SpaceCloud® Solutions
Get the benefits of our computing solutions for SpaceCloud® cloud services
SpaceCloud solutions
Computer solutions for SpaceCloud®
Unibap offers space computer solutions for large and small spacecraft. A computer solution contains everything needed in terms of power generation, calculation, data storage, communication interface, etc. Therefore, a computer solution is not comparable to an individual computer card. The computer solutions are within a power budget of less than 30 W and a volume less than 1U CubeSat unit (95 mm x 95 mm x 50 mm) and typically a weight below 250 grams. Unibap’s computer solutions are powered by 12 V DC supply.
Unibap’s iX51xx computer solutions are based on 28 nm AMD APU circuits and FPGA from Microsemi with the possibility to connect one or more Intel Vision Processing Units (VPU) circuits from the Movidus Myriad family as extra computing power. An illustration of the iX5100 configuration is shown here with the Camera Link sensor interface and radio interface for S and X bands (up to 100 Mbps).
iX5100 illustration
iX10100 illustration
Unibap’s iX101xx computer solutions are based on 14 nm AMD Ryzen APU circuits and FPGA from Microsemi with the possibility to connect one or more Intel Vision Processing Units (VPU) circuits from the Movidus Myriad family as extra computing power. An illustration of the iX101xx configuration is shown here in basic version without I/O card.
iX5100
The iX5100 is the first generation SpaceCloud® computing solution based on the 28 nm AMD x86 processor and GPU technology in combination with the Microsemi SmartFusion2 FPGA. The product can be adapted to customers' different input / output (I / O) needs. The solution also has a mini PCIe slot for expansion, for example Intel VPU ASICs from the Movidius Myriad series. PCIe generation 2.
iX10100 (ROCm support)
The iX10100 is the second generation SpaceCloud® computer solution based on the 14 nm AMD x86 Ryzen processor and GPU technology in combination with the Microsemi PolarFire FPGA. The product can be adapted to customers' different input / output (I / O) needs. The solution also has a mini PCIe slot for expansion, for example Intel VPU ASICs from the Movidius Myriad series. One of M.2. the slots can be configured as an eNVM high speed port. PCIe generation 3.
Performance comparison
An overview of iX51xx and iX10xxx families’ performance is summarized here for comparison.
| Performance | ||
|---|---|---|
| iX5xxx | iX10xxx | |
| CoreMark v1.0 | 5,842.98 (GCC8.1.0 -O3 -funroll -loops -fgcse-sm -mfpmath =both -DPERFORMANCE_RUN=1 – lrt / Heap) |
25,506.95 (GCC9.2.1 20191102 -O3 -funroll-loops -fgcse-sm -mfpmath=both -DPERFORMANCE_RUN=1 -lrt / Heap) |
| Linpack [GFLOPS] | 4.6 | 54 |
| Clpeak GPU [GFLOPS] | 87 (FP32) | 2000 (FP16) |
| FPGA DSP Cores | 72 (18×18) | 924 (18×18) |
| FPGA heterogeneous interconnect [Gbps] | 6.4 | 12.8 |
| Storage interfaces | 2 x SATA channels v3, 6 Gbps
Optional tailoring, |
2 x SATA channels v3, 6 Gbps (1 port, switeble to eNVM PCIe gen 3 x1)Optional tailoring,eNVM, PCIe gen3 x4 lanes (32GT/s) |
| Optional AI accelerators | Yes (up to 30 Gbps IO bandwidth, PCIe and USB) Multiple Intel Movidius Myriad X VPUs as example |
Yes (up to 50 Gbps IO bandwidth, PCIe and USB) Multiple Intel Movidius Myriad X VPUs as example |
AMD ROCm (for iX10 series)
Run Nvidia CUDA code on SpaceCloud®. Save development time. Caffe TensorFlow Theano MIOpen NCCL HIP
Unibap has adapted AMD’s high performance computing package (HPC), ROCm to the SpaceCloud® product family iX10 and newer.
Note. note that it is not possible to use the basic version of ROCm. Only the Unibap-adapted ROCm for SpaceCloud® can be used.
Here follows an example of converting Nvidia CUDA code with AMD HIP compiler for execution on the iX10 family.
$ hipify-perl square.cu > square.cpp // ROCm “Hipify” an Nivida CUDA code to generic cpp code.
$ hipcc square.cpp -o square_hip // Compile the cpp code with AMD “hip compiler” to either AMD or back to Nvidia.
$ ./square_hip // and finally run it on Unibap’s SpaceCloud® ROCm stack for AMD APU devices.
info: running Square CUDA example on device AMD Ryzen Embedded V1605B with Radeon Vega GFX
info: allocate host mem ( 7.63 MB) info: allocate device mem ( 7.63 MB)
info: copy Host2Device info: launch ‘vector_square’ kernel info: copy Device2Host
info: check result PASSED!