Tier change: Watson Machine Learning
  • Machine Learning
  • Frankfurt
    London
    Dallas
  • Description
    We would like to inform you that we are removing the P100 compute tier from the Watson Machine Learning service. We will still keep supporting K80 and V100 compute tiers and there will be no impact to K80 and V100 users. Current P100 users are encouraged to move workloads to the newer V100, which is a more powerful and cost-efficient alternative. For more information about V100 and its performance comparison with P100, see the NVIDIA Tesla V100 information.

    As a P100 user, you can also choose a compute tier with multiple K80 as a more affordable and equally powerful solution. However, be aware that K80 has an older architecture and it might be incompatible with frameworks that require a newer architecture. For more information about K80, see the NVIDIA Tesla Product Literature.

    We recommend that P100 users start testing workloads on V100 or K80 as soon as possible in case there are any problems that need to be addressed. If you currently have running jobs on P100, attempt to estimate the finish time. If it might go beyond our End of Support Withdrawal Date, we advise you to evaluate the option to stop early in case it can reduce your cost. For more information on how to change the compute tier and cancel jobs, read our documents for the command-line interface and the Python client.

    Here are some important dates that you need to know:
    • The End of Support Announce date was November 13, 2018.
      As of November 13, 2018, existing running jobs on P100 will continue to be supported until the End of Support Withdrawal date.
    • The End of Support Withdrawal date is December 13, 2018.
      As of December 13, 2018, any training jobs on P100 that are still running will be deleted. Save the results from your training jobs on P100 before the End of Support Withdrawal date.
    This information originated in the Watson Machine Learning is Deprecating Support for Nvidia P100 GPU article within the IBM Cloud platform blog.