Remember Vega Mobile? First announced at AMD’s CES Tech Day, the company promised that it would bring a new version of its desktop Vega chip to market, to compete in the mobile space. With an ultra-low z-height of 1.7mm (the same as Polaris) and a single HBM2 stack, the mobile chip was a potential breath of fresh air in a market that’s increasingly dominated by Nvidia.
And then… nothing.
Vega Mobile dropped off the board. Except now it’s back, as an upgrade option on the MacBook Pro (yes, the same MacBook Pro that Apple just upgraded last summer). It’s not exactly clear which systems will get the update — Apple’s landing page for the MacBook Pro refers to all 15-inch systems being able to add the GPU, but only the top-end $ 2,800 base-price edition actually lists Vega as an upcoming addition for the platform.
Anandtech has put together a chart showing how the new Vega 16 and Vega 20 might compare with the already-existent Pro 560X (equivalent to an RX 560):
There are some huge improvements here. Core counts are up by 1.25x for the maximum configuration, with 2.35x as much memory bandwidth for Vega 20 / Vega 16 compared with the Pro 560X. ROPs could also be doubled compared with that GPU (32 ROPS would make sense for a Vega 20, compared with just 16 for Pro 560X).
Mathematically, this upgrade should deliver significant performance improvements compared with the Pro 560X. Doubling ROP counts (which seems the minimal outcome) potentially improving clocks, and boosting core counts will definitely increase performance. The problem here, of course, is that Apple’s 15-inch MacBook Pros are objectively terrible deals for any gamer and it’s not clear how much these improved capabilities will translate into improved performance in various rendering or GPU-accelerated apps. It’s harder to predict those improvements overall because they tend to be application and workload specific.
It will be interesting to see just how power-efficient Vega is in this mode, however. AMD’s GPUs have developed a reputation for being markedly less power efficient than their Nvidia counterparts in part because AMD has had to push its cards far out of their ideal clock ranges in order to compete against Team Green. As the Radeon Nano demonstrated, it’s often possible to draw significantly less power for only modestly lower performance (the Radeon Nano drew 85 percent as much power as the Fury X in 4K while offering 91 percent of the Fury X’s performance). A well-optimized Vega could be a power-sipping part much more on par with what laptop users expect, despite the desktop SKU’s reputation for being power-hungry.