Can hosting provider OVHcloud convince against the hyperscalers?

Dimitrios Kechagias already carried out a comprehensive cloud provider comparison of all leading providers back in 2023. Within this provider benchmark, the 2-series of OVHcloud instances was included and received an average ranking, with the c2-7 instance even reaching 3rd place in the evaluation. To build on the existing analyses, this further benchmark follows his criteria.

Source: https://dev.to/dkechag/cloud-vm-performance-value-comparison-2023-perl-more-1kpp

OVHcloud's new server generation was recently released and is currently being delivered to the individual locations according to the following schedule.

With the new generation, the servers get a notable processor upgrade via the "AMD EPYC Milan", as it now also matches the Google Cloud and AWS instances. The AMD Milan processor is regarded in many cases as more powerful than the Intel Haswell processor. For one, AMD Milan is based on the Zen 3 microarchitecture, which offers a range of improvements including better instruction execution units, cache hierarchies and prediction algorithms, overall leading to higher performance per clock cycle. AMD Milan processors are also manufactured on a more modern process node, enabling higher performance alongside improved energy efficiency. We also find the switch to NVMe drives particularly attractive, enabling even better performance for all write access. The new generation was announced in 3 variants: "B-General Purpose, C-Compute and R-Memory Optimized".

Current news on the rollout can be found on the OVHcloud GitHub: https://github.com/ovh/public-cloud-roadmap/issues/352

The comparison test

To carry out the comparison tests, the most uniform server configurations possible are used. Here we limited ourselves to the following setup:

• OS: Ubuntu 23.02

• CPU: 2 vCores

• RAM: 8GB RAM

• Hard drive: Internal hard drive

The Geekbench test is reported as a single-core score as well as a multi-core score. Depending on the software's use case, you need to assess the single or multi score.

The single-core speed of a processor measures the speed at which a single compute core within the processor can execute operations. This speed is given in gigahertz (GHz) and indicates how fast the processor can execute the instructions assigned to it. The single-core speed of a processor is important because it represents the performance of a single compute core within the processor. This is important for applications or tasks that cannot be parallelised or where parallelisation is inefficient. The multi-core speed of a processor refers to the combined performance of all its compute cores when executing tasks that can be parallelised. Unlike single-core speed, which measures the performance of a single compute core, multi-core speed takes into account the processor's ability to process multiple tasks simultaneously. This is important because it affects a system's ability to handle parallel tasks.

The following compares OVHcloud instances with the top-ranked variants from individual cloud providers.

The evaluation clearly shows that developing the new server range paid off. Through the adjustments described above, a benchmark improvement of 60-100% was achieved. 

The following table shows a detailed view of the server configuration including monthly costs in euros and the Geekbench score.