Bit of an interesting video as we get closer to the brick wall where monolith is no longer viable. The largest die size that is possible with GPUs is around 800 mm2. This has been possible for Nvidia in the datacenter as node shrinks progressed steadily. However as node shrinks slow down drastically, both AMD and Intel going to chiplet designs for their datacenter GPUs, it's only a matter of time when Nvidia will be forced to go to chiplets themselves. In gaming, they can certainly still stick to monolith design for least a few more generations as they can always increase their die size to 800 mm2 from 600mm2 but that would also increase the cost of the gaming GPUs. They have been doing a lot of research into chiplet design shown by their own documents so it will be interesting to see when they start going into it themselves.
AMD’s Latest Radeon Vulkan Driver Adds Support For Phoenix APUs, Better Gaming Performance
https://wccftech.com/amd-radeon-vulkan-driver-adds-phoenix-apu-support-better-gaming-performance/
Intel Core i5-14600KF is 5.5% faster than i5-13600K in first Geekbench single-core test
Makes sense as it's a refresh
ZOTAC Extends Warranty Period To Five Years For NVIDIA GeForce RTX GPUs In Taiwan
https://wccftech.com/zotac-extends-warranty-period-five-years-for-nvidia-geforce-rtx-gpus-in-taiwan/
Steam Deck modded to support 32GB of LPDDR5 memory
https://videocardz.com/newz/steam-deck-modded-to-support-32gb-of-lpddr5-memory
I wonder when Valve will refresh the Deck. Hopefully they learned a lot from their first take.
PC Specs: CPU: 7800X3D || GPU: Strix 4090 || RAM: 32GB DDR5 6000 || Main SSD: WD 2TB SN850
