There is a big difference between using some or all US components and buying Cray XC40 supercomputer off the shelf. Only customization being how many cabinets of Cray XC40 compute nodes required and connected through Cray made Aries networking interconnects.
In the past, the Chinese did build, yes
build, their supercomputers with some or all US components but increasingly now using 100% indigenous parts. Cray XC40 is off the shelf supercomputer.
http://pratyush.tropmet.res.in/iitm/compute-info/
PRATYUSH is running on 3315 Intel Xeon Broadwell E5-2695 v4 2.1GHz 18C CPU on its login and compute nodes. It has 2 CPUs per node with a total of 36 cores per node. Each compute node contains 128GB DDR4 .
The system has 16 accelerator nodes powered with Intel KNL 7210 processors , 96 GB DDR4 per node with a total Peak Performance of 42.56 TFLOPS.
XC40 compute nodes are connected using Cray Aries Network On Chip. The four nodes in a blade communicates through the Aries Chip.
Sixteen blade in a chassis are connected through the Aries chip onto a backplane. Each Cabinet consists of three chassis and two cabinets forms a group. All six chassis in a group are connected in all-to-all manner using electrical cables with data transfer rate of 14Gbps. All four groups of the Cray system are connected using optical cables with a data transfer of 12.5 Gbps.
The system runs Cray Linux Environment(CLE) which is Cray’s customised version of the SUSE Linux Enterprise Server 12.2
CLE consists of two components: CLE and Compute Node Linux (CNL). The service nodes, external login nodes, and post-processing nodes of Cray XC40 run a full-featured version of Linux.
Edit:
Btw I am not claiming anything for China, just stating the facts.
If you read my past post, its rare I ever claimed anything for China, only for Singapore.
I do look out for Indian braggarts though, which don't include you.
Article did not claim its made in India, only insinuated by Indian fanboys, but did give misleading power of 6.8 petaflops for 2 machines, instead of just 4.0 petaflops for the more powerful m/c.
The XC40 can easily exceed these values by adding more compute nodes, i.e more $$.
.