The gaming benchmarks of Intel’s Arc Alchemist flagship GPU have surfaced (courtesy of LTT). The A770 features 32 Xe-Cores packed with 4,096 ALUs and 32 ray-tracing units with a peak core clock of 2.4GHz. Paired with 16GB of GDDR6 memory across a 256-bit bus, it delivers 20 TFLOPs of single-precision performance. Using an 8+6 pin power connector, it draws somewhere between 200-300W of power.
Intel promises performance on par with the RTX 3060/3060 Ti but considering the current state of the graphics drivers, it’s highly unlikely. In his video, Linus demos the gaming performance of the A770 in Shadow of the Tomb Raider using both the DX11 and DX12 APIs. In the former, the top-end Arc GPU scores a disappointing 40 FPS only to redeem itself in the latter with a satisfying 80 FPS on average.
This is almost identical to the RTX 3060 which offers 78-85 FPS on average in the same title at 1440p. Not very impressive considering that this will be the fastest Intel GPU for the foreseeable future but if priced well, it ought to serve the masses admirably.
Apparently, Intel plans to divide popular titles into tiers based on optimization and performance. The top tier will include games with the “best-in-class performance”, followed by the second tier which will consist of moderately optimized titles. Most of the games in these segments will be using the newer DX12 or Vulkan API wherein much of the optimization falls to the game/engine developers. Finally, the 3rd Tier will be chock-full of unoptimized games running older APIs like DX11 and OpenGL.