NVIDIA Jetson Nano Developer Kit is a small, powerful computer that lets you run multiple neural networks in parallel for applications like image classification, object detection, segmentation, and speech processing.

Video

What is retraining?

Following deployment of a trained AI model the intention is that edge-based inferencing hardware will then carry out the intended task of the model, whether this be to identify structural defects in a manufacturing process, monitor traffic flow across a smart city environment or carry out security surveillance in an airport. Although the intention of any AI model is to exceed human capability (96% accuracy or higher), changes in the materials or subjects being monitored may require retraining of the AI model. Rather than the original training phase needing huge GPU-accelerated compute resource such as NVIDIA DGX systems, this retraining phase is usually minor tweaking to a model so this kind of power is not required.

Even though a large amount of GPU compute power is not needed, you may still be talking about relatively large volumes of data to transfer to the retraining system and using only edge technologies like wireless or 4G / 5G will prove very restrictive. Additionally, inferencing devices naturally sit at the edge of the AI network, often in a harsh or exposed conditions, whether this be extremes of temperature, dust or weather. This combination of conditions best suit an edge optimised industrial server designed for this more challenging environment.

What is an industrial server?

Industrial servers are designed to work reliably and continuously in high temperature and/or harsh environments and are rated to operate under continuous full load. This is achieved through a combination of more efficient long-life capacitors, components with an extended temperature range and that are resistant to dust and shocks. Industrial servers also offer enhanced longevity support, revision control, design flexibility, and industrial grade reliable operation - these provide assured reliability and performance - important as failure is not an option.

Icon

Maintaining equipment
longevity

• Eliminates yearly system upgrades
• Low total cost of ownership, including system certification costs and RMA service preparation

Icon

Strict revision control
for consistency

• Reduces product validation during product lifecycle
• Avoids compatibility issues resulting from engineering changes

Icon

Industrial-grade designs
for critical applications

Take advantage of the massive amount of data from your equipment to train your AI algorithms faster and optimise your operations at scale.

The Advantech Range of Industrial GPU Servers

Advantech industrial-grade servers have better endurance and redundancy in critical application environments which commercial servers could not operate in. These servers have wider ambient temperature validation for greater reliability. They successfully reduce system shut downs caused by high ambient temperatures. Strict vibration tests make our servers more durable, especially in higher shock and vibration environments like factory automation. Advantech servers feature filters in the front panel that blocks external dust ingress while easy filter installation and replacement ensures smooth continuous operation.

Characteristics Advantech Industrial-Grade Servers Commercial Servers
Temperature Tolerance -0C - 40C
(Carrier-grade -5C - 55C)
10C - 35C
Vibration Duration 0.25 - 1G 0.25 - 0.5G
Dust Prevention Dust Filter Support -
Technical Support Long Medium to Low
Longevity 5-7 years 3 years

The Advantech SKY-6000 series are high-density GPU AI retraining industrial servers designed to meet the growing trend toward big data and analysis. The SKY-6000 series is powered by dual Intel Xeon scalable processors and support up to 10 NVIDIA GPU accelerators. They feature IPMI remote management functions and smart fan control, leading to better acoustic and the thermal management.

System

Any SKY-6000 GPU server makes an extremely powerful, reliable and versatile edge solution when connected to the Advantech MIC-7xx range of AI edge inference systems. The server can either be used with a single GPU, say an NVIDIA Tesla T4 card for enhanced inference if the embedded Jetson platform within the MIC units require additional processing. Alternatively the server can be equipped with multiple high performance NVIDIA Tesla V100 GPU cards to handle significant retraining and/or inferencing workloads.

