Talk:K computer

Latest comment: 10 years ago by 132.249.65.175 in topic What's a compute node?

What will this be used for please? edit

thanks — Preceding unsigned comment added by 67.81.199.45 (talk) 02:03, 30 October 2011 (UTC)Reply

"K sucks up enough electricity to power 10,000 homes." source edit

Computers versus Brains; Computers are good at storage and speed, but brains maintain the efficiency lead by Mark Fischetti SciAm October 25, 2011 99.181.138.228 (talk) 05:13, 2 November 2011 (UTC)Reply

What's a compute node? edit

I'm sure it's supposed to be computer node but still, what is it referring to? Nodes can be any number of things in computer terminology.Muleattack (talk) 22:36, 21 January 2012 (UTC)Reply

Here, it refers to Computer cluster nodes.Jasper Deng (talk) 22:39, 21 January 2012 (UTC)Reply
This is not correct. In Fujitsu's presentations, they used the term "node" to refer to a compute core. The numbers given in this article for number of racks and number cores/nodes per rack are very confused. There are 672 racks (not 864), each with 96 nodes (64k nodes total), each with 8 cores per node (705k cores).132.249.65.175 (talk) —Preceding undated comment added 16:24, 29 May 2013 (UTC)Reply

minecraft will still lag as hell... — Preceding unsigned comment added by 92.247.136.18 (talk) 18:28, 5 March 2012 (UTC)Reply

There seems to be confusion over the number of racks and CPUs, but if you look at the Japanese page, it specifically says (translated into English). Note there were two LINPACK tests in June and November and in each test, the number of racks and CPUs are different because in June, when it hit it's first record, the supercomputer was still under construction:

On June 20th, the LINPACK benchmark was executed for 28 hours using 672 racks (68,544 CPUs), which was in the middle of installation, and as a result, 8.162 petaflops with an execution efficiency of 93.0% was achieved. . As a result, it was ranked TOP500 [5].

On November 2, the LINPACK benchmark recorded 10.51 petaflops (floating point arithmetic at 1.051 times per second) with an execution efficiency of 93.2% (LINPACK result 10.51PFLOPS/theoretical efficiency 11.28PFLOPS), announced that it had achieved its initial performance goal. This is the final configuration consisting of 864 racks (88,128 CPUs), and completed the installation and testing of LINPACK benchmarks for 29 hours and 28 minutes on October 7th to 8th to confirm their basic operation and design performance. This is the result of running the program and measuring. [76].