[MWC] NVIDIA has showed us something remarkable: their Tegra roadmap from 2011 to 2014. If you look at the Y-axis of the graph above, you’ll see 100 – that is the performance scale relative to Tegra 2, which is the fastest dual-core SoC in volume production right now.
To put “100X Faster” into context, we’re talking about the chips computational power, not about a 100X boost in frames per second or web browsing speed. Although this type of progress is very rare (or possibly unseen before – I’m not sure), it is believable in a dynamic area like mobile computing. Also, you have to remember that the overall jump in performance includes the graphics processor (GPU) which tends to evolve much quicker than general-purpose processors (CPUs). At some point the GPU will probably become a massively parallel computing device, just like it did on PC.
In short, the takeaway is this: NVIDIA is going to crank its execution machine and release new chips on a 12-months schedule. By looking at the graph, it seems that every couple of years, you have a major architectural change that induces a 5X performance boost (doubling the number of core+better GPU?), followed by a 2X boost that might be the result of an evolution of the previous generation. If that logic holds, would that mean that “Logan” will have 8-cores? I’ll ask, but I don’t think that NVIDIA will comment.
The rest of the industry has been warned – now let’s watch the race.
Filed in ..