After the rather mild reception that Radeon VII received, gamers have been hoping that AMD’s upcoming Navi GPUs will give NVIDIA’s growing market share something to chew on. I am not contesting the fact that the lineup will be competitive, but I’m positive most AMD fans are setting their sights way too high. Here’s why:
The AMD Navi GPUs will most likely be based on the same 4th-gen GCN micro-architecture as Polaris
-o with better clocks and TDPs, thanks to TMSC’s 7nm node efficiency. The fifth iteration of GCN was mainly used in case of the HBM2 Vega and the recently released Radeon VII graphics cards, and most of the improvements pertain to the HBM memory or HBM cache. There’s a very low chance that AMD will opt for HBM memory to fuel the Navi chips, and even if they do it’ll be a ludicrous decision given the price and availability of the memory standard. It is possible though that a modified version of the GCN5 microarchitecture will power the Navi chips.
The Navi lineup won’t go up against NVIDIA’s GeForce RTX graphics cards
The highest-end Polaris card, the Radeon RX 590 struggles to keep up with the slowest Turing GPU, the GTX 1660. The 1660 Ti is pretty much out of reach for AMD’s present Radeon lineup. AMD’s Navi chips will try to recapture the mid-range segment of the GPU market, competing with the GTX 1650 (yet unannounced), 1660, 1660 Ti and potentially the RTX 2060. The big Turing is highly unlikely to find a competitor in Navi. Regardless, aggressive pricing and decent performance figures should put team red back on track, at least in the budget space.
Neither Navi 10 nor Navi 20 GPUs will have hardware level support for raytracing
Just yesterday rumors had started circulating suggesting that Navi 20 will launch in 2020 and will not only challenge NVIDIA’s top-end GeForce RTX 2080 Ti but might also support hardware-level support for real-time raytracing. I would like to clarify that although the former is a possibility, there’s a very slim chance that the Navi 20 GPU will incorporate parts specifically for accelerating raytracing performance. The GCN 5 microarchitecture hasn’t got much room left for major chip-level modifications. That will only be possible with GCN’s successor, the so-called Arcturus microarchitecture.
AMD’s Navi 10 graphics cards will most likely be 20-30% faster than present gen Polaris
People have gone wild speculating the performance of the Navi lineup, however from the few leaked benchmarks and as per my sources in the industry, the GPUs won’t be significantly faster than AMD’s current RX 500 cards. Most of the extra performance will come from the die shrink, higher-clock and the faster (GDDR6) memory standard. Furthermore, the inclusion of certain DirectX12 specific techniques like adaptive shading should also make the Navi chips more appealing.
Anyway, those are my two cents on AMD’s next-gen Navi GPU lineup. Be sure to leave your comments and suggestions below.