If you work with a NAS, an editing PC, or large amounts of data at home, classic Gigabit Ethernet quickly becomes a bottleneck. In my case, the PC and NAS were originally connected with 1 Gbit/s. That is perfectly fine for normal file storage, backups, and office data. But once large video files, project folders, or several hundred gigabytes are involved, 1Gbit becomes limiting very quickly.

In practice, 1Gbit Ethernet usually delivers around 110 MB/s. For large data sets, that is slow. A 100 GB project can take roughly 15 minutes or more to transfer. The goal was therefore a cost-effective upgrade to 10Gbit using fiber or SFP+.

Initial Situation

• NAS and PC were previously connected via 1Gbit/s.
• Practical data rate: around 110 MB/s.
• Large video and project files needed to move faster between PC and NAS.
• The upgrade had to remain affordable while also serving as an entry point into 10G-SFP+/fiber technology.

Goal of the Upgrade

The goal was a direct 10Gbit connection between the editing PC and the TrueNAS system using an affordable 10G-capable switch. The solution had to remain budget-friendly for a homelab while still delivering measurable real 10G performance.

Hardware Used

Component	Description	Cost
10G network cards	2× Intel X520 PCIe 10G cards from 10Gtek	approx. €100
SFP+ modules	2× 10Gtek 10G SFP+ modules	approx. €29
Switch	SODOLA SL-8T2XS-WEB	approx. €79
Cable	OM4 fiber cable, 50 m	approx. €35
Total	10G link PC ↔ NAS	approx. €243

The Switch: SODOLA SL-8T2XS-WEB

The SODOLA SL-8T2XS-WEB is an affordable web-managed Layer 2 switch with eight 2.5G RJ45 ports and two 10G SFP+ ports. This makes it a useful small homelab switch when you want to connect a few multi-gigabit devices and add two fast SFP+ links.

Feature	Specification
Model	SODOLA SL-8T2XS-WEB
Switch type	Web Managed Layer 2
RJ45 ports	8× 100M/1G/2.5G
SFP+ ports	2× 10G SFP+
VLAN	802.1Q VLAN, tagged/untagged ports
Additional features	QoS, IGMP Snooping, port statistics, link aggregation depending on firmware
Layer 3 / Routing	No, it should be considered a Layer 2 switch

Important: This switch is not a router and not a firewall. It can handle VLANs, trunks, and access ports. Routing between VLANs, firewall rules, DHCP, and NAT belong on a router or firewall, for example OPNsense.

Why SFP+ and Fiber?

10Gbit can also be done over copper. For my setup, however, SFP+ with fiber was the more attractive solution. SFP+ modules and fiber cables usually remain thermally more relaxed than 10GBASE-T modules over RJ45. Fiber is also electrically isolated and well suited if longer runs between rooms or floors are needed later.

For short homelab distances, 10G SFP+ modules and suitable fiber patch cables are more than enough. In this test, the PC and TrueNAS were connected via 10G SFP+ through the SODOLA switch.

Test Setup

Area	Description
Client	Windows PC / editing PC
Server	TrueNAS SCALE
Network	10G SFP+ via SODOLA SL-8T2XS-WEB
Measurement tools	iPerf3, CrystalDiskMark, SMB copy test, switch port counters, TrueNAS ethtool/ip

Test Results

1. iPerf3: PC to TrueNAS

The first test checked the direction from the Windows PC to the TrueNAS system.

Direction	Result
PC → TrueNAS	approx. 9.27–9.41 Gbit/s

This means the 10G link was almost fully utilized in this direction.

2. iPerf3: TrueNAS to PC

In the opposite direction, a single TCP stream did not use the full bandwidth. With multiple parallel streams, however, the connection scaled cleanly.

Direction	Streams	Result
TrueNAS → PC	1	approx. 1.3–1.7 Gbit/s
TrueNAS → PC	4	approx. 6.18 Gbit/s
TrueNAS → PC	8	approx. 9.49 Gbit/s

The result shows that the physical 10G link is fast enough. The limitation with a single stream is more likely related to TCP, drivers, receive queues, or operating system behavior.

3. Bidirectional iPerf3 Test

In the bidirectional test, both directions were loaded at the same time.

