Intel's single-chip cloud computer

Feb 11, 2010 by Lin Edwards report
Single Chip Cloud Computer has 48 Intel cores and runs at as low as 25 watts.

(PhysOrg.com) -- Intel Labs has recently shown off a 48-core prototype chip it calls a "single-chip cloud computer" or SCC.

Chief technology officer with , Justin Rattner, said the chip comprises 1.3-billion transistors arranged in a network of 24 tiles, each of which has two Pentium-class IA-32 cores, two L2 caches, plus a router to enable communications between cores. The system uses new software applications to control the power consumed by the cores, and to rapidly transfer data between the cores. This means data can go directly between cores without needing to go via the main memory, and this cuts the data transfer speed by 15 times. The software prevents the data being corrupted by instructing the cache sending the data to delete its copy after it is sent, and the receiving cache to delete old copies of the data before receiving the new.

The software controlling power consumption allows application developers, rather than the operating system, to decide how power consumed by the cores is controlled. The tiles can all be independently controlled, which means the on some cores can be reduced to as low as 25 watts, while others can be up to 125 watts. While some developers are not yet sure what they will do with the feature, many are interested in learning more.

Intel’s director of advanced microprocessor research, Nitin Borkar, said tasks could be programmed to run at greater rather than higher power if appropriate, or individual cores could be throttled back after they have finished their computations. This would give the system the “compute on demand” feature of traditional data centers.

Intel Labs are forming partnerships with industry and university researchers and producing 100 of the chips to enable research to refine the architecture and maximize its usefulness.

The chip was unveiled at the IEEE International Solid-State Circuits Conference in San Francisco on 8 February.

Explore further: Kickstarter project SAM kit helps teach hardware system coding

Related Stories

AMD Planning 16-Core Server Chip For 2011 Release

Apr 27, 2009

(PhysOrg.com) -- AMD is in the process of designing a server chip with up to 16-cores. Code named Interlagos, the server chip will contain between 12 and 16 cores and will be available in 2011.

Intel Discloses Details of Intel Core Microarchitecture

Mar 07, 2006

Intel Corporation today disclosed details of its forthcoming Intel Core microarchitecture, a new industry–leading foundation for Intel's multi–core server, desktop and mobile processors for computers later ...

Recommended for you

First self-contained step dimming LED tube

25 minutes ago

Samsung Electronics today introduced the industry's first AC Direct step-dimming LED linear replacement for T8 and T12 fluorescent tubes at the National Electrical Contractors Association (NECA) Convention ...

Battery system will be able to light 2,500 homes

54 minutes ago

One of the largest, most environmentally-friendly, battery-based energy storage systems in the nation will be installed at the University of California, San Diego the campus announced today (Sept. 29).

User comments : 6

Adjust slider to filter visible comments by rank

Display comments: newest first

shortcircuit
not rated yet Feb 11, 2010
I think you mean "cuts data transfer time by a factor of 15" or some such; Cutting speed 15 times isn't a desirable outcome.

I'm torn about them using a 32-bit processor as a base. I spend most of my time on workstations with 4GB or 8GB of RAM. Believe me, the latter case is worth it if only for the file cache.

While I don't particularly mind if they cut back on transistor count for things like branch prediction (who needs it with that short a pipeline?) or instruction reordering, having that much processing power stuck in a 32-bit address space will mean more code executed for marshaling large data sets and files into and out of the address space.
Buyck
not rated yet Feb 11, 2010
This type of chip will maybe aviable in 2016-2018 for everybody.
sender
not rated yet Feb 11, 2010
having that much processing power stuck in a 32-bit address space will mean more code executed for marshaling large data sets and files into and out of the address space.

I agree, maybe parallel busing the entire operation with tertiary switching wouldve helped solve that issue.
simulus
not rated yet Feb 12, 2010
There are more cores, but each one is smaller, probably not as powerful. Maybe there is an upper limit on the number of cores?
taka
not rated yet Feb 15, 2010
If you have a lot of cores then the only meaningful approach is multiple processes (one or few in each core) with multiple data (locally, also in cores as access to global memory do not scale well with increasing number of cores).

When traditional programming deal with instructions execution order and assume data is just there then here reversed approach is needed, dealing with data transport and assuming instructions are executed when they get the data they need.
Stuart21
5 / 5 (1) Feb 20, 2010
INVENTORS - DO NOT TRUST INTEL
I invented a CPU cooler - 3 times better than best - better than water. Intel have major CPU cooling problems - "Intel's microprocessors were generating so much heat that they were melting" (iht.com) - try to talk to them - they send my communications to my competitor & will not talk to me.

Winners of major 'Corporate Social Responsibility' awardS!!!

Huh!!!!

When did RICO get repealed?"

INVENTORS - DO NOT TRUST INTEL!!!