• Read the Report
• The Science Behind the Report
• Infographic

Executive Summary

This benchmarking report presents a structured evaluation of Dell’s PowerEdge R770 and R760 servers under scalable, real-world workloads. The assessment was conducted using VMware-based virtualization environments running MySQL databases with HammerDB load generators, modeled on TPROC-C benchmark protocols. 

Key performance indicators such as Transactions Per Minute (TPM), New Orders Per Minute (NOPM), and power consumption were measured across increasing concurrency and workload scales. The tests used a standardized configuration (including dual Intel Xeon processors with 32 cores per socket) to simulate realistic enterprise operations. 

The findings offer a comparative behavior profile between the two platforms, particularly across metrics like power efficiency, VM workload density, and concurrency performance—providing infrastructure teams with actionable reference points for planning and evaluation. 

What follows is a neutral benchmarking analysis based on consistent configurations. It presents a measurable behavior profile across server generations and offers a practical reference for infrastructure teams evaluating system performance, platform evolution, or deployment planning, particularly in contexts where performance per watt, VM density, and concurrency are becoming critical infrastructure traits.

This study was conducted with authorized collaboration and inputs from Dell Technologies and Intel, ensuring alignment with platform configurations and benchmark standards.

Testing Methodology

We conducted controlled benchmarking tests to assess how Dell PowerEdge R770 and R760 servers handle transaction-based workloads under repeatable, scalable conditions.

• Test Environment Configuration: The testing setup consisted of paired MySQL database virtual machines and HammerDB-based workload generators running on a virtualized infrastructure using VMware ESXi environments. Each configuration was designed to maintain consistency in virtual machine count, concurrency levels, and test duration.
• Key Metrics Captured: The benchmarking used the HammerDB tool configured to run workloads based on the TPROC-C protocol, which simulates transaction processing based on real-world scenarios. Each workload run involved a fixed number of virtual users and a defined number of warehouse units to simulate a realistic data and compute environment.  Alongside transactional throughput, power consumption during runtime was also tracked to provide insight into the performance-to-power behavior of the servers under scale. 
• Benchmarking Tool: Measurements were taken for transactions per minute (TPM), new orders per minute (NOPM), execution power draw, and network throughput. The goal was to observe server behavior in consistent conditions across different scales, ranging from small to high concurrency levels.
• Test Consistency: Each test iteration was conducted with the same control parameters to isolate the effect of server generation on performance, responsiveness, and resource usage.

Control measures

The test configuration was standardized on a 32-core CPU setup. This specification reflects the most commonly deployed core count in VMware-based virtualized environments, providing an appropriate balance between compute availability, licensing optimization, and scalability. By anchoring the benchmark on this configuration, the evaluation aligns with what infrastructure teams typically plan for in enterprise-grade deployments.

Here are the control parameters that defined our testing environment. These, as indicated before, were tweaked systematically to create credible and measurable observations.

Observation metrics

Here are the metrics we tracked (observed) based on the changes to our control parameters.

Correlation and derived efficiency metrics

Dell PowerEdge R770

A next-generation compute platform powered by Intel Xeon 6 processors (former codename Granite Rapids). Built for performance-intensive environments, the server supports high I/O throughput, accelerator integration, and scalable memory bandwidth. 

Ideal for workloads demanding concurrency stability, consistent virtualization output, and better power-to-performance efficiency under load.

Dell PowerEdge R760

An infrastructure-optimized server built for general-purpose workloads, offering dependable performance across virtualized and cloud-native operations - powered by Intel® Xeon® Gold processors. 

Designed to operate efficiently in constrained environments where space, power, and cost control are key factors.

Test Scenarios

All tests were conducted in a virtualized lab environment using ESXi infrastructure. SQL and HammerDB virtual machines were deployed in defined ratios to simulate real-world concurrency.

