Live DB — analyst tables
Test
Meta 16000 A100 CX-6 Hyperscaler Ethernet Cluster Configuration

META-A100-CX6-IB-NV-16KRSC

Meta · A100 · CX-6

Rack type
NVL8
Cluster / location
16k RSC Infiniband Cluster
Buyer profile
Hyperscaler
Networking type
Ethernet
Number of XPUs
16,000
NIC used
CX-6
NICs / XPU
1
Scale-out BW / XPU
200G
XPU / server
8

Scale-up network

1 rows
Assigned blockDefault
A100 - NVL8
Scale-UpXPU: A100 NVL81 lineshared by 9 SKUs

Scale-up Units are based on the scale-up world size (8 XPUs).

XPU World Size

2
WS
Scale-Up Network LayerPart / Model
Units#
Unit BWGbit/s uni-d
Lane Count#
Lane BWGbit/s uni-d
XPUs per Node·82,40024100
Scale-Up World Size·819,200··

Switches

2
SW
Scale-Up Network LayerPart / Model
Units#
Unit BWGbit/s uni-d
Lane Count#
Lane BWGbit/s uni-d
per SwitchNvidia NVSwitch 2.0 (NVLink 3)62,40024100
Scale-Up Switch Total··14,400··
Notes
  • XPUs per Node — 600GB Bidi

Scale-Up Parts Summary

1
ItemQty/ClusterQty/ServerAttach/XPU
Nvidia NVSwitch 2.0 (NVLink 3)12,0006.000.75

Backend network

4 rows
Assigned blockDefault
A100 - NVL8 - CX6 - 3L - IB - QM8700 - META
BackendXPU: A100 NVL84 linesshared by 1 SKU

Backend Units are based on the cluster size (16,000 XPUs).

Switches

5
SW
 Network Layer
Switch and Server SpecAggregate Port UtilizationNetwork 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·16,000200200016,00016,000016,00016,00003,200·
NICs per NodeConnectX-6 NIC - 200G8200200··1······
LeafQuantum InfiniBand HDR (QM8700)8008,00020020204016,00016,00032,0003,2003,2001
SpineQuantum InfiniBand HDR (QM8700)8008,00020020204016,00016,00032,0003,2003,2001
CoreQuantum InfiniBand HDR (QM8700)4008,0002004004016,000016,0003,2000·

Transceivers

6
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 NodesNvidia LinkX 200G DR4 QSFP5616,00016,0002004503,200
Leaf
DownlinkNvidia LinkX 200G DR4 QSFP5616,000202004503,200
UplinkNvidia LinkX 200G DR4 QSFP5616,000202004503,200
Spine
DownlinkNvidia LinkX 200G DR4 QSFP5616,000202004503,200
UplinkNvidia LinkX 200G DR4 QSFP5616,000202004503,200
Core
DownlinkNvidia LinkX 200G DR4 QSFP5616,000402004503,200
Uplink·······
Super-Core

Cables

6
CB
 Network Layer
Cable/ FiberType
Total Units#
Units per Transceiver#
Unit BWGbit/s
Lane Count#
SerDes BWGbit/s
XPU NodesNvidia LinkX 200G Single Mode16,0001200450
Leaf
DownlinkAlready counted above·····
UplinkNvidia LinkX 200G Single Mode16,0001200450
Spine
DownlinkAlready counted above·····
UplinkNvidia LinkX 200G Single Mode16,0001200450
Core
DownlinkAlready counted above·····
Uplink······
Super-Core

Backend Parts Summary

4
ItemQty/ClusterQty/ServerAttach/XPU
Nvidia LinkX 200G DR4 QSFP5696,00048.006.00
Nvidia LinkX 200G Single Mode48,00024.003.00
ConnectX-6 NIC - 200G16,0008.001.00
Quantum InfiniBand HDR (QM8700)2,0001.000.13

Frontend network

11 rows
Assigned blockDefault
A100 - NVL8 - SN5600 - Generic
FrontendXPU: A100 NVL811 linesshared by 8 SKUs

Converged Frontend + Storage Fabric

11
LayerPart / ModelUnitsAttach / XPU
·Spectrum-4 51.2T (SN5600)1920.0120
·Spectrum-3 12.8T (SN4700)10.0000
·Spectrum-2 6.4T (SN3700)290.0018
·Spectrum-2 1GbE (SN2201)2350.0147
·Nvidia LinkX 800G SR89,2910.5807
·Nvidia LinkX 400G SR49,2430.5777
·Nvidia LinkX 100G SR4 QSFP284,1020.2563
·Nvidia LinkX 800G Multimode12,7160.7947
·Nvidia LinkX 400G Multimode650.0041
·Nvidia LinkX 100G Multimode1,8270.1142
·RJ45 Cables11,2760.7048

Frontend Parts Summary

11
ItemQty/ClusterQty/ServerAttach/XPU
Spectrum-4 51.2T (SN5600)1900.0012
Spectrum-3 12.8T (SN4700)000.0000
Spectrum-2 6.4T (SN3700)300.0002
Spectrum-2 1GbE (SN2201)2300.0015
Nvidia LinkX 800G SR892900.0581
Nvidia LinkX 400G SR492400.0578
Nvidia LinkX 100G SR4 QSFP2841000.0256
Nvidia LinkX 800G Multimode1,27210.0795
Nvidia LinkX 400G Multimode700.0004
Nvidia LinkX 100G Multimode18300.0114
RJ45 Cables1,12810.0705

Compute Tray Networking Content

1 rows

Networking components inside the compute tray — accelerators, board-to-board connectors, retimers, and their per-tray bandwidth.

Assigned blockDefault
A100 - NVL8 - Generic
Compute TrayXPU: A100 NVL81 lineshared by 8 SKUs
Tray layerPart / ModelUnits per Tray
Unit BWGb/s
Lanes
Lane BWGb/s
Per switch
XPUs per TrayAmpere 1008····
A100 ServerNVLink PCIe 4.0 x16 Retimer8256···

Compute Tray Parts Summary

1
ItemQty/ClusterQty/ServerAttach/XPU
NVLink PCIe 4.0 x16 Retimer16,0008.001.00