Lenovo ThinkStation PGX Cluster Cable: 0.5m QSFP112 400G DAC
To cluster two Lenovo ThinkStation PGX AI workstations you need one 0.5m QSFP112 400G passive DAC cable across their dual 200Gb QSFP ports. The PGX is Lenovo's small-form-factor take on NVIDIA's GB10 Grace Blackwell platform, and its two QSFP ports expose the same ConnectX-7 controller as the DGX Spark, so it takes the exact cable NVIDIA approves in its Spark Stacking documentation (Amphenol NJAAKK0006 / Luxshare LMTQF022-SD-R spec). Petronella Technology Group, Inc. keeps it in stock at $159 with free US shipping, while distributors quote $179 to $229 and are usually backordered.
The ThinkStation PGX brings GB10 into Lenovo's enterprise workstation line, which is why IT teams tend to buy them through managed procurement and deploy them in pairs or small clusters rather than one at a time. When a second unit lands, the only part standing between two boxes and a working cluster is this cable, and it is the one component that does not come in the box. The PGX pairs a 20-core Arm CPU with a Blackwell GPU and 128 GB of unified LPDDR5x memory at up to 273 GB/s, then exposes clustering over two 200Gb QSFP ports.
| Cable | 0.5m QSFP112 400G passive direct-attach copper (DAC) |
| Approved spec | Amphenol NJAAKK0006 / Luxshare LMTQF022-SD-R (NVIDIA Spark Stacking docs) |
| Fits | Both 200Gb QSFP ports on the ThinkStation PGX ConnectX-7 controller |
| Platform | NVIDIA GB10 Grace Blackwell, 128 GB LPDDR5x, DGX OS |
| For two PGX units | One cable |
| For a three-node ring | Three cables (switchless, each box uses both ports) |
| Price | $159, free US shipping (in stock) |
Clustering ThinkStation PGX units: how many cables and when you need a switch
Each ThinkStation PGX has two QSFP ports, so it follows the NVIDIA-documented GB10 topology: two units link directly with one cable, and three units form a switchless ring with three cables, each box using both ports for full pairwise connectivity. Three nodes is the switchless ceiling; with only two ports per unit, four or more PGX units need a 200G-class QSFP switch. Clustering pools memory so a model too large for one 128 GB node can run across the fleet. It is a capacity play, not a raw speed boost, since each link runs around 25 GB/s, which is exactly what memory-bound inference and fine-tuning on larger models needs. A PGX also clusters cleanly with a DGX Spark or any other OEM GB10 box, because they all share the ConnectX-7 NIC and QSFP port spec.