3RU, 4,608 Fibers. Front Access Rail Optimized Panel Solution.
GPX77 Vertical Modular ODF — 19″ or ORv3 rackmount for AI fabric and hyperscale spine-leaf consolidation. Add capacity one cassette at a time, all from the front.
3RU
288 MPO Ports
4,608 Fibers
Shuffle & Pass-Through
1,536 Fibers per RU
Eighteen vertical module positions in a 3RU frame, each accepting single-row (8 MPO-8) or double-row (16 MPO-16) cassettes. Fully loaded with double-row modules: 288 MPO ports and 4,608 fibers — managed entirely from the front in a chassis only 300mm deep.
No Hot-Aisle Work
All cable handling — trunk cable management, patching, shuffle mapping, and MACs — is performed from the front. The frame is access-neutral from left or right. Technicians never need to work behind the frame, eliminating hot-aisle hazards and rear-access constraints in dense co-location environments.
4×4 and 8×8 Built In
Factory-shuffled flex-circuit cassettes distribute leaf channels across independent spine rails. A single HD77-1R8M cassette handles 4 leaf uplinks and 4 spine downlinks in a rail-optimized 4×4 configuration. Mix 100G, 200G, and 400G lanes in the same frame and reconfigure on demand without service disruption.
Incremental Spine-Collapse
Deploy exactly four or eight leaves per cassette. No day-one over-provisioning. Add one HD77 cassette for every four new leaves and grow from 4 to 72 leaf ports within the same GPX77 frame. Your fiber infrastructure scales with your transceiver consolidation, not ahead of it.
One Frame. Four Cassette Personalities.
HD77-1R8M-T2-SH4 — 4×4 Shuffle
- 1 row × 8 MPO-8 ports
- 4 leaf inputs / 4 spine outputs
- Rail-optimized channel distribution for 800G or 400G DR4 spine-collapse or graceful degradation
- 64 fibers per cassette
HD77-2R8M-T2-SH8 — 8×8 Shuffle
- 2 rows × 16 MPO-16 ports
- 8 leaf inputs / 8 spine outputs
- Higher-density 1.6T or 800G DR8 rail distribution for scale-out clusters
- 256 fibers per cassette
Technical Specifications
Deployment Configurations
Four Leaves at a Time
Replace 100G sprawl with 400G consolidation. Each HD77-1R8M-SH4 cassette connects four leaf uplinks to four spine downlinks via factory-shuffled rail distribution. A 2,000-server deployment drops from 24 spine switches to 8, and from 1,280 × 100G transceivers to 320 × 400G. Your CAPEX follows your actual server count, not your three-year forecast.
Rail-Optimized GPU Clusters
Distributed AI training demands deterministic lane mapping between GPU blades and spine switches. The 4×4 and 8×8 shuffle cassettes pre-engineer the physical layer so each spine rail receives one channel from each leaf. No field polarity guessing. No manual patch cord nests. Just validated, repeatable fiber paths from day one.
White-Space Efficiency
In leased co-location facilities, rack space is revenue. At 1,536 fibers per RU and only 300mm depth, the GPX77 maximizes patching density without protruding into aisle ways or requiring rear access clearance. Front-only operation means you can place frames back-to-back and still manage every connection.
Standards Alignment
Collapse Your Spine. Expand Your Leaves.
Discuss your 400G spine-leaf density target and cassette mix with our engineering team. See the GPX77 vertical shuffle architecture and front-access workflow live at our booth.