16 and 24GB GTDR6X and three-fan FEs

To start the last day of the week, what’s better than new information about future graphics cards NVIDIA GeForce RTX 4080 and RTX 4090 ? This is by our German colleague Igor Labs We learn more about PCB and memory and power supply configuration. In fact, we find a map of the PCB with the AD102 GPU Nvidia.

NVIDIA RTX 4000: Founder version cards with three fans and the largest custom cards

From what we know to the latest information / rumors, The RTX 4090 will have a maximum consumption of 600W. The 24 voltage converters given to us by the uP9512 controller are a group of three. Memory is connected to four VRM phases.

In terms of memory, we see 12 locations for memory chips on the map. With 1 or 2GB chips, multiple configurations are possible. We have to 24GB GDDR6X On top of that RTX-4090 On a bus from 384 bit. With all A large L2 cache is 96 MB.
For that RTX4080 And his Is hidden Lie Between 80 and 88 MBWe will be 16 GB Memory too GDDR6X On A 320-bit bus. The next high-end card issued by NVIDIA does not have GDDR7 memory. And talk about RTX4070We have 16 GB of memory, but only on GDDR6 and on the 256-bit bus.

As we have already heard, the cooling systems for the RTX 3090 Ti are the first attempt to get ready for Ada Lovelace very soon (March 29). In fact, The RTX 3090 Ti There will be a DGP of 450W, so brands are starting to produce cooling accordingly, and it is unlikely to have similar cooling systems between the RTX 3090 Ti and the future. RTX4000. Also, on the Nvidia page, we will see Founder version cards using three fans Cool the components efficiently. 3.5 slots and custom cards over 2 kg are also quite possible.