Although between 4090 and 5080 we rightly assume that the 4090 will be more powerful, it is my opinion (and everything has been calculated by NV in this sense for several generations already) that it will only be by a few percent and why ? Because :
1 Certainly the 4090 has around 16,000 shaders when the 5080 will have between 10 and 11,000
2 The 4090 has a 384bit data bus and the 5080 256bit
3 memory type GDDR6X vs GGDR7
4 IPC optimizations of the new archi
5 the TSMC node a little thinner
The reduction in shaders and buses will be compensated by the gddr7 whose bandwidth increases by at least 30% at equivalent bus/frequency compared to the GDDR6X of the 4090, as well as the archi with a better IPC.
The announced consumption is 350W compared to 450W for the 4090.
“
Memory type GDDR6X GDDR7
Maximum memory capacity 24 Gb 16 Gb
Memory bus width 384 Bit 256 Bit
Memory frequency 1313 MHz 2400 MHz
Memory bandwidth 1.01 Tb/s 973.8 Gb/s
“
The price is expected between 1300 and 1500 dollars
This is only my opinion, but there will be a visible generational leap, certainly not as much as between series 3 and 4 (3060 vs 4060, bus 30% shorter and yet much better performance and consumption), but still .