• Each test maintained a consistent mapping of user load, VM count, and warehouse configuration.
• Each test run was time-bound and triggered simultaneously across the SQL and load-generating VMs. The systems were left in a default state between runs to avoid environmental contamination.
• The same test sequence was used across both server models. Metrics were captured during the test runtime and directly after completion. The testing software was configured to log throughput and new order rates, along with network statistics. Power readings were recorded at pre-execution to isolate base system draw.\Each test point was validated through a complete run; partial passes or failures were not included in the dataset. This allowed a consistent performance line to be plotted across increasing complexity levels, while keeping the test environment stable.

Results and Observations

Each test was conducted under the same load structure for both Dell PowerEdge R770 and R760 servers, enabling a direct comparison. Metrics such as TPM, NOPM, power draw, and throughput were used to assess core performance and efficiency patterns. Each test was conducted under the same load structure for both Dell PowerEdge R770 and R760 servers, enabling a direct comparison. Metrics such as TPM, NOPM, power draw, and throughput were used to assess core performance and efficiency patterns.

The results reflect consistent behavior under concurrency and scale, backed by the architectural strengths of Intel Xeon Scalable processors. These CPUs are designed to sustain throughput under multi-VM load conditions, with performance stability across workloads and runtime stages. The validated runs form a credible reference point for infrastructure evaluations.

NOPM change with increase in virtual users

Standouts in Dell PowerEdge R770

The Dell PowerEdge R770 showed measurable efficiency gains across key metrics, with up to 27.33%[1] improvement in NOPMs under high concurrency and sustained TPM gains across scaled workloads. These results were captured in controlled and scalable configurations.

While the overall power draw increased under load, the TPM per watt metric showed relative improvement under moderate workloads, aligning with expected energy efficiency gains in next-generation hardware. The system delivered consistent output across scale levels, supporting its role in environments where performance consistency is critical across dynamic operational states.

The R770’s performance trends suggest suitability for VM-dense, transactional, and hybrid workloads that demand throughput and control across distributed compute nodes.

Inferences and Conclusion

The gains observed in the R770 were not isolated to any one load profile. Instead, improvements were distributed across light to high-load scenarios, suggesting the system is structurally better suited for environments with variable demand. The correlation analysis supported this view, showing stable relationships between input controls (VMs, users) and output metrics (TPM, NOPM, power).

These infrastructure traits - predictable scaling, consistent power-performance behavior, and I/O stability, also align with the baseline expectations of AI-adjacent workloads. While not tested explicitly for AI tasks, the R770’s behavior suggests readiness for environments where inference, streaming, or parallel compute tasks are evolving alongside traditional virtualized services.

Please also read the detailed science report (showcasing the deep logic and process followed during the benchmarking tests).

Appendix A - Testing environment assumptions

References

1.    Based on Claim ID CLM-014581 – Testing was conducted by Calsoft with servers located in Dell Labs during the June 2025 timeframe. The R770 server was running 2-socket 6th Gen Intel® Xeon® 6737P CPUs with 32 Cores, compared to the R760 server running 2-socket 5th Gen Intel® Xeon® 6548N CPU with 32 cores.

2.    Based on Claim ID CLM-014582 – Testing was conducted by Calsoft with servers located in Dell Labs during the June 2025 timeframe. The R770 server was running 2-socket 6th Gen Intel® Xeon® 6737P CPUs with 32 Cores, compared to the R760 server running 2-socket 5th Gen Intel® Xeon® 6548N CPU with 32 cores.

3.    Based on Claim ID CLM-014583 – Testing was conducted by Calsoft with servers located in Dell Labs during the June 2025 timeframe. The R770 server was running 2-socket 6th Gen Intel® Xeon® 6737P CPUs with 32 Cores, compared to the R760 server running 2-socket 5th Gen Intel® Xeon® 6548N CPU with 32 cores.

Calsoft is a leading technology-first partner, providing digital and product engineering services. For over 25 years, Calsoft has been  helping its customers solve their business challenges using technologies in Storage, virtualization, networking, security, cloud, AI/ML, IoT, and telecommunication domains.

Write to us at: marketing@calsoftinc.com 

Website: www.calsoftinc.com