[Beowulf] ARM cpu's and development boards and research

Prentice Bisbal prentice.bisbal at rutgers.edu
Wed Nov 28 07:30:21 PST 2012

On 11/27/2012 07:32 PM, Vincent Diepeveen wrote:
> On Nov 28, 2012, at 12:17 AM, Prentice Bisbal wrote:
>> On 11/27/2012 03:37 PM, Douglas Eadline wrote:
>>>> My interest in Arm has been the flip side of balancing flops to 
>>>> network
>>>> bandwidth.  A standard dual socket (AMD or Intel) can trivially 
>>>> saturate
>>>> GigE.  One option for improving the flops/network balance is to add
>>>> network bandwidth with Infiniband.  Another is a slower, cheaper, 
>>>> cooler
>>>> CPU and GigE.
>>> applause.
>> I applaud that applause.
>> What Bill has just described is known as an "Amdahl-balanced system",
>> and is the design philosophy between the IBM Blue Genes and also
>> SiCortex. In my opinion, this is the future of HPC. Use lower power,
>> slower processors, and then try to improve network performance to reduce
>> the cost of scaling out. Essentially, you want the processors to be
>> *just* fast enough to keep ahead of the networking and memory, but no
>> faster to optimize energy savings.
> For HPC the winning concept seems to be increasing corecount at 
> manycores.
> We also see how bluegene couldn't keep its concept - it's having what 
> is it 18+ cores
> now or so?

It's not 18+. It's exactly 18 cores. And only 16 are used for 
computation. One is used for operating system overhead, and the other is 
a spare.
This is exactly in keeping with the Blue Gene Concept, which is using 
low-power processors to conserve energy connected to highly-optimized 
interconnects to create a more balanced system.  'Low-power' and 'low 
core-count' are not the same thing.

Also, the networking for the Blue Gene/Q is quite a bit faster than the 
P's, which is why the Q has 4x as many cores, running at ~2x the 
clockspeed. Someone at IBM must have done the math and came to the 
conclusion that with the increase in communications performance, you can 
increase the performance of an individual node by ~8x and still have a 
balanced system.


More information about the Beowulf mailing list