[Beowulf] Register article on Cray Cascade

Steffen Persvold sp at numascale.com
Fri Nov 9 09:10:09 PST 2012


On 11/9/2012 17:17, Vincent Diepeveen wrote:
> The latency estimate taking 5 hops seems a tad optimistic to me
> except when i read the English wrong and they mean 1.7 microseconds a
> hop making it for a 5 hop 5 * 1.7 = 8.5 microseconds in total.

What they state is that nearest neighbour is 1.2 usec. Of those 1.2 
usec, only a fraction is actual fabric latency, which is why they can do 
5 fabric hops in 1.7 (meaning you have rougly 100ns per fabric hop which 
is very reasonable with 12.5/14 Gbps links).

The rest of the 1.2/1.7 usec is getting in and out of the chip (to the 
CPUs).

Cheers,
-- 
Steffen Persvold, Chief Architect NumaChip
Numascale AS - www.numascale.com
Tel: +47 92 49 25 54 Skype: spersvold

PS: Not working for Cray or have ever touched one so I can't confirm the 
numbers, but I know a thing or two about network chips and what they 
quote is not unreasonable at all :)

>
> "Not every node is only one hop away, of course. On a fully
> configured system, you are five hopes away maximum from any socket,
> so there is some latency. But the delta is pretty small with
> Dragonfly, with a minimum of about 1.2 microseconds for a short hop,
> an average of 1.5 microseconds on average, and a maximum of 1.7
> microseconds for the five-hop jump, according to Bolding."
>
> On Nov 8, 2012, at 7:13 PM, Hearns, John wrote:
>
>> Well worth a read:
>>
>>
>>
>> http://www.theregister.co.uk/2012/11/08/
>> cray_cascade_xc30_supercomputer/
>>
>>
>>
>> John Hearns | CFD Hardware Specialist | McLaren Racing Limited
>> McLaren Technology Centre, Chertsey Road, Woking, Surrey GU21 4YH, UK
>>
>>
>> T:  +44 (0) 1483 262000
>>
>> D:  +44 (0) 1483 262352
>>
>> F:  +44 (0) 1483 261928
>> E:  john.hearns at mclaren.com
>>
>> W: www.mclaren.com
>>
>>
>>
>> The contents of this email are confidential and for the exclusive
>> use of the intended recipient. If you receive this email in error
>> you should not copy it, retransmit it, use it or disclose its
>> contents but should return it to the sender immediately and delete
>> your copy.
>>
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin
>> Computing
>> To change your subscription (digest mode or unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>




More information about the Beowulf mailing list