Retraining Workflow

System

The SKY-6000 Server Range

  System System System System
Specification SKY-6100 SKY-6200 SKY-6400 SKY-6420
Key Applications • Cloud Computing
• IoT Edge Computing
• Big Data Analytics
• Cloud Computing
• HPC / Data Analytics
• DataCenter Applications
• Cloud Computing
• Big Data Analytics
• High End Enterprise Server
• Cloud Computing
• Big Data Analytics
• Research lab/National Lab
Features • 2 2.5” Hot-swap SAS/SATA drive bay
• 8 DIMM slots, Intel Optane DCPMM
• NVIDIA Tesla P4/T4 Optimized
• Unify front bezel design
• 8 2.5” Hot-swap SAS/SATA support
• 24 DIMM slots, Intel Optane DCPMM
• 4 double deck PCI-E cards or 8 single deck PCI-E cards
• Unify front bezel design
• 8 2.5”/3.5” hot-Swap SAS/ SATA support
• 12 DIMM slot, Intel Optane DCPMM
• 205W CPU support
• 6 PCI-E cards support
• Unify front bezel design
• 12 2.5”/3.5” Hot-swap SAS/SATA support
• 24 DIMM slots
• Hot-Swappable system fan design
• 11 PCI-E cards support
• Pear to Pear support
Processor Support Dual Intel® Xeon® Scalable/2nd Gen. Scalable Family Processor (Cascadelake/Skylake) with UPI up to 10.4GT/s, TDP up to 140W Dual Intel® Xeon® Scalable/2nd Gen. Scalable Family Processor (Cascadelake/Skylake) with UPI up to 10.4GT/s, TDP up to 140W Dual Intel® Xeon® Scalable/2nd Gen. Scalable Family Processor (Cascadelake/Skylake) with UPI up to 10.4GT/s, TDP up to 205W Dual Intel® Xeon® Scalable/2nd Gen. Scalable Family Processor (Cascadelake/Skylake) with UPI up to 10.4GT/s, TDP up to 160W
Serverboard SKY-6100 SKY-6200 ASMB-975I SKY-6420
Chipset Intel® C622 chipset Intel® C622 chipset Intel® C622 chipset Intel® C622 chipset
System Memory (Max) 8 DIMM slots, Up to 1TB ECC LRDIMM, Up to 2666 MHz, Intel Optane DCPMM 24 DIMM slots, Up to 3TB ECC 3DS LRDIMM, Up to 2666 MHz, Intel Optane DCPMM 12 DIMM slots, Up to 768GB ECC LRDIMM, Up to 2666 MHz, Intel Optane DCPMM 24 DIMM slots, Up to 3TB ECC 3DS LRDIMM, Up to 2666 MHz, Intel Optane DCPMM
Expansion Slots 5 PCIE 3.0 x 16 (FH, HL) 4 PCIE 3.0 x 16 (FH, 10.5"L, double deck) or 8 PCIE 3.0 x 8 (FH, 10.5"L, single deck); 1 PCIE 3.0 x 8 (FH, HL) 4 PCIE 3.0 x 16 (FH, 10.5", double deck); 1 PCIE 3.0 x 8 (FH single deck); 1 PCIE 3.0 x 4 (FH, single deck) 10 PCIE 3.0 x 16 (FH, 10.5"L, double deck); 1 PCIE 3.0 x 16 (FH, single deck)
Onboard Storage Controller Intel® C622 SATA3 (6Gb/s) controller Intel® C622 SATA3 (6Gb/s) controller Intel® C622 SATA3 (6Gb/s) controller Intel® C622 SATA3 (6Gb/s) controller
Connectivity 2 Intel® X557 10GBase-T + 1 Intel® I210 Gigabit Ethernet ports; VGA ports; 3 USB 3.0 (2 in rear, 1 internal); 2 USB 2.0 at front 2 Intel® X557 10GBase-T + 2 Intel® I210 Gigabit Ethernet ports; VGA ports; 4 USB 3.0 ports (rear); 2 USB 2.0 at front; 1 Serial port optional 2 Intel® I210 Gigabit Ethernet ports; VGA ports; 7 USB 3.0 ports (4 in rear, 2 at front, 1 type A);1 Serial port 2 Intel® X557 10GBase-T + 1x Realtek RTL8201EL-VC PHY (dedicated IPMI); VGA ports; 6 USB 3.0 ports (4 in rear, 2 at front); 2 Serial port optional
Management Controller Aspeed AST2500 BMC Aspeed AST2500 BMC Aspeed AST2500 BMC Aspeed AST2500 BMC
Management IPMI2.0; KVM with share NIC IPMI2.0; KVM with share NIC IPMI2.0; KVM with share NIC LAN; SUSI API; WISE-PaaS RMM IPMI2.0; KVM with share NIC
Peripheral Bays 2 hot-swap 2.5” drive support; 2 SAS/SATA3 ports; on board 1 M.2 2242 SATA 2 hot-swap 2.5” drive support; 2 SAS/SATA3 ports; on board 1 M.2 2242 SATA 8 hot-swap 2.5” drive support; 8 SAS/SATA3 ports; 2 internal 2.5” drive support; on board 2 M.2 2242(SATA) for OS mirror 12 hot-swap 2.5”/3.5” SAS/SATA3 ports; on board 1 M.2 2280 (SATA+PCIE)
Power Supply 1200W 1+1 platinum level redundant power supply 2000W 1+1 platinum level redundant power supply 2000W 1+1 platinum level redundant power supply 4800W 3+1 platinum level redundant power supply
Cooling System 6 high speed 4056 system fan; 1 internal 4028 system fan; 1 optional 4028 external fan 6 high speed 8038 fan; 2 for CPU, 4 for riser card cage 2 CPU fan; 3 high speed 12038 internal system fan; 2 high speed external 8038 system fan 6 high speed 12038 system fan; 4 optional external 8038 fan
Form Factor 1U chassis; enclosure: 438 x 44 x 650 mm (17.2" x 1.7" x 25.6") 2U chassis; enclosure: 438 x 88 x 760 mm (17.24" x 3.46" x 29.92") 4U chassis; enclosure: 435 x 177 x 673 mm (17.12" x 6.96" x 26.49") 4U chassis; enclosure: 438 x 176 x 770 mm (17.24" x 6.93" x 30.31")
Operating Temperature 0 ~ 35° C (32 ~ 95° F) 0 ~ 35 °C (32 ~ 95 °F) 0 ~ 35° C (32 ~ 85° F) *0 ~ 30° C (32 ~ 85.9° F) for NVIDIA Tesla P100/V100 0 ~ 35° C (32 ~ 85° F) *0 ~ 30° C (32 ~ 85.9° F) for NVIDIA Tesla V100
CONFIGURE NOW CONFIGURE NOW CONFIGURE NOW CONFIGURE NOW
Find out more Find out more