The " new " MacBook Pro 15 " equipped with the Radeon Vega came to the editor today, an interim update that was not expected soon and that makes a bit of a cringe of those who have spent a fortune in these models this summer .
Outside the GPU, no change on these laptops, that offer configurations that are strictly identical to previous versions. Good, because it allows us to compare graphical performance with much more ease!
Under the hood, we find two new graphics cards:
– the Radeon Pro Vega 16, with a clock frequency of 740Mhz and 16 calculation units
– the Radeon Pro Vega 20, with a clock frequency of 740Mhz and 16 calculation units
As a compendium, the high-end of this summer (the Radeon Pro 560X) offers lower frequencies (300Mhz) and 16 calculation units :
In addition to the difference in architecture and frequency, AMD is not very local on its site, just know that these GPUs are etched in 14nm (like the 560X) FinFET and embark HBM2 memory (2nd generation). Apple conjures up " 60% extra power " without, however, specifying the points of comparison and the associated tests.
While waiting for our application benchmarks to come (it takes time), here are our first tests in GeekBench (Metal), CineBench (OpenGL) and Luxmark (OpenCL). In OpenGL, precisely, Cinebench seems to indicate quite similar scores between the 560X and the Vega 16 (which have the same number of units), but the engine used here is aging and does not incorporate the latest technology from Apple (Metal):
For its part, GeekBench is a little more relevant under Metal, exactly. The 4 GPUs are pretty well staged, and the gain is about 10% between the 560X and the Vega 16 and 20% with the Vega 20.
Finally, Luxmark joins Cinebench, with a 560X that pays the luxury of exceeding the Vega 16, the gain between the Vega 20 and the 560X not being here " than " 15%.
Obviously, these benchs remain very theoretical and we will soon test these GPUs under real conditions, taking again the many tests done this summer on previous models. At first glance, the price difference requested (we still talk about 300 to 420 € additional) we seem quite consistent in terms of actual performance observed (we are far from 60%), but for some treatments, any extra graphics power is always good to take!