Huawei is making its Ascend AI GPU software toolkit open-source to better compete against CUDA

News
By published

Huawei is getting better at making AI GPUs. Now it wants to increase adoption of its technology on the software side

Huawei Ascend AI chip
(Image credit: Huawei)

Huawei has greenlit a plan to open-source its CANN software toolkit for its Ascend AI GPUs. According to the South China Morning Post, Huawei is making this move in order to better compete against Nvidia and its closed-source CUDA software ecosystem.

Open-sourcing CANN, or Compute Architecture for Neural Networks, could further speed up developer innovation and improve the usability of Huawei's Ascend AI GPUs. Furthermore, Huawei has reportedly already held talks with China's leading AI players, business partners, universities, and research institutions on how to build an open-sourced Ascend ecosystem.

Aaron Klotz
Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

7 Comments Comment from the forums
  • -Fran-
    Why is contributing to OpenCL so darn hard? (both rethoric and literal)

    Just make OCL better as it was, and is, the open tool we should agree on improving for this.

    Regards.
    Reply
  • bit_user
    I'm curious to know whether they followed AMD's play of trying to make a CUDA work-alike (i.e. HIP), or if they developed something completely from scratch (a bit more like what Intel tried with oneAPI).
    Reply
  • bit_user
    -Fran- said:
    Just make OCL better as it was, and is, the open tool we should agree on improving for this.
    Intel has been the lone stalwart really pushing to improve OpenCL. Their oneAPI is built on it.

    Battlemage turns out some decent performance numbers, considering its hardware specs. These are some of the first compute benchmarks on it. I wonder how performance has matured.
    https://www.phoronix.com/review/intel-arc-b580-gpu-compute
    Reply
  • thestryker
    -Fran- said:
    Why is contributing to OpenCL so darn hard?
    In this case it may be as simple as political. I certainly agree with you though!
    Reply
  • bit_user
    thestryker said:
    In this case it may be as simple as political. I certainly agree with you though!
    Or maybe it's just a case of: "monkey see, monkey do". People see how successful Nvidia has been and are trying to follow their same playbook.

    IMO, you'll never beat Nvidia at their own game. You need to do an end-run, like Cerebras did.
    Reply
  • zsydeepsky
    -Fran- said:
    Why is contributing to OpenCL so darn hard? (both rethoric and literal)

    Just make OCL better as it was, and is, the open tool we should agree on improving for this.

    Regards.

    OpenCL is a US entity.

    IEEE is also a US entity. In 2019, in response to US sanctions, they banned Huawei, despite Huawei being their biggest contributor to patents and standards.

    So it's reasonable that Huawei will not contribute to anything that's based in the US if they have a choice.
    Reply
  • bit_user
    zsydeepsky said:
    OpenCL is a US entity.
    Yes, I think Khronos is.

    FWIW, if you avoid Khronos, then you also must avoid OpenGL and Vulkan.
    Reply