Part 2 of 2 — The Workload Evidence

In Part 1, we established the case that a data center SSD and its underlying CPU platform should be assessed together. Now we present the evidence. AMD's solutions engineering team and Micron's workload engineers recently tested this concept using a workload critical to virtually every enterprise: a relational database executing business intelligence and decision support queries.

The Test Setup

The study employed PostgreSQL, among the most commonly used open source relational databases, executing a BI/DSS workload on a server equipped with 5th Gen AMD EPYC processors. The workload adhered to the TPC-H methodology that AMD applied for the analysis, engineered to replicate the intensive, analytical queries powering dashboards, reports and planning applications.

The comparison was intentionally realistic: Micron's latest PCIe Gen5 drive, the Micron 9550, measured against an earlier Micron PCIe Gen4 drive, the 7450. Identical processor platform, identical database, identical workload. The only difference was the SSD generation, making the findings straightforward to understand.

One note for anyone referencing these figures within your organization: they originate from a vendor workload study applying a TPC-H based methodology, not an officially audited TPC result, so they should be interpreted as compelling directional guidance for planning purposes rather than a certainty for your specific situation.

The Results

The Gen5 drive demonstrated substantial improvements in queries per hour, expressed as QphH:

1.45x the QphH of the Gen4 SSD at scale factor 1000, representing a dataset of approximately 1 TB
1.33x the QphH at scale factor 300, representing roughly a 300 GB dataset

Two key observations emerge. First, these represent substantial improvements from modifying just one component within an otherwise stable system a rare and valuable discovery for database teams. Second, the performance gains scale with data volume: larger datasets showed greater improvements. This pattern makes sense: as datasets exceed available memory capacity, storage becomes the primary bottleneck and faster storage delivers more meaningful benefits.

Should You Always Upgrade to Gen5?

Not necessarily. The study itself acknowledges this reality. The key insight isn't that Gen4 is outdated. The Gen5 findings demonstrate that the 9550 is an excellent option when maximizing BI/DSS throughput is the primary objective, while the Gen4 drive remains a strong choice for standard relational database workloads.

This gives infrastructure teams a clear decision framework:

Opt for Gen5 when analytics performance is critical to your business, datasets are substantial and expanding or you're designing infrastructure built to remain effective through multiple hardware generations and evolving workloads, including AI related applications.

Stick with Gen4 when workloads are predictable and conventional, budget constraints exist and current performance satisfies business requirements.

The optimal choice depends on your specific workload which is exactly why evaluating your workload before making purchases, as covered earlier in this series, delivers real value.

Why Collaborative Testing Deserves Your Attention

It's tempting to skip over the partnership aspect and focus solely on the results. That would be a mistake. These findings are credible precisely because both companies validate them together under real-world scenarios. Verified configurations provide server manufacturers with a more reliable foundation and give enterprise customers infrastructure optimized for actual data center conditions not just theoretical benchmark performance.

This is the benchmark standard worth applying to any storage and compute recommendation: has this particular combination undergone testing together, using a workload that matches yours?

Converting Evidence Into Implementation

This is where a skilled partner demonstrates their value. Understanding that a Gen5 SSD can deliver up to 1.45x more database queries per hour is helpful. Understanding whether that improvement justifies the investment in your situation, how many drives and lanes your architecture truly requires and how the memory and networking components should be configured around them is what transforms a benchmark into an actionable decision.

As a Micron reseller partner, Beyond Limits Technologies assists organizations in making exactly that conversion: analyzing your database and analytics workloads, identifying the right SSD generation and platform for them, and building everything into a system that delivers the performance the validation data indicates it should.

Considering PCIe Gen5 storage for your upcoming database or analytics project? Reach out to Beyond Limits Technologies for a consultation, and we'll configure it based on your workload, not marketing specifications.