"NVIDIA today released the upgraded GPU computing server" dgx-2h ", which is equipped with up to 16 Tesla V100 GPUs like the previous generation dgx-2, but the thermal design power consumption is open from 350W to 450W, and the performance is improved.
Tesla V100 is by far the most powerful GPU computing card. Its core is manufactured by 12NM process. It integrates more than 21 billion transistors with a record area of 815 mm2. Its architecture is based on Volta. It has 5120 fp32 single precision floating-point cores, 2560 fp64 double precision floating-point cores, 640 tensor cores and 16 / 32GB hbm2 video memory.
Lao Huang shows Tesla V100
The first generation computing server dgx-1 based on Tesla V100 has only eight cards, which is equipped with dual Xeon e5-2600 v4. The second generation dgx-2 has 16 fast cards. At the same time, all aspects are upgraded. For example, it is equipped with dual Xeon platinum 8168 24 core processor.
Dgx-2h is still 16 Tesla V100, all of which are 32GB hbm2 video memory, with a total of 512gb. With the release of thermal design power consumption, the frequency has increased (specific unpublished), and the tensor calculation performance has been improved from 1.92pflops to 2.1pflops (amplitude about 9%), but the semi single double precision performance has not been published.
The shape of dgx-2h is basically unchanged
internal structure
In addition, the new system processor is upgraded to two 24 core Xeon platinum 8174 (not listed in the official public model or customized version), and the network is upgraded from dual 25gbe to dual 100gbe, with a maximum memory of 1.5tb DDR4 and storage of 30-60tb nvme.
The power consumption of the whole machine has increased from 10 kW to 12 kW, and the weight has increased by 20 pounds (about 9 kg), which is likely to be a new cooling system.
The price did not say, but it must be higher than the previous generation's $399000.
Specification comparison of the third generation Tesla V100 server
Dgx-2h / 2 specification comparison, technical zone
NVIDIA released an upgraded version of GPU computing server dgx-2h. The thermal design power consumption is open from 350W to 450W
Shasta, Cray's new generation supercomputing platform, adopts 1U specification and has 8 amdromepyc built in
With the advent of black five, Intel CPU began to return to the recommended retail price for listing
Desktop cascade lake-x will still evolve on the basis of old technology and old architecture. There is no surprise
Intel will reduce the number of processors in the DIY market by up to 2 million in Q4. Motherboard and graphics card manufacturers will download enthusiast app
Create an electronic circle of your contacts
Pay attention to wechat of electronic enthusiasts
Interesting and informative information and technology dry goods
Focus on enthusiast class
Lock in the live broadcast of the latest course activities and technologies and collect them
0 collections
Share:, comment
Lin Chaowen PCB Design: pads tutorial, pads video tutorial, Zheng Zhenyu teacher: Altium designer tutorial, Altium designer video tutorial, Zhang Fei actual combat electronic video tutorial, Zhu Youpeng teacher: Hisilicon hi3518e tutorial, hi3518e video tutorial, Li Zeng teacher: signal integrity tutorial, high-speed circuit simulation tutorial, Huawei Hongmeng system tutorial, harmonyos video tutorial, saisheng: EMC design tutorial, EMC video tutorial Mr. Du Yang: STM32 tutorial, STM32 video tutorial, Tang zuolin: basic C language tutorial, basic C language video tutorial, Zhang Fei: Buck Power tutorial, buck power video tutorial, punctual atom: FPGA tutorial, FPGA video tutorial, Mr. Wei Dongshan: embedded tutorial, Embedded video tutorial Zhang Xianfeng: C language basic video tutorial Xu Xiaogang: Modbus communication video tutorial Wang Zhentao: NB IOT development video tutorial mill: FPGA tutorial, zynq video tutorial c language video tutorial linux driver development video tutorial Zhu Youpeng: u-boot source code analysis video tutorial harmonyos,
Press and hold the slider and drag to the far right
Learn about new features
Published, relevant recommendations
Ten years of electric competition, hand in hand - Texas Instruments China Education in 2018
The 2018 annual meeting of Texas Instruments Chinese educators was successfully held in Wuhan a few days ago. President of Texas Instruments (TI) global educational technology, education and enterprises
Published on November 21, 2018 at 11:55
•
248 readings
How does the Intel Q33 express chipset enable 2D and 3D?
Published on November 21, 2018 at 11:47
•
7 readings
I7-7820x vs i7-8086k for gaming / streaming?
Published on November 21, 2018 at 11:43
•
6 readings
Introduction of design example of deep learning processor system
On January 9, 2010, Yi Shan of deephi in the edge track of XDF 2018 in Frankfurt
Published at 06:16, November 21, 2018
•
0 readings
Introduction to successful cases of artix-7 100t FPGA
See why Cobham chose to use the Artix-7 100T FPGA ...
Published at 06:14, November 21, 2018
•
6 readings
The market needs photoelectric sensors with higher integration
Published at 16:56, November 20, 2018
•
10 readings
ATMEL ARM7 flash microcontroller has a density of 64KB
Published at 16:50, November 20, 2018
•
14 readings
The Matthew effect in the chip industry intensifies, and TSMC takes the lead in occupying the high point of 7Nm
Not long ago, amd released the world's first 7 nm process CPU and GPU in Beijing, which attracted wide attention. 7Nm is known as semiconductor process
Published on November 20, 2018 16:11
•
150 readings
Cray's new generation supercomputing platform Shasta adopts 1U specification
Amd Rome, Intel cascade Lake... Two next-generation high-performance computing platforms are coming, especially
Published at 15:58, November 20, 2018
•
274 readings
In 2018, Apple will launch a new version of MacBook Air, which can
In addition to lightweight, magnesium lithium alloy has better dissimilarity, ductility, easy forging, and stronger strength than magnesium alloy. Anli material 20
Published in 2
Our other product:
