Micro cracking on BGA will be a short term issue with this kind of temperature behavior.
Small talk from Taiwan
Engineers who worked on RX3080 and RX3090 design told on local closed group chat that thermal design for cards that have FE boards and even less dense custom ones are indeed horrible. And despite that thanks to special throttling and local power management of the main chip they are able to keep its temperature more or less ok - on the board itself it can be areas of too high temperature, especially near VRM and RAM same as in referenced thermal images above.
BGA and memory issues are very probable.
Next thing - Nvidia actually plan shift main focus and promotion on cards with 16GB and more memory. Including 20Gb RX3080 and 24GB RX 3090. All cards with RAM below 16GB will be called "low end and obsolete models" already in 2021. RX3xxx update will focus to bring all memory sizes to 16GB except 1050/1650 level cards. GT1030 replacement also coming with around 3x performance.
Main issue is super thermally dense boards due to decision to shorten PCB
NVIDIA PG133 Founders Edition board:
Long term problems are very probable.
Issue with high memory temps
In reality 3080 and 3090 require proper water cooling systems with at least 360mm radiators, but we still have all sites claiming how it is nice and ok.
CUDA Performance
Board design
NVIDIA PG132 reference board for 3080:
NVIDIA PG132 reference board 3090:
Next time then press will tell you something about very costly top board design and expense on extra memory bus - you can spit them into their prostitute face, as board is literally the same, just less power phases.
NVIDIA PG133 Founders Edition board:
Colorful GPUs
Note how it is first time that better process did not help much.
One guy in industry I talked with told that Ampere GPUs will be very prone to BGA issues and board burnouts. Life of such cards if it is not used with water cooling can be limited to 2 years (if used heavily).
As expected - design of the cooling is not best for 3080 FE.
Quadro card based on Ampere will be having less disabled CUDA cores
Actual number of CUDA cores in GA102 is around 5500, but some can be disabled.
Good youtube video must be like this one
More leaked benchmarks
3090 ti or a titan based on 3090 should be grate to see its capabilities
RTX 3080 Benchmarks
In average RTX 3080 is 33% faster than RTX 2080 Ti and 54% faster than RTX 2080 Super.
It looks like you're new here. If you want to get involved, click one of these buttons!