Lambda runs the first production co-packaged optics GPU cluster
NVIDIA and Lambda say a 10,000-GPU cluster already uses co-packaged optics switches, years before many expected the technology in production.
On this page · 1 min read
Key takeaways
- 1At SC26 NVIDIA and Lambda disclosed a production cluster of more than 10,000 GPUs on NVIDIA’s co-packaged optics (CPO) switches.
- 2The Quantum-X CPO switch has 144 ports at 800 Gbps, fed by 18 pluggable light sources that can be serviced.
- 3CPO cuts power by moving optics onto the switch package, so more of a fixed power budget can go to GPUs.
Original source · servethehome.com
NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster… It is FINALLY real
Published
What happened
Co-packaged optics put the optical modules on the same package as the switch chip instead of in pluggable modules on the front panel. The industry has tried to bring that from the lab to production for decades. At SC26, according to ServeTheHome, NVIDIA and Lambda described a production cluster with more than 10,000 GPUs that already runs it.
NVIDIA’s first CPO version is in its Quantum-X switch, with Spectrum-X to follow in a second generation. The Quantum-X CPO switch has 144 MPO/MTP ports at 800 Gbps each. Instead of 144 optical modules, it uses 18 pluggable light sources on the front of the chassis. Fewer parts can fail, and they are easier to cool. Because the optics sit at the switch package, the high-speed copper links that would otherwise carry signals to the front cages are not needed, which cuts power use.
Why it matters
In a power-limited facility, every watt spent on networking is a watt not spent on GPUs. ServeTheHome argues the question is no longer whether CPO will be deployed but how quickly it is adopted.
In one sentence:optics power is a limit on how many GPUs fit in a data center, so CPO changes how dense AI networks can get.
What it means for you
- Serviceable light sources were the main obstacle for CPO in switches, since light sources are the part that fails in pluggable optics. This design makes them replaceable.
- Plan fibre and MPO/MTP cabling for 800G ports if you evaluate Quantum-X class switches.
- Expect GPU hosting providers to quote CPO-based fabrics as a density and power advantage.