To make it easier for you guys, try to get inside console designers heads.
- Put hard limits on paper.
- Put manufacturing nodes on paper.
- Put die size and energy requirements as well.
- Put AMD current offering on paper.
- Put console price on paper.
After that try to design console.
Think about it, designers work with hard limits and they are using following :
- 7nm node (at most 7nm+ with slightly higher density and 15% higher clocks for same wattage).
- AMD APU
- Pricing console at $499 max
- Navi/Zen2 combo (RT included), I/Os, gobs of GDDR6 RAM, wide memory controler, SSD, BD drive - all within 200W
So, now look at what AMD is offering and what they have to include there.
- 8 core Zen2 at 3.2GHz is around 30W
- 40CU Navi at 1.8GHZ is 200W (without RT, 220W with 8GB of memory)
- 16-20GB of GDDR6 ~ 30W
- I/O, SSD, disc drive ~ 15W
How do you fit all of this inside 320-350mm² chip and 180-200W of power consumption?
Fact is - YOU CANT. Now increase 40CU part to 56CUs and add RT hardware.
This is why 12TF without new node is bogus. Its right infront of your eyes. They are not talking about TF. They are talking about SSD, new CPUs, physics, 60fps etc. but not TFs.
Why? Because we are stuck. We where at 10TF 4 years ago. So where is 30TF GPU? Why arent we advancing faster? Why does new Nvidia flagship after 2 years deliver 10% increase in FLOPs, but 40-50% faster performance in games?
Face it, tech is getting mature. Node advencment is slower and slower, but also more expensive. So what we are getting is what its confined inside hard limits, now what people expect and want.