MLU270-S Series Intelligent Processing Card User Manual V1.0.0 Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual Revision History Document Name MLU270-S Series Intelligent Processing Card User Manual Revision Created Date Revision History V1.0.0 2019.08.20 Date Revision No. 2019.08.20 V1.0.0 Notes First Release 0 1 2 3 4 5 Technical Supports contacts: service@cambricon.com www.cambricon.com Content Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 1. Product Brief...................................................................................................................................4 2. Product Specifications...................................................................................................................6 2.1 Performance specifications.................................................................................................6 2.2 Software specifications........................................................................................................6 2.3 Operating Environment...................................................................................................... 8 2.4 Mechanical Specifications...................................................................................................8 2.5 Package Size and Weight................................................................................................... 9 2.6 Thermal Specifications........................................................................................................ 9 2.7 Power Supplies Specifications..........................................................................................12 3. Cambricon NeuWare Development Environment................................................................... 13 4. Compliance.................................................................................................................................. 14 Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 1. Product Brief Figure.1 MLU270-S Series Intelligent Processing Card Specially designed MLU270-S series dedicated to AI inferences and data center accelerations with high EER (Energy Efficiency Rate) The SIYUAN 270 ASIC has been designed based on several innovative technologies in the architecture of processors owned by Cambricon. Along with it, many up-to-date features have been integrated into a standard HHHL PCIe card, which can be inserted into a modern AI PC/server so as to provide a huge extension of calculation power of AI inferences. MLU270-S series has a moderate TDP of 70W and it can provide a calculation power as high as 4 times of the previous generation. It can be widely used in applications such as vision, voice, natural language processing, legacy machine learning, and many other AI scenarios, and it can be used in an AI inferences platform to make it work with even higher EER. Brand-new Cambricon MLUv02 Architecture The MLUv02 architecture is not just a simple update from previous generation, but a brand new design based on NOC (Network on Chip), which will guarantee the parallel efficiency of execution of 16 NPU cluster within the SIYUN 270 ASIC. The dataflow will be compressed within the chip by dedicated hardware engine and this will increase the volume and bandwidth effectively. The new architecture can fully support all the AI accuracies such as INT16, INT8, INT4, FP32, FP16, and provide necessary calculation power to many kinds of Neural Network. In a word, the new architecture can provide both good universality and best performances to customers at the same time. One more step on the performances of inferences When using INT8 accuracy for AI inference calculations, the performance of non-sparse networks is improved up to 4 times as much as that of the previous generation. MLU270-S series can provide a great EER as high as 40 times of a normal CPU. There are embedded newly designed code/decode hardware for videos and pictures, so when the system have to handle this kind of tasks, MLU270-S series will decrease the load of CPU and the occupation Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual for PCIe bandwidth, so as to increase the performances even more. Calculating flexibility and programmability SIYUAN 270 ASIC can provide supports to many kinds of Neural Networks. Cambricon NeuWare software stack can easily deploy inference environment.BANG Lang.CHENG environment can be directly customized for computing resources meeting all kinds of AI customization requirements,professional but not specialized. Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 2. Product Specifications 2.1 Performance specifications Table 1 MLU270-S Series Intelligent Processing Card Hardware Specifications Card Type Processor Architecture Core Clock Peak INT8 Calculation accuracy Video Decoding Support Total Memory Size Memory Bus Width Memory Bandwidth System Interfaces PCIe ID Outline TDP ECC Supporting Heat Dissipation MLU270-S4/V4/D4/S4K/V4K/D4K Cambricon MLUv02 MLU270-S4/V4/D4:1 GHz, MLU270-S4K/V4K/D4K:1.2 GHz MLU270-D4/D4K 64TOPSDense, MLU270-S4/V4 128TOPS (Dense) MLU270-S4K/V4K 160TOPS(Dense) INT16, INT8, INT4, FP32, FP16 Yes 16GB 256-bit 102.4 GB/s PCI Express Gen3, x16, Supporting Lane Reversal PCIe Vendor ID PCIe Device ID PCIe Sub-Vendor ID 0xCABC PCIe Sub-System ID 0x0012 68.9mm * 167.5mm, Single Slot 70W Yes Passive 0xCABC 0x0270 2.2 Software specifications Table 2 MLU270-S Series Intelligent Processing Card Software Specifications PCIe Base address PF 164bit:
BAR0: 256MB prefetchable BAR2: 64 MB prefetchable BAR4: 64 MB prefetchable VF464bit BAR0: 256MB prefetchable BAR2: 64 MB prefetchable BAR4: 64 MB prefetchable 0x8E(Write 0x8FRead SMBus8bit Address SMBUS Register is 32-bit wide, and below describes how to read a registerS:SlaveM:Master Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual Direction M S M S S M M S S M M S M S S M S M M S S M M S S M M S S M M S M S Registers Total Card Power Card Temperature Chip Temperature Power Brake Table 3 SMBUS Registers Reading Bits 1 8 1 8 1 1 8 1 8 1 8 1 8 1 8 1 1 Content START SLAVE ADDRESSWrite ACK REGISTER ADDRESS ACK RE START SLAVE ADDRESSRead ACK DATA[7:0]
ACK DATA[15:8]
ACK DATA[23:16]
ACK DATA[31:24]
NACK STOP Table 4 SMBUS Registers Description Address ACCESS Description 0x01 0x02 0x03 0x05 RO RO RO WO Card Power consumption, Float Data, Unit W Card Temperature, Float Data, Unit Chip Temperature, Float Data, Unit The main frequency is reduced to 25% of current frequency when write 0x04,and restore to pre-frequency when write 0x01
[15:0] Vendor ID :0xCABC
[31:16] Device ID:0x0270
[15:0] Sub-Vendor ID:0xCABC
[31:16] Sub-System ID,for example 0x0012 Display device typee.g 0X12 means device type is S4 Display vendor name Display hardware revision Display Firmware Revisione.g 0x02120016 means the main chip version is 0x02,device type is 0x12,main version is 0x00,subversion is 0x1,patch number is 0x6 Dispaly manufacturing time,e.g 0x1811 means manufactured in 2018.11 Display Device IDe.g 0x20023 means device id is 20023 PCIE Vendor ID and Device ID PCIE Sub-Vendor ID and Sub-System ID Device Name Vendor Name Hardware Revision Firmware Revision 0xA0 0xA1 0xF0 0xF1 0xF2 0xF3 Manufacturing Time 0xF4 Device ID 0xF5 RO RO RO RO RO RO RO RO Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 2.3 Operating Environment Table 5 MLU270-S Series Intelligent Processing Card Working Environment Operating Temperature Storage Temperature Operating Humidity Storage Humidity 045C
-4075 5%95% Relative Humidity 5%95% Relative Humidity 2.4 Mechanical Specifications Figure. 2 MLU270-S Series Intelligent Processing Card Dimensions (High bracket) Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual Figure.3 MLU270-S Series Intelligent Processing Card Dimensions (Low bracket) 2.5 Package Size and Weight Type Single Card Gift package with single card Weight 310g 760g Size 167.5mm*68.9mm*18.1mm 385mm*230mm*97mm Table 6 MLU270-S Series Package Size and Weight Notes NA Including gift package,MLU270 series card,standby panel,PE bag,desiccant 16 cards each box 7.7kg 600mm*400mm*253mm Industrial Box Gift Box Notes:The weight is a measured value with a tolerance of +-10%
600mm*400mm*253mm 5.6kg 6 gift packages each box 2.6 Thermal Specifications 2.6.1 MLU270-S Power and Temperature Specifications Table 7 MLU270-S Series Thermal Design Specifications Item TDP(Thermal Design Power) Recommended Operating Tj Slowdown Tj Slowdown rate Shutdown Tj Parameters 70W 0-85 87 50%
95 2.6.2 MLU270-S Series System Pressure Drop The figure below is got by measurement :
Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual Figure.4 MLU270-S System Pressure Drop Table 8 MLU270-S Heatsink Airflow Rate (CFM) Pressure Drop (mmAq) 3 3.5 4.3 5.4 6.7 8 5.1 7.0 10.5 16.6 25.6 36.5 2.6.3 MLU270-S Series Airflow Directions Support MLU270-S series can be placed at the front and rear part of the chassis, and the airflow can be in either directions as below:
Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual Figure.5 Airflow Directions Support 2.6.4 MLU270-S Series Operating Temperature and Airflow Requirements MLU270-S Series can support operatingTDP at Temperature of 0-50 (Temperature at the entrance of heatsink). Refer to below table for more information about minimum Airflow requirement Table 9 MLU270-S Temperature of Ambient vs Min Airflow Requirement Air Temperature at Card Inlet Min Airflow Requirements CFM) 30 35 40 45 50 55not recommend 3 3.5 4.3 5.4 6.7 8 Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 2.6.5 MLU270-S Average Temperature of Inlet vs Min Airflow Requirement Figure.6 Air Inlet Temperature vs Min Airflow Requirement 2.7 Power Supplies Specifications Table 10Input Voltage Requirement Power Supply PCIe edge connector12V PCIe edge connector3V3 Min 11.04V 3.0V Normal 12V 3.3V Max 12.96V 3.63V Table 11 Current Requirement Power Supply Peak Current Moving Average PCIe edge connector12V 20A 17A 13A 200us 1ms 5ms Table 12 Power Capping Power Capping Threshold Power Capping Response time typical Power Capping Response time max Table 13 Power Brake PB# PCIe pin assignment Power Brake response timetypical PB# input insertion low time min Power brake hardware slowdown factor 70W 50ms 100ms B30 150us 250ms 4x Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 3. Cambricon NeuWare Development Environment NeuWare can fully support all kinds of mainstream programming framework, such as TensorFlow, Caffe, PyTorch, MXNet and so on.With above mentioned Programming Frameworks, users can easily and conveniently develop and deploy their Deep Learning Applications on Cambricon MLU270-S Series Intelligent Processing Cards.At the same time, NeuWare provides complete runtime system and driver software to speed up the system integration procedure. Learning Library, Runtime NeuWare also provides a full set of software tools such as Application Development, Function Debugging and Performance Optimization. The Application Development Tools include Machine and Domain-Specific (for example Video Analysis) SDK; The Function Debugging Tools can fulfill all the requirements from different levels of Programming Framework and Function Library; The Performance Optimization Tools for performances analysis and system monitoring. Library, Compiler, Model retraining tools include tools Figure.7 Cambricon NeuWare For more information,please visit www.cambricon.com Tel: 86-10-83030003 E-mail: business@cambricon.com Address: 11/F, Block D, Zhizhen Tower, No.7 Zhichun Road, Haidian District, Beijing, China Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual 4. Compliance The MLU270-S Series is compliant with the regulations listed in this section. Compliance marks, including the FCC ID numbers, can be found on the labels of each devices. United States Federal Communications Commission (FCC) This device complies with Part 15 of the FCC Rules. Operation is subject to the following two conditions: (1) This device may not cause harmful interference, and (2) this device must accept any interference received, including interference that may cause undesired operation. This equipment has been tested and found to comply with the limits for a Class B digital device, pursuant to Part 15 of the FCC Rules. These limits are designed to provide reasonable protection against harmful interference in a residential installation. This equipment generates, uses and can radiate radio frequency energy and, if not installed in accordance with the instructions, may cause harmful interference to radio communications. However, there is no guarantee that interference will not occur in a particular installation. If this equipment does cause interference to radio or television reception, which can be determined by turning the equipment off and on, the user is encouraged to try to correct the interference by one or more of the following measures:
Reorient or relocate the receiving antenna Increase the separation between the equipment and receiver Connect the equipment into an outlet on a circuit different from that to which the receiver is connected Consult the dealer or an experienced radio/TV technician for help Underwriters Laboratories (UL) UL Listed Product Logo for MLU270-S Series Intelligent Processing Cardsmodel name MLU270-S. Cambricon Technologies, MLU270-S Series Intelligent Processing Card User Manual