DDN unwraps AI400X3, Infinia updates in Hamburg before AI world turns to Paris

CCH - the Congress Center Hamburg pic: shutterstock / editorial use only

DDN chose the International Supercomputing Conference this week to officially launch its latest appliance, the AI400X3, together with enhancements to its AI software offering Infinia 2.1.

The firm said these will form the core of the DDN Data Intelligence Platform strategy it unveiled at Supercomputing 2024 last year.

The ISC announcement came less than a month after it trumpeted closer collaboration with Nvidia. Every storage firm is battling to be the GPU giant’s data partner of choice at the moment, given the vast amounts of data AI needs – and the vast amounts of cash customers are pouring into AI.

Needless to say, the AI400X3 promises “seamless integration” with Nvidia’s DGX and GB200 platforms, as well as its BlueField DPUs and Spectrum-X network platform.

It should deliver 70 percent higher write throughput and 55 per cent higher read throughput compared to the previous generation. (When it first mentioned the platform at Supercomputing DDN claimed a 60 percent performance boost overall.)

For your reference

DDN presenters told the IT Press Tour that the new platform also promises enterprise grade multitenancy, with per-tenant isolation, VLANS, quotes and access controls. And at the storage level, proactive drive failure management.

Critically, given the tech world’s Nvidia obsession, “Nvidia-approved reference architectures featuring the AI400X3” are already available, including for Cloud Partners running H100, H200 and B200 systems, and GB200 systems.

Infinia 2.1 delivers expanded support for “complex, high-throughput data pipelines”. This includes native integrations with Datadog, Chronosphere and OpenTelemetry.

A Hadoop Connector has been launched in preview, to support “high-performance Hadoop and Spark workloads using native S3-compatible access.”

Together, the two should accelerate time to value for both model training and inference as well as complex simulations.

At the same time, the vendor is promising lower operation cost, through simplified architectures and the need to management fewer VMs.

It felt a little old-school that the vendor chose to highlight the platform at ISC, given that the whole AI world had turned its attention to Paris, where Jensen Huang was preparing to kick off GTC. However, the Paris event was apparently a surprise to the industry, put together at incredibly short notice.