.png){kind=logo}
Optimizing Gen5 Storage Performance for Your Gen5 Computing Platform
- Oct 30, 2024
- 4 min read
In order to ensure your PCIe Gen5 NVMe storage performs optimally you will need to make sure the host platform is configured properly, and that you are properly testing the storage configuration.
This blog will examine 3 critical factors associated with PCIe Gen5 NVMe Storage:
1) Memory (system RAM)
2) System/Mainboard BIOS Settings
3) Benchmark Utility Settings/Test Scripts
These are of key importance, and apply to desktops, workstation and server platforms. However, but specific chipsets and motherboard combinations may have unique requirements.
System Memory
PCIe Gen5 storage media benefits from high-performance memory (system RAM). As a general rule, you will want to install as much memory as your budget allows for.
The type of memory used, and how this memory is configured varied from system to system. We recommend consulting the motherboard/platform user guide for a list of supported memory-modules.
Selecting the right type of memory can have a major impact on system performance. Dual-channel memory for example, is recommended for workstations or platforms with single CPUs, as it can optimize available bandwidth and transfer speed. (two modules installed into two memory lanes) can optimize bandwidth and improve transfer speeds.
In addition, some systems may provide various BIOS settings associated with system memory.
Motherboard BIOS Settings
Although PCIe Gen5 NVMe media is now widely available, it is not yet considered mainstream, as its most commonly used for specialized professional applications rather than those associated with consumers or SMBs. As a result, most Gen5 computing platforms are often paired with PCIe Gen4 NVMe media by default, and their BIOS configurations will reflect this.
When prepping a system for Gen5 storage media, it’s important to check the platform’s BIOS configuration menu to ensure Gen5 related settings are enabled and optimized. As the ideal configuration varies from system to system, be sure to consult your platform/motherboard’s user guide. PCIe Gen5 connectivity may be limited to specific slots, or even certain types of PCIe devices.
Some general settings are outlined below:
Memory Related Settings: As touched upon earlier, several memory settings may be provided by the system’s BIOS Configuration Menu:
NUMA (non-uniform memory access): This setting may also be presented along the line of “NUMA nodes per socket”. Although this setting can affect performance for any platform, the default NUMA settings for some motherboards may not be ideal for NVMe storage configurations. If the option is present, you may need to manually set the system to operate in NPS4 mode.
Memory Frequency: Memory frequency settings can have a big impact on performance. 5600 Hz is recommended for Desktops while 4800Hz is ideal for Server and Workstations.
Other Settings
PCIe MPS (Max Payload Size) can affects sequential performance. As a general rule, lowering the Max Payload Size improves compatibility with PCIe devices, but may result in performance loss. In most cases, it can be left in the default configuration, but using the maximum of 4096 is recommended for modern PCIe devices. If 4096 is not the system’s default setting, you want to consider checking the system/motherboard user guide for recommendations.
VT-d: VT-d (short for "virtualization for technology direct I/O access") is associated with some Intel-based server and workstation motherboards. Though its primary purpose is for virtualization platforms (it enables direct access to the host hardware from a virtual machine), it can have an adverse effect on transfer performance in some instances. If you are experiencing sub-par performance, try disabling the VT-d setting.
Power Management: While often overlooked, Power Management related BIOS settings can have a big impact on performance, especially if the system has been configured to operate in a “Green” Power Savings mode. In fact, many systems are now shipped with these modes enabled.
If your platform’s BIOS menu provides Power Management settings, make sure to prioritize system performance over efficiency.
Optimize Queue Depth and Thread Count for Accurate NVMe Benchmark Performance
Performance Benchmarks such as CrystalDiskMark (CDM) can be used to test storage media by simulating a wide range of workloads. However, such utilities may not be configured to test NVMe storage devices by default. In many cases, adjustments are required.
The most common are Queue Depth and Thread Count.
Queue Depth: By and large, NVMe storage media has been designed to execute a huge number of concurrent tasks, especially when compared to SAS/SATA SSDs or hard disks. The Queue depth of and NVMe SSD (the number of I/O requests a device can handle at one time) is measured in the tens of thousands, as opposed to tens or hundreds for a SAS or SATA drive.
As a general rule, increasing queue depth will result in higher performance, as the test will be simulating denser workloads associated with NVMe storage applications.
Real-World Applications:
Thread Count: As PCIe-devices, NVMe’s multi-threaded architecture was designed to work in conjunction with modern multi-core CPUs, enabling parallel I/O processing across multiple threads (A.K.A tasks or jobs).
This approach eliminates bottlenecks inherent in single-threaded protocols like SATA, ensuring benchmarks reflect the true capabilities of NVMe storage.
Properly configuring the number of threads (often represented as “workers” or “jobs”) is essential to unlock accurate performance measurements for NVMe storage.
HighPoint publishes benchmark reports and guidelines for their various NVMe products, including Rocket7600 and 1600 series Gen5 AICs and Adapters. Contact us for more information
HighPoint publishes various test scripts and setting recommendations for tested NVMe media with various benchmarking tools including FIO (Flexible I/O Tester), Iometer, and CDM.
For more information and sample scripts, please contact our Support Department
Learn More
This explains NVMe’s multi-threaded architecture, emphasizing proper thread configuration and HighPoint’s benchmarking tools for accurate performance evaluation.
This is super helpful for optimizing my new Gen5 setup! Speaking of optimizing, sometimes I just want to chill with a simple game. Have you ever tried a Scratch Cards Idle Game (No Download)? It's surprisingly relaxing and a nice brain break.
The first time Pixel Flow grabbed me was when I lost a level and immediately knew why. That matters. A lot of mobile games just feel messy when you fail. This one usually makes sense. You used the wrong color early. You clogged your slots. You didn’t peel the outer layer first. For anyone who hasn’t seen it, that’s what the game is built around. Matching colors on layered voxel shapes, but with a very limited margin for bad timing. I kept playing because it felt learnable.
Get beads out if you want something that scratches both the relaxing and stressful parts of puzzle games at the same time. The visuals and motion are smooth, but the pressure is real once the belt gets crowded and you know there’s only one useful route out of the mess.
This article really highlights how crucial every detail is for top performance. It reminds me of the strategic thinking you need in a game like blockblast, where every move affects your overall score. Great practical advice for anyone with Gen5 hardware!