1. Home
  2. Rack Servers
  3. IBM Power servers Reviews
  4. User Review of IBM Power servers
Best infrastructure for AI workload
May 01, 2019

Best infrastructure for AI workload

Anonymous | TrustRadius Reviewer
Score 10 out of 10
Vetted Review
Reseller

Modules Used

  • Power System AC922

Overall Satisfaction with IBM Power

We use IBM AC922 servers as a GPU based infrastructure for a deep learning use case. We leverage it to address use cases across industries and it's leveraged across the USA.
  • Great on-premise infrastructure for AI.
  • Only GPU infrastructure which uses PCIe Gen 4 and has a Nvlink between GPU and CPU.
  • Comes bundled with WML Base which could be used by data scientists.
  • Product documentation needs to be kept at pace with changes with the various libraries that need to be installed.
  • No other changes are required in AC922 and we are happy with the infrastructure.
  • Great infrastructure for Enterprise DL use cases.
  • The ROI depends on the use case, but with the top-notch performance it provides, customers will realize the immense value in a period of just 2 years.
IBM has been a leader in HPC and we did initial work and found the results encouraging.
If there is a huge data repository that is both structured and unstructured, IBM AC922 is the best fit when compared to its competitors. It's like comparing a race between an F1 car and a NASCAR. If the models built are not large (LMS) and DL is not required then IBM AC922 would be overkill.