1.5 exaFLOPS

[Klaus]

I think of this more as a publicity stunt than as a serious research project.

[George K]

@Klaus said in 1.5 exaFLOPS:

I think of this more as a publicity stunt than as a serious research project.

Gotta wonder.

By the way, how's that "SETI at Home" thing working out?

[Aqua Letifer]

@George-K said in 1.5 exaFLOPS:

@Klaus said in 1.5 exaFLOPS:

I think of this more as a publicity stunt than as a serious research project.

Gotta wonder.

By the way, how's that "SETI at Home" thing working out?

The amount of data they were able to crunch has been incredible. Never, ever would have been able to do that without the screen saver.

[Klaus]

One thing that is noteworthy here is that one ExaFLOP isn't one ExaFLOP.

[jon-nyc]

Oh yeah? How many is it?

[Klaus]

It's like saying that one engine with 1000hp in one place can do the same things as 1000 engines in 1000 places with 1hp.

[jon-nyc]

I once knocked up three women and had triplets in 3 months.

[Klaus]

@jon-nyc said in 1.5 exaFLOPS:

I once knocked up three women and had triplets in 3 months.

Exactly. Amdahl's law is a good starting point to understand the limitations of parallel computing. It basically says that sequential tasks - like child bearing - cannot be sped up by parallelization. The kinds of problems where the distributed computing architectures like Seti@Home or Fold@Home work at all are called "embarrassingly parallel" problems.

[Copper]

We did some wonderful things with Mr. Amdahl's machines in the 70s and 80s.

I was working at the Amdahl lab in Sunnyvale when they installed their first parallel processor to exceed a Gigaflop. It was a big secret. There were no signs or identification on the machine so nobody would know what it was. In the couple months I was there it mostly sat idle, at least while we were around it.

We were working next to it on the floor but weren't supposed to know what it was. But one of the Amdahl guys told us. The type of stuff we were doing wouldn't have been able to take advantage of it, at least not much.

[Klaus]

@Copper said in 1.5 exaFLOPS:

I was working at the Amdahl lab in Sunnyvale when they installed their first parallel processor to exceed a Gigaflop.

You mean a single processor with multiple cores? Or do you mean multiple processors?

In any case, I thought that a Cray was the first computer to exceed a Gigaflop. What that a Cray?

[Copper]

It was an Amdahl machine, it was their (Amdahl's) first machine capable of a Gigaflop. This was early 80s.

And it was not a single processor, it had many CPs. I don't remember exactly how many, my fuzzy recollection is that it was hundreds, not a few or a dozen. I don't remember them competing with anything Cray made.

I believe it used the same air-cooled processors, the 470, that was current in their mainframes at the time. The 470 was the competition for IBM 370 that ruled the world at that time. Air-cooling was a big deal, the IBM mainframes were water cooled and that added to expense and complexity.