How does RDNA architecture improve GPU performance?

30 Haz 20:31

RDNA architecture is a huge step up from the older GCN design, and it’s not just about raw compute power. AMD pushed for more efficient use of resources here—think smaller execution units packed tighter with better scheduler logic. What I love most is how RDNA 2 (like in my Steam Deck) handles memory. Infinity Cache reduces bandwidth bottlenecks, so even with lower memory speeds, you get smoother performance in games like *Hades* or *Stray* compared to GCN-based GPUs. Efficiency matters a lot too. RDNA runs at lower power draw while delivering better frames per watt, which is why consoles and handhelds swear by it now. My Deck practically sips power for the performance it gives, whereas an older AMD GPU? You’d be battling thermal throttling or having to underclock. It’s not just “more cores = better”—architecture tweaks like wave-level parallelism and improved texture compression make a real difference.

30 Haz 22:24

Hmm, RDNA'nın mimarisiyle ilgili en çok merak ettiğim şey aslında bu "Compute Units" denen şeyin nasıl bu kadar verimli çalıştığı ya? Eski GCN mimarisine göre kaç kat hızlanma sağlıyorlar acaba, bunda yükü nasıl azaltmışlar ki bu performans patlamasını yaşatıyorlar?