Live DB — analyst tables
Test
HS-Feynman NVL72-CX10-IB-4R
HS-Feynman NVL72-CX10-IB-4R
Default Hyperscaler · Feynman · CX10
Rack type
Oberon NVL72
Cluster / location
Default Hyperscaler Oberon NVL72 x800 Q3400 (non-CPO)
Buyer profile
Hyperscaler
Networking type
Infiniband
Number of XPUs
93,312
NIC used
CX10
NICs / XPU
3
Scale-out BW / XPU
4.8T
XPU / server
72
Scale-up network
5 rows
Assigned blockDefault
Feynman - Oberon NVL72 - Generic
Scale-UpXPU: Feynman Oberon NVL725 linesshared by 10 SKUs
Scale-up Units are based on the scale-up world size (72 XPUs).
XPU World Size
2
WS| Scale-Up Network Layer | Part / Model | Units# | Unit BWGbit/s uni-d | Port Count# | Port BWGbit/s uni-d |
|---|---|---|---|---|---|
| XPUs per Node | · | 72 | 28,800 | 144 | 400 |
| Scale-Up World Size | · | 72 | 2,073,600 | · | · |
Switches
4
SW| Scale-Up Network Layer | Part / Model | Units# | Unit BWGbit/s uni-d | Port Count# | Port BWGbit/s uni-d |
|---|---|---|---|---|---|
| per Switch | Nvidia NVSwitch 8.0 (NVLink 8) | 1 | 115,200 | 288 | 400 |
| Layer 1 Switches | Nvidia NVSwitch 8.0 (NVLink 8) | 18 | 2,073,600 | 288 | 400 |
| Layer 2 Switches | Nvidia NVSwitch 8.0 (NVLink 8) | 0 | 0 | 288 | 400 |
| Scale-Up Switch Total | · | 18 | 2,073,600 | · | · |
Connections & cabling
12
CN| Scale-Up Network Layer | Part / Model | Units# | BW / PortGbit/s uni-d | Ports# | Bidi / Port# | Total BW# | BW / Connector | Cable / Model | Cable Units# |
|---|---|---|---|---|---|---|---|---|---|
| Optical Engines | NV 3.2T Optical Engine (OE) for NVLink Switch side CPO | 0 | 200 | 16 | · | 0 | 3,200 | NV 1.6T 8 Fiber Pair FAU | 0 |
| External Laser Source | NV External Laser Source (ELS) 8 LD 6.4T for NVLink CPO Switch | 0 | 200 | 32 | · | 0 | 6,400 | · | · |
| L2 Switch Total | |||||||||
| Male Connectors to GPU | NV Sentry HD 448Gb/s 144DP (Male) | 0 | 400 | 72 | · | 2,073,600 | 28,800 | Already Counted Below | · |
| Female Connectors to GPU | NV Sentry HD 448Gb/s 144DP (Female) | 0 | 400 | 72 | · | 2,073,600 | 28,800 | · | · |
| Optical Engines | NV 3.2T Optical Engine (OE) for NVLink Switch side CPO | 0 | 200 | 16 | · | 1,209,600 | 3,200 | NV 1.6T 8 Fiber Pair FAU | 0 |
| External Laser Source | NV External Laser Source (ELS) 8 LD 6.4T for NVLink CPO Switch | 0 | 200 | 32 | · | 1,209,600 | 6,400 | · | · |
| L1 Switch Total | |||||||||
| Male Connectors | NV Sentry HD 448Gb/s 144DP (Male) | 72 | 400 | 72 | · | 2,073,600 | 28,800 | NV 224Gb/s SkewClear EXD Gen 2 Bidi Cable | 10,368 |
| Female Connectors | NV Sentry HD 448Gb/s 144DP (Female) | 72 | 400 | 72 | · | 2,073,600 | 28,800 | Already Counted Below | · |
| XPU Total | ** One connector per GPU | · | · | · | · | · | 28,800 | · | · |
| · | NVL Blade Column Spine Chassis (NvLink 7) | 1 | · | · | · | · | · | · | · |
| · | NVL Kyber Canister Midplane PCB (NvLink 7) | 0 | · | · | · | · | · | · | · |
| - | NV 224Gb/s SkewClear EXD Gen 2 Bidi Cable | 10,368 | · | · | · | · | · | · | · |
Scale-Up Parts Summary
5
| Item | Qty/Cluster | Qty/Server | Attach/XPU |
|---|---|---|---|
| Nvidia NVSwitch 8.0 (NVLink 8) | 23,328 | 18.00 | 0.25 |
| NV Sentry HD 448Gb/s 144DP (Male) | 93,312 | 72.00 | 1.00 |
| NV Sentry HD 448Gb/s 144DP (Female) | 93,312 | 72.00 | 1.00 |
| NV 224Gb/s SkewClear EXD Gen 2 Bidi Cable | 26,873,856 | 20,736.00 | 288.00 |
| NVL Blade Column Spine Chassis (NvLink 7) | 1,296 | 1.00 | 0.01 |
Backend network
4 rows
Assigned blockDefault
Feynman - Oberon NVL72 - CX10 - 3L - IB - X800-Q3400-RA - Generic
BackendXPU: Feynman Oberon NVL724 linesshared by 2 SKUs
Backend Units are based on the cluster size (93,312 XPUs).
