By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Pemalite said:

I will quote my other post:

"So doing some digging. Looks like RDNA is a "Hybrid" design that bridges GCN to RDNA... Just like VLIW4, but more extreme.
Lots of changes under the hood, but it's still very easy for developers to leverage it if they have worked with GCN for a long time.

The doubling of the ROPS is a big boost... And one bottleneck of GCN gone. - Reducing wavefront size and increasing SIMD size so they match is another big Pro' as well and another GCN bottleneck gone.
Uplift in bandwidth and chip-wide delta colour compression and much improved caches resolves another GCN bottleneck.

It's going to result in a substantial uptick in performance per teraflop either way... I would not be surprised to see a 50% performance uplift over the RX 590."

Things like Primitive Shaders are now working as well, finally.

As for Vega... Vega was actually extremely efficient, at lower clocks and voltages, AMD though wasn't able to compete though and decided to throw out efficiency in order to garner as much performance as possible and drive the clockrates up.

Just remember though, Navi is Polaris's replacement, not Vega. Vega 7 will still be AMD's high-end part going forward.

And if we can... Try and stick to better sources than Polygon, like Anandtech.

RDNA has a SIMD32 and a SIMD64 mode which corresponds with their dual scalar units and is somewhat similar in concept to what Intel has been doing a long time with variable SIMD widths on their GPUs ... (Intel has SIMD32/SIMD16/SIMD8/SIMD4x2 modes)

As for primitive shaders, I'd be interested in seeing an API exposed for it to see what it's programming model is actually like as an interesting comparison to Turing's mesh shaders. "Primitive shaders" are still a massive mystery since we still don't if it's truly equivalent to Turing's mesh shaders or something completely different ...