With this range of applications, a flexibly configurable embedded system helps the system integrator to both guarantee the necessary performance and avoid unnecessary investments.

The high-performance Edge AI servers of the Spectra PowerBox 500 series offer this possibility. They are highly scalable - the integrated Intel C246 chipset supports more than 20 different 8th and 9th generation Intel processors from Celeron to Xeon. The outstanding graphics performance is achieved by the graphics controller integrated on the chip. The required storage capacity is provided by two 2.5" SATA drives and can be expanded using several mSATA and NVME SSD slots.

The real flexibility is provided by the integrated MXM interface for additional GPU expansion. The new Quadro MXM GPU modules MXM-RTX3000 and MXM-T1000 are available. They use the Nvidia Quadro Turing GPU architecture based on the latest 12 nm process.

There is no need to compromise when it comes to computing power and performance. The MXM-RTX3000 features 1920 Cuda cores, 5.3 TFLOPs of peak FP32 high-end computing power, parallel integer execution, AI computing tensor core and specialized RT core for ray tracing. Together, these fully integrated GPU capabilities provide everything needed to handle complex visual image calculations. The MXM-T1000 features 896 Cuda cores, 2.6 TFLOPs of peak FP32 computing power and only 50W power consumption for low-power, high-speed computing.

Both models support GDDR6 memory. The MXM-RTX3000 has a single card capacity of up to 6 GB and a memory bandwidth of 336 GB/s. The MXM-T1000 has a capacity of 4 GB and a memory bandwidth of 192 GB/s. The fast memory smoothes the reading of data and multiplies the computing power.

A highly effective cooling concept enables fanless operation. This is achieved through the thermal separation of CPU and GPU and additional copper heat pipes in the aluminum housing, which ensure even better heat dissipation. If the processor and GPU module are expanded to the maximum, an external cooling system can easily be connected.