Possible gpu arch mismatch: gfx900, please check compiler: -march=<gpu> flag · Issue #49 · ROCm-Developer-Tools/aomp · GitHub
![2019 13″ MacBook Pro [8th,4C,U] + RX Vega 56 Nano @ 32Gbps-TB3 (AORUS Gaming Box) + macOS 10.15.3 & Win10 2004 [itsage] | External GPU Builds 2019 13″ MacBook Pro [8th,4C,U] + RX Vega 56 Nano @ 32Gbps-TB3 (AORUS Gaming Box) + macOS 10.15.3 & Win10 2004 [itsage] | External GPU Builds](https://egpu.io/wp-content/uploads/2020/01/2019-13-in-macbook-pro-gigabyte-rx-vega-nano-gaming-box-luxmark-ball.jpg)
2019 13″ MacBook Pro [8th,4C,U] + RX Vega 56 Nano @ 32Gbps-TB3 (AORUS Gaming Box) + macOS 10.15.3 & Win10 2004 [itsage] | External GPU Builds
ConvBWD triggers GPU memory access fault on GFX803, not on GFX900 · Issue #84 · ROCmSoftwarePlatform/MIOpen · GitHub
遠坂小町 on Twitter: "[KOMACHI] AMD GPU Table for Vega & Navi Version 2.2A (8/5/2019) https://t.co/ietzMEC2hR" / Twitter
![blue nugroho on Twitter: "@markus_augsten no it is what it is, GFX100x = PS5 = Navi10lite no functional differences with GFX101x = Navi10 gpu" / Twitter blue nugroho on Twitter: "@markus_augsten no it is what it is, GFX100x = PS5 = Navi10lite no functional differences with GFX101x = Navi10 gpu" / Twitter](https://pbs.twimg.com/media/EMY62eyUEAAPHF6.jpg)
blue nugroho on Twitter: "@markus_augsten no it is what it is, GFX100x = PS5 = Navi10lite no functional differences with GFX101x = Navi10 gpu" / Twitter
![遠坂小町 on Twitter: "[CompuBench] Compute Performance of AMD 7340:C1 https://t.co/Z1Z513X4aP GFX ID = GFX 1012 (Navi 14). CL_DEVICE_MAX_COMPUTE_UNITS : 12. (24). *(CompuBench displays half of the actual CU number (RDNA GPU Only).)" / Twitter 遠坂小町 on Twitter: "[CompuBench] Compute Performance of AMD 7340:C1 https://t.co/Z1Z513X4aP GFX ID = GFX 1012 (Navi 14). CL_DEVICE_MAX_COMPUTE_UNITS : 12. (24). *(CompuBench displays half of the actual CU number (RDNA GPU Only).)" / Twitter](https://pbs.twimg.com/media/D_Sq5-1UcAAoBNF.png)