Previous image Index Next image
Index Previous image Next image Start slideshow (2 s)

Image 28 of 35



The original supercomputer, the Cray-1, was set up at Los Alamos National Laboratory in 1976.
The Cray-1 was capable of 80 MFLOPS (or, according to another source, 138 - 250 MFLOPS).

// --- Wiki Note: FLOPS is an acronym meaning ''FLoating point Operations Per Second''.
It is a measure of a computers performance.
A couple of things to keep in mind when talking FLOPS though:
Often, it is a rather theoretical assessment of a computers power. In the ''real'' world,
system buses, RAM type, and I/O speeds, etc. all effect the overall system performance.
Performance-per-watt, MIPS (Millions Instructions Per Second) etc. should also be considered.
Nevertheless, a FLOPS measurement will gives us a rough estimate of a computers performance.
Something we can use to compare computers with! --- //

The Cray - 1 was of course for years and years
the best known and most successful supercomputer.

Even today, we have conversations where people say something along the lines of:
''Well, well, very good, but still not a Cray computer ...''

Well, how good was the 80 MFLOPS Cray - 1 actually? Configured with its 1 million word RAM and all.
That Cray thing that weighed 5.5 tons (including the freon refrigeration system).
That Cray thing, that consumed about 115 kW of power;
(Cooling and storage doubling the figure) ...

After seeing the Cray 1 in the museum. Breathing in the ultimate, uber cool computer of all times ...
(Back home) I decided to run some statistics on my office PC. Just to compare.
// --- I do realize that it is as fair as comparing apples and oranges, but still, it was
kind of shocking to see how (some values from) a measly office computer of 2009 compares
with the uber computer of all times... ---//

I downloaded CPUBench and ran it on my office computer to get some values.

CPUBENCH on my office windows PC October 2009

And the measly office PC wins the day ...At 382 MFLOPS, its nearly 5 times faster than the CRAY.

CPUBENCH on my office windows PC October 2009

But, back in the 80s me and my friends were of course convinced that the CRAY 1 computer was just awesome
and uber cool like nothing else.
And pretty close to being comparable to a human brain.....


Well, how wrong we were...

Sure, there is no consensus regarding the computational power of the human brain.
Still, a number of experts have suggested that the brain's processing power is around 100 trillion neuron updates per second.

Using that estimate, it is expected that the first unoptimized simulations of a human brain in real time will
require a computer capable of 10^18 FLOPS (= 1 exa FLOPS= 1000 Peta FLOPS = 1.000.000 Tera FLOPS = GFLOPS).
// --- I.e. If the brain is capable of 100 petaflops, (most people thinks that) you need 10 times more to run a simulation of it. --- //

Lets just say its a lot more than the Cray....

About the numbers (for the brains processing power):
Hans Moravec of Carnegie Mellon University started by estimating the human brain to be doing app. 100 million MIPS.
Then Nick Bostrom of Yale estimated it to be 1,000 times higher than Moravecs estimate....
And recently Ray Kurzweil have estimated the computing capacity required for a human brain -functional- simulation to
be in the 10^15 - 10^16 FLOPS range (chart).
Full brain simulations (in the exa range) to be reached by super computers in the 2020s (Singularity is near), perhaps earlier,
according to Kurzweils predictions in the singularity book.

The race is on. Japanese politicians have decided to fund the construction of a 10 petaflop computer system by 2011.
A lot faster than current recordholder Blue Gene.
// --- Blue Gene: IBMs private partnership with a Swiss Institute, to emulate the activity of brain neurons. Appropriately called Blue Brain.
Is currently reaching sustained speeds of 500 Tera FLOPS (Blue Gene) ---//

The latest Cray machine - the Cray XT5 - is doing 1.3 petaflops (Nov. 2008).
And the fastest (Oct. 2009) super computer now is the IBM road runner at 1.7 Peta FLOPS.
Busy modelling the decay of the US nuclear arsenal.
Perhaps a more reasonable thing to do than Mind Uploading?

October 2009

Pics from visit to the Computer Museum, Mountain View, CA. - Visible Storage display items.
And my own graphs of course ..