A new report from Moore’s Law is Dead claims that NVIDIA could be prepping to launch a new TITAN RTX halo product. This would be based on a maxed-out 4nm AD102 silicon. For those who don’t know NVIDIA only enabled 128 of the 144 SMs on the AD102 silicon for the RTX 4090. They also only used 21 Gbps memory that they had on the previous generation. This gives NVIDIA lots of headroom not only for an TITAN RTX SKU, but also an RTX 4090 Ti as well.
So with a completely maxed-out AD102 you would get all 144 SMs, giving you a total of 18.432 CUDA cores. NVIDIA could also use faster 24 Gbps memory, which would give the card a 1152 GB/s memory bandwidth as well as a stock power limit closer to the 600W design limit of the 12VHPWR connector.
Moore’s Law is Dead also posted what they claim to be the first real-world pictures of this card. This card has a massive 4-slot footprint and uses the same dual-axial flow-through design as the RTX 4090 Founders Edition. Being a thicker card, it should be able to handle higher thermal loads. The last RTX TITAN product from NVIDIA sold in the $2500 range, do you think this card will be more expensive than that?
Via Moore’s Law is Dead (YouTube)
