View Post

So, Navi is between 55 - 85% more power efficient in gaming than the Polaris and Vega 14 nm cards, and more or less matches the Turing cards. That's about the same position AMD was back in 2016, with Polaris compared to their 28 nm GCN cards and Nvidia's Maxwell ones.

The silver lining is, unless the RDNA architecture offered no IPC gains whatsoever, AMD's difference to Nvidia on 7nm should diminish compared to where it was with 14 - 16 nm (Pascal was a whopping 60 - 65% more power efficient than Polaris at launch) because the newer node doesn't seem to decrease power consumption as much as the previous shrink did. That is, assuming Nvidia's first 7nm cards are mostly a shrink of their Turing architecture, which seems probable, given Nvidia, like Intel, doesn't seem to release new architectures with node shrinks.

On the other hand - impressed with the Zen 2 chips. It's the first time in many years AMD competes and even surpasses Intel, matching core to core, clock to clock. However, Intel still seems to have a distinct gaming advantage. I would say the reason being developers favor Intel tools and instructions more than it is the fault of single-core performance at this point.

Last edited by haxxiy - on 07 July 2019