Japanese team discovers 'super disc' material

A Japanese research team has found a material that could be used to make a low-price super disc with data storage capacity thousands of times greater than a DVD, the lead scientist said Monday.....

WCG Scientist should start also to think about ways to design projects that can make use of such crunching power. When this power will be available for supercomputers, mere mortals like us will probably have 10 or 20 times more crunching power available.

As an example the fastest supercomputer in the world has a capability of a sustained 1700 TFlop/sec crunching power.
Let's suppose that we take the NRW project. It finished at 20 billion points.
Now 0.7 Million points equal 1 TFlop. With 1700 Tflop you finish the whole project in 17 seconds.
FAAH is approaching 60 billion points. That is one minute.
Ok. I know all these numbers do not relate as easily. The supercomputer is benchmarked with Linpack, and WCG projects do not relate in terms of calculations so easily to those benachmarks. Calculations are not always on flops but integers etc. And the ratio 0.7Million points = 1TFlop is also for sure approximate.
Nevertheless the example gives orders of magnitudes which gives us a glimpse of what will be possible in the future. As crunchers will update their systems the collective crunching power will increase. We are at 400 TFlop/day but in two or three years we should be over the Petaflop for sure.
And GPU crunching can also give a terrible boost to existing systems immediately today.

All this is excellent news for humanity, but scientist will be under pressure to use all that power the best possible way.

Let's see hum.....

If we agree on the 1TFlop = 700'000 points

1700 TFlop/sec X 0.7 million points = 1.19 billion points/sec.

20 billion points / 1.19 billion points/sec = 16.8 sec

Seems right. But Sgt.Joe please do not hesitate to tell me were the error is.

20,000,000,000 / 700,000 = 28,751.43 TFL / 1,700 TFL sustained = 17 days

Would the 400 TFL WCG machine exclusively do RICE, than 71 Days or 2.37 months.

Test:

20,000,000,000 RICE points / 292,000,000 points WCG daily last Monday = 68.5 days it would at this power take on our grid.

I.e. 17 days it would take on the fastest computer we know of. In the context of WCG that makes sense, mi penso.

Calculus may get rusty... my Omron is too.

Sek the supercomputer linpack benchmarks are of 1'700 TFL or 1.7 PETAFLOP per SECOND not per day.

So I think here lies the problem. WCG is measured in TERAFLOP/DAY the supercomputer performance is in TERAFLOP/SECOND.

If we talk about the same TERAFLOP definition and that is One Thousand Billion Floating Point calculations, then the question is the timeframe. The supercomputer is measured PER SECOND.
For WCG I suppose we measure PER DAY. So in a day you have 86'400 seconds. We have to take this into account.

Just to be very precise, we use the term MEGAFLOPS and TERAFLOPS not allways rightly. It is MEGAFLOP and TERAFLOP. MEGAFLOPS is a shortening for MEGAFLOP/second and same for TERAFLOPS.

So when we say a 400 TERAFLOP machine. This means the machine can do 400 TERAFLOP. But this means nothing about its effective computing power. How long does it take to reach 400 TERAFLOP, a day? 30 years?. A 400 TERAFLOPS machine means 400 TERAFLOP/second capable which is a precise capability.

The fastest supercomputer in the world can crunch in a day 146'880 PETAFLOP if he can work in a sustained mode.

I think the error lies in the 700'000 points = 1 TFLOP. It is a relation of points to TFLOP. This is probably very wrong. We should completely review this.

As an example my Boinc benchmark gives a total rounded of 144 GFLOP/sec on my i7 980X at 4Ghz. I have 11 of them that makes an aggregate 1'584 GFLOP/sec or rounded 1.6 TFLOP/sec. This means in a day if I generate 600kpoints then the ratio would be
1.6 TFLOP/sec X 86400 sec = 135'360 TFLOP = 600'000 points

The ratio is then 1 TFLOP = 4.4 WCG points

This is probably a much more realistic estimate.