We’ve been hearing a lot about AMD’s upcoming Navi GPUs. These 7nm based video cards are going to be featured in both Microsoft and Sony’s next-gen consoles. Furthermore, it’ll be the first new architecture in AMD’s aging mid-range GPU lineup in quite a while. Team red has been stuck on Polaris for almost half a decade. Now, an unnamed AMD Radeon GPU has been spotted over at CompuBench which might turn out to be the elusive Navi chip.
The leaked benchmark compares the supposed Navi GPU to the Radeon RX 580. However, it’s worth noting that the latter is run using OpenGL while the former leverages the much faster DirectX API. As such, it’d be unfair to directly pit the two, but just for the sake of it, Navi appears to be 30-40% faster than the RX 580. The DX scores of the 580 should be notably higher and should close the gap down to around 20-30%. This isn’t the interesting part, however.
A bunch of compute-intensive benchmarks
This purported Navi chip appears to have 20 CUs which translates into 1280 shaders. Comparatively, the Radeon VII, although a high-end part has 3,840 shaders with the remaining 4 CUs disabled. This puts the shader performance of the Polaris successor at roughly a third of the recently launched 7nm Vega GPU. Looking at the performance figures above, it is highly likely that AMD will be looking towards GDDR6 memory to make up for the limited CUs. HBM2 is not an option given the shortage and its high price.
To sum up, if (a big if) this is the Navi GPU at all, then it seems to be optimized for sheer graphics performance, rather than compute-based calculations unlike NVIDIA’s chips. Given that team green’s GTX 1660 Ti is priced similar to ($179 tbe) but pretty much blows the RX 580 out of the water, AMD will have to launch the Navi GPUs at rather tempting prices. We’ll keep an ear out for any more Navi related rumors. Cheers!