Switches
5
SW Network Layer | Switch and Server Spec | Aggregate Port Utilization | Network Bandwidth Aggregate | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SwitchPart Name / Model | Units# | Unit BWGbit/s | Port BWGbit/s | Downlink Ports / SW# | Uplink Ports / SW# | Max Ports# | Downlink Ports# | Uplink Ports# | Total Ports# | Downlink BWTbit/s | Uplink BWTbit/s | OversubRatio | |
| XPUs | · | 93,312 | 4,800 | 4,800 | 0 | 93,312 | 93,312 | 0 | 93,312 | 93,312 | 0 | 447,897.6 | · |
| NICs per Node | ConnectX-10 NIC - 1.6T | 12 | 1,600 | 100 | · | · | 8 | · | · | · | · | · | · |
| Leaf | Quantum-X800-Q3400-RA (Air Cooled) | 7,776 | 115,200 | 800 | 72 | 72 | 144 | 559,872 | 559,872 | 1,119,744 | 447,897.6 | 447,897.6 | 1 |
| Spine | Quantum-X800-Q3400-RA (Air Cooled) | 7,776 | 115,200 | 800 | 72 | 72 | 144 | 559,872 | 559,872 | 1,119,744 | 447,897.6 | 447,897.6 | 1 |
| Core | Quantum-X800-Q3400-RA (Air Cooled) | 3,888 | 115,200 | 800 | 144 | · | 144 | 559,872 | 0 | 559,872 | 447,897.6 | 0 | · |
Transceivers
7
CN Network Layer | Transceiver / CablePart Name / Model | Total Units# | Units per SW/GPU# | Unit BWGbit/s | Lanes# | BW per LaneGbit/s | Agg. BWTbit/s |
|---|---|---|---|---|---|---|---|
| XPU Nodes | Nvidia LinkX 1.6T DR8 (SiPho) | 93,312 | 93,312 | 4,800 | 8 | 600 | 447,897.6 |
| Leaf | |||||||
| Downlink | Nvidia LinkX 1.6T DR8 (SiPho) | 279,936 | 36 | 1,600 | 8 | 200 | 447,897.6 |
| Uplink | Nvidia LinkX 1.6T DR8 (SiPho) | 279,936 | 36 | 1,600 | 8 | 200 | 447,897.6 |
| Spine | |||||||
| Downlink | Nvidia LinkX 1.6T DR8 (SiPho) | 279,936 | 36 | 1,600 | 8 | 200 | 447,897.6 |
| Uplink | Nvidia LinkX 1.6T DR8 (SiPho) | 279,936 | 36 | 1,600 | 8 | 200 | 447,897.6 |
| Core | |||||||
| Downlink | Nvidia LinkX 1.6T DR8 (SiPho) | 279,936 | 72 | 1,600 | 8 | 200 | 447,897.6 |
| Uplink | · | · | · | · | · | · | · |
| Super-Core | CPO is used here - so there is no need for Transceivers on Switches | · | · | · | · | · | · |
Cables
7
CB Network Layer | Cable/ FiberType | Total Units# | Units per Transceiver# | Unit BWGbit/s | Lane Count# | SerDes BWGbit/s |
|---|---|---|---|---|---|---|
| XPU Nodes | Nvidia LinkX 800G Single Mode | 559,872 | 6 | 800 | 4 | 200 |
| Leaf | ||||||
| Downlink | Already counted above | · | · | · | · | · |
| Uplink | Nvidia LinkX 800G Single Mode | 559,872 | 2 | 800 | 4 | 200 |
| Spine | ||||||
| Downlink | Already counted above | · | · | · | · | · |
| Uplink | Nvidia LinkX 800G Single Mode | 559,872 | 2 | 800 | 4 | 200 |
| Core | ||||||
| Downlink | Already counted above | · | · | · | · | · |
| Uplink | · | · | · | · | · | · |
| Super-Core | · | · | · | · | · | · |
Backend Parts Summary
4
| Item | Qty/Cluster | Qty/Server | Attach/XPU |
|---|---|---|---|
| Nvidia LinkX 1.6T DR8 (SiPho) | 1,492,992 | 1,152.00 | 16.00 |
| Nvidia LinkX 800G Single Mode | 1,679,616 | 1,296.00 | 18.00 |
| ConnectX-10 NIC - 1.6T | 279,936 | 216.00 | 3.00 |
| Quantum-X800-Q3400-RA (Air Cooled) | 19,440 | 15.00 | 0.21 |
Frontend network
11 rows
Assigned blockDefault
