Can someone explain the basic architecture of Super computer?
I am not a CS or EE person but can add some additional comments to Concard and SKC's comments
1. Regular computer has 1 CPU (central processing unit), and attached memory (RAM, and various cache memories - L1, L2 etc that have much faster access to the CPU each way.
2. You can only make this so fast by (a) increasing clock speed, (b) playing with CPU architecture - design and also CISC, RISC
3. In the past few decades Intel (& other) CPUs also have many "cores" - that many computing units on the same chip - perhaps sharing L1, L2.. cache and memory...
Also may have associated GPUs (graphic processing units) in order to offload graphics load off the main CPUs
4. Multi-cores only work when the task can be split into several parallel tasks - and the results are then perhaps combined to give the output.
5. In Supercomputers - you now add many many CPUs - so basically a processing array that process massive computations in parallel. They may have independent or shared memory (place where the program, data and intermediate results are temporarily stored). It only works when the task can be done in parallel -
examples are solutions of partial differential equations which ultimately boil down to inversion of a matrix - either full or a "band" matrix. There are probably other examples ..
Needed for this in addition to hardware wiring is the architecture and software (e.g. for CPUs and memory to work to be used fast and effectively) -
so architecture, memory and cpu interface and software (the operating system) are needed to build this thing. India has plenty of expertise to build these - for tasks that may be needed. CPU and memory and interconnects are available off the shelf
----
This is my understanding - a decades ago Chemical Engineer - who has never worked in EE/CS - except for that one programming course on the Ganga plain
But I am pretty sure that I am close to what it is