[Beowulf] Rackable / SGI
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Eugen Leitl eugen at leitl.orgSat Apr 4 07:06:22 PDT 2009
- Previous message: [Beowulf] Rackable / SGI
- Next message: [Beowulf] Rackable / SGI
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Fri, Apr 03, 2009 at 01:32:13PM -0700, Greg Lindahl wrote: > > Will have to do with embedded memory or stacked 3d memory a la > > http://www.cc.gatech.edu/~loh/Papers/isca2008-3Ddram.pdf > > We've been building bigger and bigger SMPs for a long time, making > changes to improve the memory system as needed. How is multicore any Off-die memory bandwidth and latency are limited, so many codes start running into memory bottlenecks even at moderate number of cores (quad-cores seem to be a sweet spot). Mounting memory on top of CPU cores, and 3d stacks with through vias appears a cheap way to increase lanes and shrink geometries, which allows for higher data rates and less heat dissipation. Of course this arguably won't scale to a kilocore on-die (or wafer), and SMP emulating a shared memory will probably break down way before (synchronizing lots of writes to the same location is expensive in this relativistic shared-nothing universe). > different? In the computer industry, change is normal. Incremental changes, sure. -- Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
- Previous message: [Beowulf] Rackable / SGI
- Next message: [Beowulf] Rackable / SGI
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
