[Beowulf] Thought that this might be of interest
Joe Landman
landman at scalableinformatics.com
Mon Nov 6 08:04:00 PST 2006
Richard Walsh wrote:
> Joe Landman wrote:
>> We are working on a machine with 2 Opteron 2218s. For laughs, I ran
>> streams on it. Here are 1,2,3,4 processor data.
>>
>> 1 processor:
>>
>> Copy: 5713.9944 0.0591 0.0560 0.0600
>> Scale: 5713.9822 0.0587 0.0560 0.0600
>> Add: 5454.2389 0.0911 0.0880 0.0920
>> Triad: 5454.1576 0.0916 0.0880 0.0920
>>
> Mmmm ... these actually seem low considering the bus bandwidth is
> supposed to be 10.67 GB/sec
Heh.... don't shoot the measurer ... :)
I agree BTW that I expected this to be about 7++ GB/s per socket. This
is a pathscale built binary. Will try with PGI and Intel as well.
> in the socket F. Using the 75% rule (1 silent read, 2 actual reads,
> and 1 write) from the triad shouldn't
> we get:
>
> 10.67 * .75 = ~8 GBs/sec
Possibly... The bandwidth per socket is 10.67 GB/s. So the two thread
on one socket should be pretty darned close to that. I am surprised
that it wasn't.
Moreover, other folks seem to note what we observed:
http://tweakers.net/reviews/646/5 with about 5.3 GB/s BW/core. I think
affinity is keeping two threads on the same core, so it might take a
little bit of work to get a real 1 thread/socket number.
[...]
>> 4 processors:
>> Copy: 13332.7765 0.0302 0.0240 0.0360
>> Scale: 13332.7765 0.0262 0.0240 0.0320
>> Add: 13332.6220 0.0400 0.0360 0.0480
>> Triad: 14999.0756 0.0391 0.0320 0.0480
> These are the numbers I would expect here ... about 70% of 2 *
> 10.67 GBs/sec.
> I think that your out-"liar" is actually giving you the correct
> numbers.
Possible. I was expecting something north of 12 GB/s.
--
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web : http://www.scalableinformatics.com
phone: +1 734 786 8423
fax : +1 734 786 8452 or +1 866 888 3112
cell : +1 734 612 4615
More information about the Beowulf
mailing list