AMD Ryzen 5 5300G Engineering Sample Sold on Ebay and Benchmarked

Until you need to use Direct Storage.

 

That's a myth.

 

RDNA2 is bit more bandwidth efficient than last of GCN (Vega). By itself it would not be big improvement.
If there was die size investment to IC, I am sure, graphical performance of APUs could even double.

I would not be surprised if 8 RDNA2 CUs run circles around 10/11 GCN CUs. But problem is different. Transistor cost of 10 Vega CUs vs 8 RDNA2 CUs + IC.
My guess is that IC investment would start to show worth when APUs use 16CUs+.

 

Current: Storage (compressed) => RAM (compressed) => CPU (decompression) => RAM (Inflated) => PCIe transfer of inflated data towards GPU/VRAM

Direct Storage: Storage (compressed) => PCIe transfer of compressed data towards GPU/VRAM => VRAM (compressed) => GPU (decompression) => VRAM (inflated)

Guess what? Direct Storage saves a lot of PCIe bandwidth when it gets to provisioning of new data.

 

So... seeing that nobody commented about this again, I'm guessing that nobody really cares about the AV1 support?

 

It might still have it, the video decoder engine isn't really linked to the architecture, they can update it independently, as seen with ravenridge and renoir which both have vega igpus, yet feature VCN hardware like navi.

 

Hope you are right!