|
Current projects ::
PRACE - towards petascale
Petapath in partnership with HP and SGI has delivered two petascale prototype systems as part of the
PRACE programme. The system is a testbed for new ideas in usability, scalability and efficiency of
large computer installations.
Read more...
HPC for Enterprise - market study
Petapath embarked on a fully funded market study to examine the impact
of heterogeneous computing model at an enterprise level.
Download the programme summary, and
if it sounds interesting, we are always looking for new participants.
|
|
What we do ::
In short: deliver performance.
Absolute performance. Price performance. System efficiency. Whichever the business requires.
Today, where enterprise can no longer rely on CPU clock speed scaling to increase performance,
new approaches are necessary. Petapath has worked successfully with a number of companies,
implementing both the optimization and acceleration strategies, in particular:
- Traditional code optimization
- Code refactoring and effective parallelization
- OpenMP and OpenCL (also CUDA C)
- System-level optimization and architecture
- Proof-of-concept and prototype development through to production code
|
|
Business benefits ::
"Obtaining even a modest 20 percent improvement in execution time through choice of processor architectures or
high-performance libraries would equate to a saving of more than $300,000 per annum on a 500 node single line-of-business grid.
Taken over a 10,000 CPU grid,
this small change could save in the order of $6 million each year. Combined with application optimizations,
savings of around $10 million annually are possible. Cost savings are only half the story.
The real value is often taken to be the opportunities that can be realized as a result of the efficiency improvements."
Dr. Alistair Dunlop, "Is your grid burning money"
read full article
|
