PB65MX4 Test Results: How Much Can One PCIe Slot Really Deliver?
1. Test Results at a Glance
In short: The rear PCIe 4.0 slot reaches full performance, both USB interfaces deliver the speeds expected from their respective specifications, and the 10GbE SFP+ ports reach near line rate. Running both 10GbE ports simultaneously delivers around 19 Gbit/s total. Temperatures remain controlled with passive cooling, and the card worked normally across three different motherboard platforms.
2. Test Setup
3. Rear M.2 Slots (PCIe 4.0 x4): 5190–5261 MB/s Read
The two rear M.2 slots use PCIe 4.0 x4 and serve as the main storage interfaces on the card.
AJA System Test produced the following results:
Read speeds remained between 5190 and 5261 MB/s, which is within the normal performance range of PCIe 4.0 x4 SSDs.
This indicates that the M.2 interface itself is not becoming a bottleneck, allowing the SSDs to deliver their available performance.
Side-by-side comparison: rear M.2 (3039/5261 and 4103/5190), Type-E (1000/985), and rear-bracket USB-C (1880/1872).
AJA System Test was run with 1920×1080 HD-1080i, a 1GB test file, and 10-bit YUV. Different testing tools, file sizes, and access patterns can produce different results, so these figures should not be directly compared with CrystalDiskMark results.
4. USB Interface Testing: One at 20Gbps, One at 10Gbps
The card provides two USB interfaces, and each delivered performance corresponding to its specified bandwidth:
Two points are worth noting:
- The measured performance of the two interfaces differs by nearly 2× — 1880 MB/s vs. 1000 MB/s — which is consistent with the difference between their 20Gbps and 10Gbps specifications. This indicates that each interface is getting the bandwidth expected from its specification without noticeably sharing or limiting the other.
- Read and write performance are nearly symmetrical — 1880 / 1872 MB/s and 1000 / 985 MB/s — with no obvious read-heavy or write-heavy imbalance.
5. 10GbE SFP+ Testing: 9.2–9.5 Gbit/s on One Port, Around 19 Gbit/s on Two
Single Port: Stable at 9.2–9.5 Gbit/s in iperf3 Sampling
[SUM] 399.01-400.00 sec 1.07 GBytes 9.23 Gbits/sec
[SUM] 400.01-401.00 sec 1.10 GBytes 9.48 Gbits/sec
[SUM] 364.01-365.00 sec 1.10 GBytes 9.37 Gbits/sec
[SUM] 385.01-386.00 sec 1.07 GBytes 9.40 Gbits/sec For 10GbE, actual usable throughput is normally around 9.3–9.5 Gbit/s after protocol overhead.
The measured results fall within this range, indicating that the link is operating close to the practical limit of a 10GbE connection.
Two Ports Running Simultaneously: 9.4–9.8 Gbit/s Each, Around 19 Gbit/s Total
One 10G optical module was installed in each of the two front M.2 slots. Both links were tested simultaneously with iperf3.
In Windows, they were recognized as two independent Realtek PCIe 10GbE network adapters, with both links negotiating at 10.0 Gbps.
Both 10GbE adapters negotiate at 10.0 Gbps, with both links running close to line rate simultaneously.
This result is particularly useful for understanding the available upstream bandwidth.
With both 10GbE ports running simultaneously, the combined throughput reaches nearly 20 Gbit/s, while the upstream interface is PCIe 4.0 x4, providing 64Gbps of raw bandwidth.
Even with both ports running at full speed, the PCIe 4.0 x4 upstream link still has substantial bandwidth available. The two ports therefore do not need to compete heavily for the upstream connection, allowing both to operate close to line rate.
For applications such as NAS, software routers, and direct connections between multiple systems, this means the two 10GbE ports can actually be used simultaneously at near full speed rather than simply being two ports that cannot maintain performance when used together.
6. Temperature Testing: 71.7°C Maximum Under Load
An HIKMICRO infrared thermal camera was used during testing:
The thermal image shows that heat is concentrated around the main controller and heatsink area. The temperature distribution forms a relatively regular rectangular hot zone, with a gradual temperature transition toward the edges.
This is consistent with the expected behavior of a passively cooled design, rather than indicating a localized overheating issue.
One distinction is important: 71.7°C is the measured surface temperature of the device, not the operating ambient temperature.
The card's specified operating ambient temperature range is 0°C to +70°C. These are two different measurements.
The all-aluminum enclosure also acts as part of the cooling structure. The card uses no fan, which means there is no fan noise, dust accumulation, or bearing wear associated with active cooling.
7. Cross-Platform Verification on Three Motherboards
The same PB65MX4 was installed and tested on three different motherboards:
- ASUS PRIME B650M-A WIFI (AMD)
- ASUS TUF GAMING B760M (Intel)
- MSI Z890 (Intel)
Three points stand out from the cross-platform testing:
- All three motherboards recognized the card normally without additional configuration.
- The motherboard does not need to support PCIe bifurcation. This is one of the key differences between the southbridge-based design and conventional PCIe expansion cards that rely on motherboard PCIe bifurcation. It is also why the card can work across different motherboard platforms without additional configuration.
- Storage, USB, and network interfaces were all tested on the three platforms, covering both AMD and Intel systems, as well as B-series and Z-series chipsets.
For users, this means changing platforms does not necessarily mean changing the expansion card. Whether moving from AMD to Intel or upgrading from B650 to Z890, the PB65MX4 can continue to be used.