
4.9 PFlops «Lomonosov-2» supercomputer
Current rank in Top500 rating - #107 (November 2019).
Current rank in Top50 rating - #1 (September 2019).
«LOMONOSOV-2» KEY FEATURES
Peak performance |
4,946 PFlops |
Linpack performance |
2,478 PFlops |
Compute nodes |
1 696 |
Main processor type |
Intel Haswell-EP E5-2697v3, 2.6 GHz, 14 cores |
Accelerator type |
NVidia Tesla K40M |
Total cores |
64 384 |
RAM per node |
64 GB |
Main network |
Infiniband FDR |
I/O network |
Infiniband FDR |
System network |
Gigabit Ethernet |
Management network |
Gigabit Ethernet |
Operating system |
CentOS 7 |
Manufacturer |
T-Platforms |
Year of issue |
2017 |
Partition Compute |
Compute nodes |
1487 |
Main processor type |
Intel Haswell-EP E5-2697v3, 2.6 GHz, 14 cores |
Accelerator type |
NVidia Tesla K40M |
RAM per node |
64 GB |
Main network |
Infiniband FDR |
I/O network |
Infiniband FDR |
Management network |
Gigabit Ethernet |
Operating system |
CentOS 7 |
Partition Test |
Compute nodes |
32 |
Main processor type |
Intel Haswell-EP E5-2697v3, 2.6 GHz, 14 cores |
Accelerator type |
NVidia Tesla K40M |
RAM per node |
64 GB |
Main network |
Infiniband FDR |
I/O network |
Infiniband FDR |
Management network |
Gigabit Ethernet |
Operating system |
CentOS 7 |
Partition Pascal |
Compute nodes |
160 |
Main processor type |
Intel Xeon Gold 6126, 2.6 GHz, 12 cores |
Accelerator type |
2 x Nvidia P100, 3584 Cuda cores, 16 GB |
RAM per node |
96 GB |
Main network |
Infiniband FDR |
I/O network |
Infiniband FDR |
Management network |
Gigabit Ethernet |
Operating system |
CentOS 7 |
Partition Phi |
Compute nodes |
4 |
Main processor type |
Intel(R) Xeon Phi(TM) CPU 7230, 1.30GHz, 64 cores |
Accelerator type |
- |
RAM per node |
112 GB |
Main network |
Infiniband FDR |
I/O network |
Infiniband FDR |
Management network |
Gigabit Ethernet |
Operating system |
CentOS 7 |
Storage system |
Home Directory Storage |
221 TB |
Work File Storage |
925 TB |
Service storage |
246 TB |
Partition |
nodes/type |
memory |
drives |
max core-hours |
max/default time (hours to the task) |
max tasks |
max running tasks |
max nodes per task, no more |
compute |
1504/K40 |
64 GB (4,5 GB/core) |
- |
- |
48/24 |
8 |
4 |
100 |
pascal |
160/P100 |
64 GB (5,3 GB/core) |
- |
- |
20/20 |
8 |
4 |
40 |
test |
32/K40 |
64 GB (4,5 GB/core) |
- |
- |
15 min |
3 |
1 |
15 |
