back to article IBM US nuke-lab beast 'Sequoia' is top of the flops (petaflops, that is)

For the second time in the past two years, a new supercomputer has taken the top ranking in the Top 500 list of supercomputers – and it does not use a hybrid CPU-GPU architecture. But the question everyone will be asking at the International Super Computing conference in Hamburg, Germany today is whether this is the last hurrah …

COMMENTS

This topic is closed for new posts.
Anonymous Coward

I dare someone to ask whether or not it could run Crysis and think they are being even slightly amusing.

2
2
Silver badge
Joke

Don't be silly

Crysis is not available on BlueGene

3
0
Anonymous Coward

Why settle for simply running Crysis?

With that much computational power, it should be able to simulate several machines running Crysis at the same time.

1
0
Silver badge
Happy

Re: Don't be silly

Ok then- can it run Metro* and do more than two tasks?

/ducks and runs

* In a VM- I'm not that optimistic...

0
0
Anonymous Coward

I'm not a tree hugger or anything but...

...it's just a bleeding shame they put all this CPU grunt into running nuclear weapon simulations. If it turned its attention exclusively to say something like medical research (e.g protein folding etc..) just think of the benefits to humanity? Just saying.

0
0
Anonymous Coward

Re: I'm not a tree hugger or anything but...

Yeah, you're right. They should just test fire the nukes.

5
0
Anonymous Coward

Re: I'm not a tree hugger or anything but...

Sadly, the two uses for these extremely powerful clusters (that I can remember) are for simulating nukes and the weather. Neither of which seems to have much lasting benefit to humanity. Even hurricane tracking and forecasting is shoved off to state university grade HPC clusters. However, BOINC--with a heavier focus on math, astronomy and medicine--seems to be doing rather well with enough processing power to put it in 4th place.

0
0
Thumb Up

Re: I'm not a tree hugger or anything but...

Yahhhhhh!!!!!!! Meet Jesus, Be God!!!!!!!!!

MEGA - FUCKING - NUKES!!!!!!!

Just imagine that - drilling a 10,000 meter deep hole into Yellowstone Park and dropping a 100 megaton fusion nuke down it, with a million gallon tank of deutrium on top!!!!!!!!

"I am greater than god!!!!!!!!!!!!!

1
2

That's silly number

"...running across the 1.57 million PowerPC cores inside the box."

"I always try and believe 6 impossible things before breakfast", said Alice.

1
0
Anonymous Coward

Re: That's silly number

I'll just go round it off with breakfast at Milliways, then.

1
0
Childcatcher

Is it me??????

Or does anyone reading this feel the same as when they watch Clarkson on top gear testing a car that 0.0000000000001% of the population will ever own or need? I for one have never experienced under-steer on the M25 or the need to calculate how quick the world could go fooom after pressing the appropriate buttons.................

2
4
JLH

Re: Is it me??????

I don't agree with you.

The techniques used to build, manage and run jobs on beasts like this filter down to the HPC machines used by 'normal' business and academia.

4
1
Bronze badge
Windows

Re: Is it me??????

Plus I take it you have never experienced the deep frustration of knowing exactly how to calculate something (perhaps vaguely useful) but realising that it would take something like 9 x 10^7 seconds to actually run the calculation on the machines you have available...

0
0
Anonymous Coward

Bar charts, we've heard of them

If only there were some graphical way to compare the various machines specs.

2
0

Uses ZFS

This IBM supercomputer uses Lustre + ZFS. The reason ZFS was chosen is because it scales better, provide data integrity, etc.

http://www.youtube.com/watch?v=c5ASf53v4lI

http://zfsonlinux.org/docs/LUG11_ZFS_on_Linux_for_Lustre.pdf

I wonder why they did not choose the IBM GPFS filesystem.

0
0
Bronze badge

What About Latency?

While the article notes that several things besides the raw number of floating-point operations per second is relevant to rating the real performance of a supercomputer, one thing that wasn't mentioned was latency - or single-thread performance. Not everything can be parallelized well.

0
0
Coat

Larry Ellison's wet dream

When they decide to put an Oracle DB on it and work out the license cost based on a per core calculation.

0
0

1.57 million PowerPC cores

... for when 1.56 million PowerPC cores is just not enough

0
0

@ John Savard

"one thing that wasn't mentioned was latency - or single-thread performance. Not everything can be parallelized well."

If your problem doesn't parallelize well and single thread performance is your main concern then you would buy a different (and much cheaper) machine.

0
0
This topic is closed for new posts.

Forums