Talk:Blue Waters

Comments
Why the big push to have "petascale" mentioned AND linked in the article so much when the link doesn't exist and the word almost seems to be invented just for the Blue Waters stories out there? Even the definitions for "petascale" that are out there don't really make sense with how they use it. We don't usually refer to a home office as a gigascale computing center.

But can it run Crysis 3? — Preceding unsigned comment added by 69.121.18.113 (talk) 15:02, 31 March 2013 (UTC)


 * If it is true that petascale is a marketing nonce word, then it doesn't belong in WP. David Spector (talk) 00:38, 9 June 2013 (UTC)

Critique?
Should we add a criticism section? Blue Waters borders on fraud on the part of IBM as the contractor. They knew in 2008 when bidding on the project that NVIDIA GPUs (currently 936 gigaflops[1]) are already more powerful for physics computations than Power7 (projected 256 gigaflops[2]) will be. And this is after DARPA already threw $244 million at the Power7 to speed it up.

Numerical kernels will be need to be ported to the new Power7 vector units, so it will be at least 2012 before many codes will be ready to run on the system.

[1] http://en.wikipedia.org/wiki/Nvidia_Tesla [2] http://www.theregister.co.uk/2008/07/11/ibm_power7_ncsa/ GrEp (talk) 04:35, 15 May 2009 (UTC)

-- Note that Tesla slows down to less than 10% of that performance when you use it to do double-precision arithmetic, as is most common in scientific simulation software. That's purely the core, and doesn't account for data movement to/from the Tesla or from its memory to execution units. Note also that the porting effort of a code to a new narrow SIMD architecture is fairly low, especially compared to the completely different programming model presented by GPGPU hardware. It might be a lot of money, but the NSF program officers are not idiots. There's also a lot of work being done other than getting IBM to design, build, and install the machine. In other words, I don't see any fraud here. Full disclosure: I'm a PhD student at UIUC funded by the Blue Waters project. 98.222.133.234 (talk) 06:11, 3 September 2009 (UTC) -- One more thing, I forgot to note: we're doing application performance work for Blue Waters right now, using both a preliminary Power6 system and a cycle-accurate simulator for Power7. Our contract with NSF actually does specify that we meet the benchmark numbers for various particular simulation targets by the time the system is up and running. 98.222.133.234 (talk) 06:13, 3 September 2009 (UTC)

I called it about two years ago :) Looks to be a flop. http://chronicle.com/article/In-University-Supercomputing/126979/ GrEp (talk) 05:15, 6 April 2011 (UTC)


 * Who are you? That article does not label the project a flop, not even by implication. David Spector (talk) 18:13, 4 November 2011 (UTC)

Established or introduced?
Are supercomputers established or introduced and in which year was this computer first operated? Cray-1 is within an introduction category. Dates field in the infobox and categories should be completed once this is ascertained. - Shiftchange (talk) 01:07, 23 October 2016 (UTC)