The Provision of HPC Resources to Top Universities

Recently, the University of Melbourne is ranked #1 in Australia and #33 in the world, according to the Times Higher Education World University Rankings of 2015-2016 [1]. The rankings are based on a balanced metric between citations, industry income, international outlook, research, and teaching.

Although it is not a subject heavily reasearched, there is some evidence that the local provision of high performance computing correlates with improved research results [2]. This, of course, is evident from first principles. Ceteris paribus, if research is dependent on processing increasingly large datasets, then those research institutions that have that capacity will be more successful that those that do not. "Appearance on the Top 500 list is associated with a contemporaneous increase in NSF funding levels as well as a contemporaneous increase in the number of publications... The conclusion is that consistent investments in HPC at even modest levels are strongly correlated to research competitiveness."

The following dataset compares the presence local HPC systems for general availability based on cores, and compares it to the Times Higher Education ranking. It's written in free text, and it's based on a manual search public records which is subject to human error. Nevertheless, it does give a reasonable idea of on-campus availability of compute resources. Some campuses are, of course, providers of national facilities as well, or they have close access to national facilities. The measure of cores is not taking into account age and clockspeed etc either, but with the broad brushstrokes being used here it is a reasonable to pretend that overall computational capacity is equivalent.

Cores are not, of course, the only metric that should be under consideration. Throughput is probably the most important, as that is what generates research. However given that figure is not readily available cores provides a prima facie indication of the dedication of the institution to HPC support. In this regard the University of Melbourne, considered to be the top-ranked university in Australia, lags behind the average of the rest of the Top 50 TLS list. The question is raised of course, where would the university be if it had a Top 50 average level of HPC computing resources.

1. University of Oxford. 7,650 cores total (c.f.,, plus access to national HPC facility Archer (UK National Supercomputing Service).

2. California Institute of Technology 21,600 cores

3. Stanford University. 13,000 cores, plus 2,040 GPUs (

4. University of Cambridge. 14,720 core , plus access to national HPC facility Archer (UK National Supercomputing Service).

5. Massachusetts Institute of Technology 48,672+ cores (c.f.,, with other partners e.g., Harvard..,

6. Harvard University Odyssey partners with MGHPCC, above "Additionally, there are over 1,000,000 NVIDIA GPU cores which can greatly increase the speed of parallel processing jobs."

7. Princeton University, 25,524 cores

8. Imperial College London 13,824 cores (main system)

9. ETH Zurich – Swiss Federal Institute of Technology Zurich 45,000+ Provisioned through Swiss HPC

10. University of California, Berkeley 8,000+ (plus 180,000 GPGPU)

10. University of Chicago 11,256 cores (main system)

12. Yale University 22,000+

13. University of Pennsylvania 2,304 cores (This is a strange outlier, and I suspect they must have other resources)

14. University of California, Los Angeles 8,020+ (plus faculty-specific systems)

15. University College London 10,944

16. Columbia University 7,376 cores,

17. Johns Hopkins University over 22,500 cores - Plus shared with Maryland

18. Duke University 7,000 cores -

19. Cornell University - 8,516+

20. Northwestern University 11,800+ CPU cores -

21. University of Michigan-Ann Arbor 27,000 cores -

22. University of Toronto 65,536 core on BGQ system plus 30,240 on GPC -

23. Carnegie Mellon University 2,888+

24. National University of Singapore - 6,000+ 1,300 PC Grid, 16,000+ GPGPU,

25. London School of Economics and Political Science 416 cores -

25. University of Washington-Seattle 10,784 cores -

27. University of Edinburgh, 98,304 cores. and ARCHER 118,080 processing cores

28. Karolinska Institute 19,800 cores - (Note: this is also the national facility)

29. Peking University Unknown; almost certainly has access to national facilities.

30. École Polytechnique Fédérale de Lausanne - 26,592 cores,;;;

30. LMU Munich - 10,752 cores,

33. Georgia Institute of Technology - 30,000+

33. University of Melbourne - 2,380 cores maximum, includes: 300 HPC cores (physical partition). An addition 920 are available through the NeCTAR research cloud allocation (cloud and cloudtest partitions), plus an temporary allocation of 1160 cores to punim0095. Also, 5 GPGPU nodes. ; plus shared access to VLSCI (2112 cores), in some cases (

35. Tsinghua University Unkown; 9,216 in 2011

36. University of British Columbia 9,600 cores plus other Compute Canada resources

36. University of Illinois at Urbana-Champaign - 9,056 cores

36. King’s College London 1,464 cores -, plus access to ARCHER.

39. University of Tokyo - 556,104 cores (Note: national facility) Plus others:;;;

40. KU Leuven 3,520 cores -

41. University of California, San Diego 67,936 cores -

42. McGill University 20,400 cores -

43. Heidelberg University - 9,384+ cores,

43. University of Hong Kong 4,080+ cores -

45. University of Wisconsin-Madison - 3,616+ core for HPC They also use a campus-wide HTC Condor implementation;

46. Technical University of Munich 10,752 cores -
See also LMU Munich, shared facility.

47. Australian National University 57,864 cores - (Note: This is a national facility)

48. University of California, Santa Barbara - 4,358 cores

49. Hong Kong University of Science and Technology 4,080+ cores - See University of Hong Kong.

50. University of Texas at Austin - 36,288+ cores;jsessionid=E75B7A0528ED3008...


1] Phil Baty (ed), Times Higher Education World University Rankings, 22 September 2016

2] Apon A., Ahalt S., Dantuluri V, et. al., (2010) "High Performance Coputing Instrumentation and
Research Productivity in U.S. Universities", Journal of Information Technology Impact, Vol 10,
No 2, p87-98

Thanks to Erica Hoehn in her assistance in compiling the URLs for HPC resources