Feynman - Oberon NVL72 - SN6600 - Generic
FrontendXPU: Feynman Oberon NVL7211 linesshared by 11 SKUs
Converged Frontend + Storage Fabric
11
| Layer | Part / Model | Units | Attach / XPU |
|---|---|---|---|
| · | Spectrum-6 102.4T x1600 (SN6600) | 1,120 | 0.0120 |
| · | Spectrum-4 25.6T (SN5400) | 5 | 0.0000 |
| · | Spectrum-3 12.8T (SN4700) | 167 | 0.0018 |
| · | Spectrum-2 6.4T (SN3700) | 1,370 | 0.0147 |
| · | Nvidia LinkX 1.6T DR8 (SiPho) | 54,187 | 0.5807 |
| · | Nvidia LinkX 800G DR4 | 53,904 | 0.5777 |
| · | Nvidia LinkX 200G SR4 | 23,920 | 0.2563 |
| · | Nvidia LinkX 800G Multimode | 74,159 | 0.7947 |
| · | Nvidia LinkX 400G Multimode | 380 | 0.0041 |
| · | Nvidia LinkX 200G Multimode | 10,655 | 0.1142 |
| · | RJ45 Cables | 65,763 | 0.7048 |
Frontend Parts Summary
11
| Item | Qty/Cluster | Qty/Server | Attach/XPU |
|---|---|---|---|
| Spectrum-6 102.4T x1600 (SN6600) | 1,120 | 1 | 0.0120 |
| Spectrum-4 25.6T (SN5400) | 5 | 0 | 0.0001 |
| Spectrum-3 12.8T (SN4700) | 167 | 0 | 0.0018 |
| Spectrum-2 6.4T (SN3700) | 1,370 | 1 | 0.0147 |
| Nvidia LinkX 1.6T DR8 (SiPho) | 54,187 | 42 | 0.5807 |
| Nvidia LinkX 800G DR4 | 53,904 | 42 | 0.5777 |
| Nvidia LinkX 200G SR4 | 23,920 | 18 | 0.2563 |
| Nvidia LinkX 800G Multimode | 74,159 | 57 | 0.7947 |
| Nvidia LinkX 400G Multimode | 380 | 0 | 0.0041 |
| Nvidia LinkX 200G Multimode | 10,655 | 8 | 0.1142 |
| RJ45 Cables | 65,763 | 51 | 0.7048 |
Compute Tray Networking Content
4 rows
Networking components inside the compute tray — accelerators, board-to-board connectors, retimers, and their per-tray bandwidth.
Assigned blockDefault
Feynman - Oberon NVL72 - Generic
Compute TrayXPU: Feynman Oberon NVL724 linesshared by 11 SKUs
| Tray layer | Part / Model | Units per Tray | Unit BWGb/s | Lanes | Lane BWGb/s | Per switch |
|---|---|---|---|---|---|---|
| XPUs per Tray | Rubin 200 | 4 | · | · | · | · |
| PCIe 6.0 Connector (Male) | NV Paladin Board to Board 32 Pin 4p8c (Male) | 2 | 512 | · | · | · |
| PCIe 6.0 Connector (Male) | NV Paladin Board to Board 64 Pin 4p16c (Male) | 8 | 1,024 | · | · | · |
| PCIe 6.0 Connector (Female) | NV Paladin Board to Board 32 Pin 4p8c (Female) | 4 | 512 | · | · | · |
| PCIe 6.0 Connector (Female) | NV Paladin Board to Board 64 Pin 4p16c (Female) | 16 | 1,024 | · | · | · |
| PCIe 6.0 Connector (Male) | NV Paladin Board to Board 32 Pin 4p8c (Male) | 2 | 512 | · | · | · |
| PCIe 6.0 Connector (Male) | NV Paladin Board to Board 64 Pin 4p16c (Male) | 8 | 1,024 | · | · | · |
Notes
- PCIe 6.0 Connector (Male) · NV Paladin Board to Board 32 Pin 4p8c (Male) — 2 pins = 1DP
- PCIe 6.0 Connector (Male) · NV Paladin Board to Board 64 Pin 4p16c (Male) — 2DP = 1Lane
Compute Tray Parts Summary
4
| Item | Qty/Cluster | Qty/Server | Attach/XPU |
|---|---|---|---|
| NV Paladin Board to Board 32 Pin 4p8c (Male) | 93,312 | 72.00 | 1.00 |
| NV Paladin Board to Board 64 Pin 4p16c (Male) | 373,248 | 288.00 | 4.00 |
| NV Paladin Board to Board 32 Pin 4p8c (Female) | 93,312 | 72.00 | 1.00 |
| NV Paladin Board to Board 64 Pin 4p16c (Female) | 373,248 | 288.00 | 4.00 |