tired of what is essentially the same games since the PS2 era, only with better graphics.
And for that we need huge leaps in CPU technology, plus maybe GPU compute technology.
Nah, we need full-body VR. Also AR.
Why not? The upcoming flagship card from AMD is going to just about hit 100 TFLOP and the PS6 is 5/6 years and 2 process shrinks away.
The Fury X was AMDs flagship card and launched just over 5 years before the PS5 launched and was 8.6 TFLOP. 5 years and two process shrinks later and it was handily surpassed by the consoles.
Trends of the past can't 100% be used for predicting what the future will bring. Smaller nodes are a lot more expensive. RAM prices haven't really been dropping like in the past. Power consumption targets for consumer electronic devices are more of a thing than in the past.
If the PS6 and NextBox can offer 2x raw TF compute of a 3090 then that would be fantastic and is probably around the realistic top end. But if they have big improvements in upscaling tech, ML tech, data locality management, data decompression technologies, CPU throughput and scalability, RAM capacity, etc., then being below whatever top-end Nvidia or AMD GPUs are available by the end of this year won't really matter.