I'm actually more worried about the declining number of Chinese supercomputers on the list. It could mean three things, the rivals are investing more aggressively than China, China is waiting for homemade processors before going all out, or China's investment on supercomputer is declining and only aiming at superficial status like the number one spot. I hope it is either the first reason or second.
To me, access to the supercomputers for the scientists all over the nation is as important as the computing power itself. After all, the supercomputer is only as good as the users. I hope Chinese scientists all over China who need access to the supercomputer to do the research can get them, and not only selected few with connection.
I read a post by someone who actually got access on TH-2, and he said that the machine is not fully utilized. The problem is software -- off the self software won't work -- it has to be specially modified to take advantage of the speed, so unless you develop the software yourself and/or have access to the source code to modified it, you cannot really make use of it.
-- actually, it's worse. It's not fast enough to make porting of software, even if it is available, to be worthwhile. Neither is developing software for it, because imported supercomputers are faster.
Nothing more than a waste of money -- by the time they develop the computer, it is already hopelessly behind in technology. The money would be better spend buying foreign supercomputer and develop home grown software for it.
with the fastest machine having speed of 20 petaflop.
By that time, though, the fastest supercomputer is expected to have speed exceeding 1000 petaflop (1 exaflop).
The fastest machine made in India now is PARAM Yuva II, which is about 0.38 petaflop. Top put it in perspective, TH-2 is 33 petaflop -- so native Indian supercomputer is 1-2% as powerful as TH-2.
Note also 20 petaflop compare to 1000+ petaflop is also around 1-2%. So you can say India is planning to stand still and not falling further behind.
