Their first GDDR7 module packs a massive 16 gigabits, with each module holding 2 gigs. It's clocked at 32 gigabits per second in PAM3, but can drop to 28 to improve yield, overall performance, and cost-effectiveness in the early stages.
Samsung says these babies are 20% more power-efficient, too, running at a voltage of just 1.1 volts instead of the standard 1.2. They've also got a new packaging material and some slick circuit design that reduces heat and lowers GDDR7's thermal resistance by 70% compared to GDDR6.
Now, get this: rumors are flying that the flagship card in the RTX 50 series, probably called the RTX 5090, will be the first to flaunt this GDDR7 goodness. And guess what? They're saying it could almost double the performance!
Even with a 256-bit bus, GDDR7 is hitting 1.18 terabytes per second at 37 gigabits per second. That's way faster than the 384-bit 24-gigabit per second GDDR6. It's like a turbo boost for the graphics firepower of our games.
