Energy Saving¶
Due to high energy prices and reductions in funding, IT4Innovations has implemented a set of energy saving measures on the supercomputing clusters. The measures are selected to minimize the performance impact and achieve significant cost, energy, and carbon footprint reduction effect.
The energy saving measures are effective as of 1.2.2023.
Karolina¶
Measures¶
The CPU core and GPU streaming multiprocessors frequency limit is implemented for the Karolina supercomputer:
| Measure | Value |
|---|---|
| Compute nodes cn[001-720] CPU core frequency limit |
2.100 GHz |
| Accelerated compute nodes acn[001-72] CPU core frequency limit |
2.600 GHz |
| Accelerated compute nodes acn[001-72] GPU SMs frequency limit |
1.290 GHz |
Performance Impact¶
The performance impact depends on the arithmetic intensity of the executed workload. The arithmetic intensity is a measure of floating-point operations (FLOPs) performed by a given code (or code section) relative to the amount of memory accesses (Bytes) that are required to support those operations. It is defined as a FLOP per Byte ratio (F/B).Arithmetic intensity is a characteristic of the computational algorithm.
In general, the processor frequency capping has low performance impact for memory bound computations (arithmetic intensity below the ridge point). For processor bound computations (arithmetic intensity above the ridge point), the impact is proportional to the frequency reduction.
On Karolina, runtime increase up to 16% is observed for arithmeticaly intensive CPU workloads and up to 10% for intensive GPU workloads. No slowdown is observed for memory bound workloads.
Energy Efficiency¶
The energy efficiency in floating point operations per energy unit is increased by up to 30% for both the CPU and GPU workloads. The efficiency depends on the arithmetic intensity, however energy savings are always achieved.
Barbora¶
None implemented yet.
NVIDIA DGX-2¶
None implemented yet.
Complementary Systems¶
None implemented yet.