Tell me some use case in ANY field of science which can be done in say the fastest supercomputer (which is Tianhe -2 ?) but not in the second or third fastest one ? If the first one take an hour probably the second fastest one will take minutes more. At those humongous scale of processing power record books hardly matter rather we are investing in engg streams which specialize in leveraging those computing power in diff aspects of everyday life !
There are none.
Like it or not, the supercomputer race is, to put it bluntly, more about bragging rights than for any scientific purposes outside of extremely esoteric subjects like hurricane analyses, nuclear explosion simulations, or calculating earthquake propagation. Here is the kicker -- designing a 'stealth' aircraft does not require a supercomputer.
» Supercomputing vs. Distributed Computing: A Government Primer
The advantage of supercomputers is that since data can move between processors rapidly, all of the processors can work together on the same tasks. Supercomputers are suited for highly-complex, real-time applications and simulations.
The phrase 'real-time' does not mean at the immediate moment. It mean calculating and predicting relationships between factors as they occurs, whether those relationships are simulated (at anytime) or is actually happening now like the real hurricane bearing down on your position. Obviously, we cannot detonate a nuclear explosive device just to observe, experiment, or take measurements. So simulation is the next best thing and the more accurate the data inputs, the more accurate the simulation. Same for weather phenomenons.
There is a difference between a 'supercomputer' and 'supercomputing' capabilities. Regarding the source above, distributed computing is a 'supercomputer' of sorts, or rather it is an aggregate of many many 'weaker' contributors into a 'supercomputing' architecture.
For our 'stealth' aircraft, can a 'supercomputer' help ? Absolutely. The more complex a body, the greater the interactions between many diffracted/reflected signals from many structures. But unlike a nuclear explosion or a hurricane, this complex body can be broken up into many discrete smaller structures where those interactions are lesser in complexity and quantity.
For example...We can assign a 'regular' computer to calculate the diffracted/reflected signal paths off a wing, mathematically 'freeze' those paths, and virtually plug them into the fuselage to see how those signals affects the fuselage. The 'supercomputing capabilities' of distributed computing is better financially speaking. That does not mean that 'regular' computer is anything from your local Joe's PC Shack. That 'regular' computer is custom designed and can run into the four, five, or even six figures price range.
Further...Unlike a nuclear explosion or a hurricane which are unitary events, an aircraft is a composite of many events, such as the fuselage, the wings, or the cockpit. The items are often finished at different times throughout the entire project, then assembled together. So a 'supercomputer' to analyze the small UHF antenna's RCS is overkill when a 'regular' computer will do just fine. The 'supercomputer' can do the job in a couple milliseconds, of course, but then the UHF antenna will be sitting on the shelf waiting for final assembly on the fuselage.
So take these contests with a grain of salt.
why wait for Volta when Pascal got the power now 