Direction	Result
PC → TrueNAS simultaneously	approx. 8.95–8.97 Gbit/s
TrueNAS → PC simultaneously	approx. 9.43–9.46 Gbit/s

This means the connection was able to deliver nearly full-duplex 10G performance in the test. Under maximum load, TCP retransmits occurred, but there were no port errors on the switch. This points more toward buffer, TCP, or driver behavior under full load than toward a physical issue.

4. Switch Port Counters

After the load tests, the error counters of the SFP+ ports were checked.

Port	Link	Send Errors	RX Errors	Assessment
Port 9	10Gbps Full	0	0	clean
Port 10	10Gbps Full	0	0	clean

Even after high data volumes and bidirectional tests, the 10G ports remained error-free.

5. CrystalDiskMark on the Network Drive

A CrystalDiskMark test was also performed on the TrueNAS network drive.

Test	Read	Write
SEQ1M Q8T1	approx. 1068 MB/s	approx. 698 MB/s
SEQ128K Q32T1	approx. 797 MB/s	approx. 599 MB/s
RND4K Q32T16	approx. 72 MB/s	approx. 39 MB/s
RND4K Q1T1	approx. 10.6 MB/s	approx. 8.5 MB/s

For large video files and sequential transfers, the SEQ values are the most relevant. The achieved values are very solid for an affordable 10G homelab setup.

6. SMB Copy Test and SMB Multichannel

When copying from the NAS to the local NVMe SSD of the PC, the speed initially reached only around 279 MB/s. The analysis showed that SMB Multichannel was disabled on TrueNAS. After enabling it, real-world copy performance improved significantly.

State	NAS → PC
SMB Multichannel disabled	approx. 279 MB/s
SMB Multichannel enabled	approx. 593 MB/s

Windows then confirmed an active SMB Multichannel connection. This is an important point for practical use: the network hardware can deliver 10G, but SMB and operating system settings also decide how much of that performance actually arrives in daily use.

7. TrueNAS NIC Statistics

The network interface on TrueNAS was also checked. The relevant error counters remained clean.

Counter	Result
RX errors	0
TX errors	0
RX CRC errors	0
RX frame errors	0
RX FIFO errors	0
RX missed errors	0
RX no buffer count	0

This showed no signs of physical errors on the fiber link, SFP+ modules, or network card.

8. Temperature Under Load

During the load tests, the switch reported around 63 °C. For a compact and affordable 10G SFP+ switch under heavy load, this is warm but was not critical in the test. There were no link drops, no throttling, and no port errors.

What Did the Upgrade Deliver?

Scenario	Speed
Before: 1Gbit NAS to PC	approx. 110 MB/s
After: SMB copy NAS to PC	approx. 593 MB/s
After: CrystalDiskMark read	up to approx. 1068 MB/s
After: iPerf3 net throughput	up to approx. 9.5 Gbit/s

This significantly increased real-world copy performance compared to 1Gbit. Instead of around 110 MB/s, several hundred MB/s are now possible in daily use. Under optimal conditions, read speeds over the network drive can even reach around 1 GB/s.

Conclusion

The affordable entry into 10Gbit via SFP+ and fiber was worth it. For around €243, it was possible to build a fast connection between the PC and TrueNAS that reaches nearly full 10Gbit performance in iPerf3 and is also significantly faster than classic 1Gbit networking in real-world SMB transfers.

The SODOLA SL-8T2XS-WEB is not an enterprise core switch. The web interface is simple and the firmware feels budget-oriented. However, the raw switching performance in the test was convincing: the 10G ports remained error-free even under load.

Important: 10Gbit does not end when the fiber cable is plugged in. To achieve good real-world results, SMB, drivers, the operating system, NAS configuration, and storage performance also have to be configured properly. In this test, enabling SMB Multichannel on TrueNAS delivered a significant performance increase.

For homelab use, video editing, large file storage, and fast NAS access, this solution is very attractive from a price-performance perspective. Anyone who needs maximum stability, centralized management, and long-term vendor support should consider higher-end switches. But for an affordable entry into 10G SFP+, this setup delivers a strong result.

