The UK's Fastest Supercomputer

bmsleight writes "The Guardian has a story on the HECToR, The largest supercomputer in the UK — around five times more powerful than its predecessor, HPCx, which is also at the University of Edinburgh. It measures up well internationally, sitting at 17 in the top500.org list of the most powerful computers in the world."

Re:17th isn't good enough

[Ed Avis]

I don't think research is a pissing contest to see who has the most powerful computer.

The metric for 'most powerful' also seems flawed. If you just count operations per second, then a large enough cluster of Linux PCs will appear 'more powerful' than any supercomputer, even if they are connected by UUCP over 2400 baud modems. Yet the supercomputer is much faster at most difficult computational tasks because it has faster connections between the nodes. The Linux cluster would only outperform it for drawing a large picture of the Mandelbrot set or other 'embarrassingly parallel' problems. This isn't an academic distinction; even gigabit Ethernet is much slower (higher latency) than the links used in a real supercomputer.

top500.org do categorize each supercomputer as 'cluster' or whatever but I think their ranking is just on raw operations per second.

Cray XT4 Supercomputer

[icehawk55]

The article really didn't say much about HECToR itself. It's a 60 cabinet Cray XT4 system that currently has over 5500 AMD dual core processors. We'll be upgrading it in stages over the next couple of years to over 250 Teraflops. Including some cabinets of the new Black Widow Vector product, now called the Cray X2 system. The Cray team, myself part of it, is actually a multinational effort. I'm a US citizen who is headed over to maintain the system, we have a Brit on the team and the third is also from outside the UK. It's an interesting situation. The biggest UK system, being maintained by two expats and a local. (-: ice_hawk55

Re:But does it know...

[DMoylan]

are you thinking of hactar, designer of the ultimate weapon?

http://hhgproject.org/entries/hactar.html [hhgproject.org]

top500 uses linpack to measure performance.

[flaming-opus]

Your example of the 2400 baud modems for a linux cluster isn't completely accurate, as linpack does a little bit of communication, though the point is well taken. The top500 list only uses linpack to measure performance, and linpack represents a very easy problem to solve. Essentially, the top500 list is a list of which machines do a really good job of solving a trivially difficult problem. The hpcc benchmarks (http://icl.cs.utk.edu/hpcc/) are a lot more interesting; though, even these need to be read with some caution.

Ranking supercomputers is a really hard problem. Each application has different needs for communication latency, bandwidth, programming model, cache size, memory bandwidth, and computational throughput. Then you have to ask: how much optimization can I do to the benchmark? Am I going to be able to do the same amount of optimization for each of my applications? How easy is it to extract this performance? The guys writing the software for these things are usually professors or post-docs in the hard sciences, not in supercomputing.

HPCx is not at the Uni of Edinburgh

[Anonymous Coward]

HPCx ain't at the University of Edinburgh. It's partly funded by them but is actually housed at the Daresbury Laboratory [wikipedia.org] site of the Science and Technology Facilities Council [stfc.ac.uk], and is partly looked after by the Computational Science and Engineering department at said Lab. Once upon a time, HPCx was in the top ten machines in the world. It's since been overtaken by many commodity cluster architectures and most recently Blue Gene of course.

Gallery link

[Anonymous Coward]

If you want to see the pictures, here they are [hector.ac.uk].

Re:Poor comparison

[icehawk55]

160 GB Hard Drive vs RAM. Not a good comparison. They didn't mention the SAN storage we have, the tape backups, etc etc. I know. I know. I'm being picky.