Yorkshire Building Society (YBS)
Rate this documentShare with colleagues
© 2015 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. The only warranties for HP
products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed
as constituting an additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein.
4AA5-9253ENW, June 2015
Case study | Universit Paul Sabatier
Sign up for updates
hp.com/go/getupdated
the distributed parallelism of our software
code without taking up a huge amount of
space.”
In UPS’s benchmark tests, a single 4.3U HP
Moonshot System produced total gigaFLOPS
equivalent to 17.2U of Dell PowerEdge R620
rack servers. The Dell servers would also
required 2.4 times more power to get the
same output as HP Moonshot. What’s more,
when idle the HP Moonshot System consumed
0.8 watts per core compared to 5.25 watts per
core for the Dell servers.
“The power savings from HP Moonshot is
enormous,” Scemama remarks. “That not only
meets our objectives to be more green but
also reduces our energy costs dramatically.”
The right balance of
performance and eiciency
HP Moonshot also provides UPS with optimal
performance for its quantum Monte Carlo
project. While it’s easy to expect a faster CPU
to run calculations faster, some very complex
software code can execute only so quickly. In
fact because CPU speeds have increased so
dramatically, it has become more and more
diicult for certain code to approach the
theoretical performance peaks on today’s
high-performance processors.
Due to the unique characteristics of UPS’s
algorithms, they perform much better when
spread across large numbers of compute
nodes that deliver suicient but not excessive
speed. The university’s benchmark tests
proved this point.
A Dell PowerEdge R620 with the Intel Xeon
Processor E5-2670 and 16 cores completed
7,200 quantum Monte Carlo blocks in 13,216.0
seconds. By comparison, the HP Moonshot
System based on the Intel® Atom™ Processor
C2730 with 1,440 total cores completed the
same workload in 768.12 seconds—17.2 times
faster.
“HP Moonshot provides the right balance of
performance and eiciency, enabling our
researchers to run their quantum Monte
Carlo simulations with the desired speed and
the least amount of energy consumption,”
notes Scemama. “Based on our studies HP
Moonshot’s CPUs, which are low-frequency
and not highly vectorized, prove to be very
performance-eicient for the types of
calculations we perform.”
He concludes, “HP Moonshot has opened up
a whole new perspective on how we run our
quantum Monte Carlo simulations, and that’s
exactly what we were looking for. Using the
insights we gained from these initial studies
we’ll now see how our other software code
performs on HP Moonshot. By proving HP
Moonshot here in the lab, we hope to bring
highly energy-eicient software code into the
commercial marketplace to help businesses
more green and cost-eective.”
Customer at a glance
Hardware
• HP Moonshot System
• HP ProLiant m350 Servers
• HP Moonshot 1500 Chassis
Software
• Ubuntu Linux
• Quantum Monte Carlo simulations
Our partners support