I hope to attend.
The following advert (apology) might be of (great?) interest to this
subject:
Intel debuts new Xeon Phi coprocessors
Bizarre as it may first seem, but the last time that Intel released a
co-processor was way back in 1987 when the 80387 launched. This
coprocessor plugged into the motherboard next to your 80386, adding the
ability to make complex Floating Point calculations that weren�t part of
the original 8086 instruction set. Since then however all Intel (and
compatible clones) have had FPU (floating point units) built in,
negating the need for a coprocessor.
That�s all set to change in 2013 however as Intel has announced a whole
new series of exciting coprocessors � the Xeon Phi series. Like previous
co-processors the Xeon Phi has been specifically designed to speed up
certain types of calculations, in this case the highly parallel
compute-intensive workloads found in many server rooms and supercomputer
clusters.
The Xeon Phi is based on the Intel MIC (Many Integrated Cores)
architecture, the name giving a strong hint that each Xeon Phi is based
around an array of multiple processors all working together in parallel.
The first model, the Xeon Phi 5110P includes 60 such cores and 8GB of
its own on-board memory. Packaged as a 16x PCI-E 2.0 card so that you
can install multiple Xeon Phis in your server, Intel claims it can
calculate as much as up 1 Teraflop of double precision peak performance,
at least ten times more than a standard 8086 CPU can manage.
Unlike competing coprocessors the Xeon Phi can be addressed using common
programming languages and tools, which will make it far easier and
quicker to start using.
On 02/01/2013 20:58, Phil Last wrote:
> BAA London - the London based discussion group for the APL family of
> array languages.
>
> The January Symposium will be held on Friday 25 at:
>
> The Albion
> 2/3 New Bridge Street
> City of London
> London
> EC4V 6AA
>
>
https://maps.google.co.uk/maps?q=the+albion+ec4v+6aa
>
> Upstairs room booked from 14.30 'til 17.30
>
> Meet earlier downstairs for a drink and/or a chat.
> ----------
> Agenda:
>
> Following Gert M�ller's talk about parallel processing with GPUs back