V444_2026 compute nodes for HPC cluster

All tenders are checked against the list of exclusion criteria to ensure that they meet the requirements described in the specifications (see Appendix, List of evaluation criteria). The total throughput is calculated in a simplified manner from the “SPECCPU2017 Floating Point Rate” baseline value summed over all computing nodes (standard and high meme nodes): total throughput = N × “SPECCPU2017 fprate base” (standard computing nodes) + “SPECCPU2017 fprate base” (high meme computing nodes) Here, “SPECCPU2017 fprate base” is the base value for the benchmark “SPEC CPU 2017 Floating Point Rate Benchmark” on a computing node, which must be shown in the offer for the offered computing nodes. The specified benchmark value must be demonstrable for the respective processor type and system configuration (e.g. by providing the link to an official publication on www.spec.org or by providing a certified and traceable benchmark measurement report on identical hardware architecture). The most economical bid shall be determined using the simple indicative methodology. It is necessary to achieve 1000 performance points. This reaches the offer with the highest total throughput performance of the HIPOLE partition. The performance points for the other offers result correspondingly from the ratio to this maximum value. The ratio of performance points of a provider divided by the total price offered results in a key figure, K = performance points / total price, which must be maximized by the prov…

Price = 100% ; The offer with the lowest price receives 100 points, then linear skimming.