The SCA8000 packs eight powerful NVIDIA Tesla V100 SXM2 GPUs connected via NVIDIA NVLink™ in a single GPU expansion accelerator

Find a Distributor

With up to four PCI-SIG PCIe Cable 3.0 compliant links to the host server up to 100m away, the SCA8000 supports a flexible upgrade path for new and existing datacenters with the power of NVLink without upgrading server infrastructure. With advanced, independent IPMI system monitoring and full featured SNMP interface not available in any other GPU accelerator with NVLink, the SCA8000 fits seamlessly into any size datacenter.


  • Dimensions: 29.5” deep x 17.2” wide 5.25” high (3U)
  • Four Optional  PCIe x16 slots for customer I/O cards face the rear of the chassis
  • Supports four half-height, half-length, single-wide PCIe x16 cards with IO bracket access
  • Front panel LEDs
  • Four rear panel PCIe x16 Gen3 cable interfaces
  • 5 hot-swap fans
  • Weight: 68lbs  as standard OSS configuration
  • Eight Tesla V100
  • 1 petaFLOPS (GPU FP16) performance
  • 256GB total GPU HBM2 ECC Memory
  • 40,960 CUDA® Cores
Main Backplane
  • 300GB/s NVLink 2.0 bidirectional peer-to-peer bandwidth
  • Four PCIe 3.0 x16 cable inputs to rear of enclosure (OSS HIB based )
  • PCIe 3.0 switch fabric manages PCIe cross connects from cables to GPUs and I/O Slots
  • 2x RJ45 connectors for IPMI v2.0 System Monitor
  • 1x HD DB-9 serial port for IPMI network configuration
  • 6000W redundant power subsystem
  • Four 2,000-watt front removable, hot-swap supplies
  • 3+1 redundant with full current sharing operation
  • 2,000W each at 208-277VAC, 15A max input
  • 1,500W each at 90-124VAC, 15A max input
  • 15A breaker and IEC C14 power input at rear for each supply
  • +12V and +12V standby voltage outputs
  • All +12V power rails shared on copper bus bar delivery system
Power Cords
  • The SCA8000 ships with C13-C14 15A PDU type power cords
  • 240V power cord for PDUs
    • OSS Part number:  OSS-CBL-PWR-C13-C14-15A-8
    • IEC C13 to IEC C14, Straight, 14AWG, 15A, 8' (2.44m)
  • Other power cords available on request
System monitoring/alarming
  • Fully IPMI v2.0 compliant monitoring, control & alarming system Temperature
    • Monitors inlet & exhaust temps
    • Fan speed auto adjusts by temp
    • Alarm set-points for over temp
  • Fans
    • Monitors all system fan tachs
    • PWM fan speed control
    • Alarms for slow or failed fans
  • Power
    • Alarms for supply failure
    • Alarms for voltages out of range
    • Monitors output voltage rails
    • Monitors supply telemetry
  • Add-in Cards
    • Alarms for abnormal card telemetry
    • Monitors add-in card I2C SM bus
  • Interface

Supports SNMP and RCMP

CLI or web GUI

  • Remote chassis and canister LED tagging
  • Five dual 80 x 80 x 38mm fans on internal fan tray
  • All fans are 141CFM each
  • All fans PWM monitored and speed controlled by the IPMI system monitor
  • Fans hot-swap from small top cover near front of chassis
  • Power supplies separately cooled from internal 25mm fans
Operating Environment
  • Temperature range:
    • Operating: 5°—35°C
    • Storage: -40°—85°C
  • Humidity range:
    • Operating: 10% to 80% relative (non-condensing)
    • Non-operating: 5% to 95% relative (non-condensing)
  • Altitude range:
    • Operating: 0 to 10,000 ft.
    • Storage: 0 to 50,000 ft.
    • Maximum noise rating of 74.5dB(A)
Agency compliance

Designed to meet the following agency certifications with testing currently pending:

FCC - Part 15 of the FCC Rules, Class A, 47CFR, Canada ICES-003, issue 4, Class A, UL/IEC 60950-1, Canada: CSA C22.2 No. 60950-1, Argentina: IEC60950-1, Japan: VCCI, Class A, Australia/New Zealand AS/NZS CISPR 22, Class A, IEC 60950-1 (CB Certificate and CB Test Report), CE Mark (EN55022 Class A, EN60950-1, EN55024, EN61000-3-2, EN61000-3-3), CISPR 22, CISPR 24, Class A, CE Emissions 2004-108EC, RoHS compliance (Directive 2002/95/EC), CCN NWGQ, NWGQ7


Specifications for SXM2 GPUs

Tesla V100 SXM2
Tesla P100 SXM2
GPU Architecture Volta Pascal
NVIDIA Tensor Cores
Double-Precision Performance
7.8 Tflops
5.3 Tflops
Single-Precision Performance
15.7 Tflops
10.6 Tflops
Half-Precision Performance N/A 21.2 Tflops
Tensor Performance
125 Tflops
GPU Memory 16GB HBM2 16 GB CoWoS HBM2
Memory Bandwidth 900 GB/sec 732 GB/sec
System Interface NVIDIA NVLink NVIDIA NVLink
Form Factor SXM2 SXM2
Max Power Consumption 300 W 300W
Thermal Solution Passive Passive
Compute APIs CUDA. DirectCOmpute, OpenCL, OpenACC CUDA. DirectCOmpute, OpenCL, OpenACC