"Games are still not multithreaded properly then."
Not true.
Dx12 supports multi-core gaming natively. ANY 3d game Engine coded with DX12 or Vulkan will run as a multi-core multi-threaded game. Asynchronous Compute allows the CPU to send data non-serially or Asynchronously to the dGPU or GPU cores in the APU. This why the consoles all run 8 core Jaguar APU's. And YES ACE is managed by the GPU but it impacts CPU performance.
DX11 ONLY allows a serial data stream to the shader pipelines and doesnot come close to matching the performance of DX12. Tom;s Hardware tested DX11 multi-core gaming and determined the was virtually no difference between 4 or 6 -10 cores and a large jump between 2 to 4. The O/S overhead impacts 2 core and this is alleviated with 4 cores.
Will Gamespot rerun these tests when Vega is released?
What the benchmarks have proven is GAMESPOT is biased to favor Intel.
Using NVidia dGPU disables Asynchronous Compute. NVDA does not support it and ACE and ASP are major AMD CPU force multipliers. Instead RYZEN was judged by a biased report.
Star Swarm should be rerun using RX 480, better would be two as RX 480 scales easily with DX12.
This site also IGNORED or OMITTED 3dMark API Drawcall Feature Overhead test. This test shows AMD with a huge advantage. Obviousy Gamespot does not want to report ANY results that show Intel unfavorably
Using ANY NVidia dGPU card CRIPPLES Ryzen as the DX12 benchmarks are run with Asynchronous Compute DISABLED. You might try running Star Swarm with at least ONE RX 480. Try 2 as DX12 supports Explicit Multi-adaptor in other words multiple GPU's scale the performance. This is NOT Crossfire.
You also OMITTED 3dMark API Drawcall Feature Overhead test. This shows AMD/RAdeon far superior to Intel/GeForce see the ANAND TECH report back in early 2015. AMD hardware is optimized for DX12.
You also OMITTED ChessBase Fritzmark which shows RYZEN CRUSHING ALL Intel 8 core 16 Thread CPU's
Chessbase FritzMark is probably the best system stress test available.
CPU-z also shows Ryzen as the clear single core winner.
http://valid.x86.fr/bench/rjmzdu/1
OF COURSE IF YOU INTENDED TO SHOW RYZEN IN A BAD LIGHT AND INTENTIONALLY DESIGNED THE BENCHMARKS TO SHOW INTEL FAVORABLY THEN YOU WILL NOT REDO THE DX12 TESTS WITH RX480.
akamateau's comments