This article contains affiliate links. The price remains the same for you, and we receive a small commission for the recommendation.

The HP EliteDesk 800 G6 Mini is a compact and capable platform for small virtualization environments. With a modern Intel CPU, low power consumption and a small form factor, it is well suited as an edge server, lab system, home server or small business virtualization host.

1. Concept: Running Proxmox VE on the EliteDesk G6 Mini

The EliteDesk 800 G6 Mini provides a solid foundation for Proxmox VE:

• Intel Core i5 / i7 10th generation CPU
• up to 64 GB DDR4 RAM
• NVMe SSD storage
• very low power consumption
• compact form factor

Typical use cases include:

• small Proxmox environments
• local infrastructure at customer sites
• Windows VMs for business software
• monitoring with Checkmk, Prometheus or Grafana
• backup, storage or proxy services

2. The Problem: Intel I219-LM Network Controller

The EliteDesk 800 G6 Mini commonly uses an integrated network controller called Intel Ethernet Connection I219-LM. Under Linux, this NIC is handled by the e1000e driver.

In combination with Proxmox VE, this controller may occasionally cause network instability.

Typical symptoms include:

• short network interruptions
• the Proxmox host becoming temporarily unreachable
• virtual machines losing network connectivity
• monitoring systems reporting host outages

The kernel log may show messages such as:

e1000e: Detected Hardware Unit Hang
NETDEV WATCHDOG: transmit queue timed out

3. Technical Explanation

The Intel I219-LM is not a traditional server-grade PCIe network controller. It is an onboard controller closely integrated with the chipset. The Linux driver uses several hardware offloading features such as TSO, GSO, GRO and Energy Efficient Ethernet.

Under certain conditions, especially in virtualization environments or with specific power-saving and driver combinations, the controller may stop processing packets correctly. The driver then reports this as a Hardware Unit Hang.

4. Software Workaround

A common stabilization method is to disable selected offloading features.

Disable offloading

ethtool -K nic0 tso off
ethtool -K nic0 gso off
ethtool -K nic0 gro off

Disable Energy Efficient Ethernet

ethtool --set-eee nic0 eee off

To make these settings persistent in Proxmox, add them to /etc/network/interfaces:

auto nic0
iface nic0 inet manual
    post-up ethtool -K nic0 tso off gso off gro off
    post-up ethtool --set-eee nic0 eee off

Then reload the network configuration:

ifreload -a

These changes can significantly improve stability. However, they should be considered a workaround rather than a final fix.

5. Hardware Solution: Additional Network Interface

For production systems, a dedicated network interface is the cleaner solution. Depending on the specific model and adapter, the EliteDesk 800 G6 Mini can be expanded internally with an additional network interface.

Recommended controller families include:

• Intel i210
• Intel i350

These controllers are widely used in server environments and are well supported by Linux.

Benefits of a dedicated NIC:

• more stable Linux operation
• fewer power-management related issues
• better suitability for virtualization
• cleaner separation of management and VM networking

6. Conclusion

The HP EliteDesk 800 G6 Mini is a very interesting platform for small Proxmox deployments. It is compact, energy-efficient and powerful enough for many infrastructure tasks.

However, the integrated Intel I219-LM network controller can be problematic under Linux and Proxmox. Software-side adjustments can reduce the risk by disabling selected offloading and power-saving features.

For production environments, I recommend using a dedicated Intel network interface, ideally with an i210 or i350 controller. This turns the compact business PC into a stable and efficient virtualization platform.

Fraudulent payment requests are currently circulating, allegedly from the debt collection service provider EOS. Some of these fake letters are professionally designed, but have features that indicate an attempt at fraud: for example, false judge's names are given or an Italian IBAN is given for the transfer.

EOS itself warns against these e-mails on its website and offers those affected the opportunity to check their authenticity using the claim number. If this 11-digit claim number is missing from the letter, this is another clear indication of attempted fraud.

Recently, some security holes in the Exchange servers have been closed, but the fraud attempts continue.
Heise.de reports a fraud attempt in which previously stolen e-mail communication was used to initiate a payment by the recipient via a typo-squatting domain in which, for example, numbers are substituted for letters (he1se.de).