Now what really surprises me here is that they could add more bandwidth to deal with more RAM... I always thought they (Nvidia) laser cut stuff like that when deactivating clusters.

Card is faster becouse it has more and faster memory? That cant be 😛

Undying:

Card is faster becouse it has more and faster memory? That cant be 😛

AMD GPU seem to love memory, on my 6700XT you gain more with OC the mem than OC the CPU. On this the gain is also on expended bandwith size, remember the gap with NVidia 730/740 from 64, 128 and 256 version... The big one was a tiny gamer and the small one were slow with office with the same GPU.

Hilbert Hagedoorn:

A group of skilled individuals from Brazil, known for their previous work on enhancing a GeForce RTX 2080 graphics card to have 16 GB of memory, have taken on a new challenge with the AMD Radeon RX 5600 XT graphics card. Radeon 5600 XT Gets 16GB Graphics memory modded and increases in performance

I like things like that, People that try something just because it could be done.

So no laser cut the traces are still there but it is 192 bit as far there is no 4th chip connected to those traces ...to bad this is a difficult task as in is too hard for the average user to solder new memory chips I would have loved to see this kind of market be available everywhere !

Why not a socketable memory? Why not a socketable chip? I always wondered about this, but seems nobody wanted it. Manufacture will definitely get more profit, easier upgrade and last longer hw for consumer. I dream for atx size motherboard that has 2 socketable chip for cpu and gpu with socketable memory for each. Remove the pcie slot for the socket space, put more usb port for peripheral.