Antares

Introduction#

In 2025, the cloud server market is vibrant and competitive, with various manufacturers and cloud service providers focusing on MC flourishing. Various cloud server solutions have already secured a place in the MC server field due to their convenience, elastic scalability, and technical support advantages. So, in this increasingly rich cloud solution landscape, do we still need to build a physical machine to set up an MC server?
Life goes on, and so does the tinkering. As an MC player and a tinkerer, let's play with the ancient relic of the blue giant—Intel Xeon E5-2696 v3.

Note: The core discussion of this article is to build a small MC server that can run stably 24/7 on a low budget.

Hardware Configuration and Testing Environment#

The smooth operation of an MC server largely depends on the performance of the CPU, and MC is a game that heavily relies on single-core performance. Therefore, we need to choose a processor with a higher clock frequency. In comparison, the E5-2666 V3 is actually more suitable for running an MC server and is cheaper, but since I have a need for a home server that requires features like running multiple virtual machines, I chose the E5-2696 V3, which has stronger multi-core performance.

Notes:

• CPU Base Frequency: The base clock frequency of the CPU.
• Single-Core Turbo Frequency: The maximum frequency the CPU can reach under single-core load.
• All-Core Frequency: The frequency the CPU can reach when all cores are fully loaded.
• TDP (Thermal Design Power): The maximum heat generated by the CPU under normal working load. The higher the value, the stronger the cooling capability required.

Cautions:

• In practical use, the E5 series CPUs often cannot stably run at the single-core maximum turbo frequency in most application scenarios. If the server's load is slightly higher, the CPU frequency will drop to the all-core frequency. My suggestion is to look directly at its all-core frequency rather than the single-core turbo frequency.
• This test does not use any methods to boost CPU performance.
• "Affected by Trump's reciprocal tariffs," the price of the E5-2666 V3 rose from 43 yuan in March to 73 yuan in April.

Hardware Configuration List#

The power supply and case are leftover from my previous builds. Since the server will install the Proxmox system, daily management can be done via the web, so the graphics card is only needed during system installation and can be temporarily borrowed from the main machine.

Software Environment Description#

Proxmox VE is a free and open-source server virtualization platform that allows you to run multiple independent "virtual computers" (virtual machines or containers) on a single physical machine, thereby improving resource utilization and reducing costs. You can easily create, manage, and back up these virtual environments, achieve application isolation, and even migrate them to other servers without stopping services. Proxmox VE has two virtualization "tricks": virtual machines (KVM, capable of running various operating systems) and containers (LXC, a lighter-weight Linux environment), enabling both individuals and enterprises to manage servers more efficiently.

Due to the use of virtual machines, there may be a certain performance drop (about 5% performance loss). If it's just for running an MC server, other systems can be used for better performance.

This test will run Windows Server 2022 in a Proxmox VE virtualized environment, using Java 22 and Paper 1.21.4 as the Minecraft server core, allocating a maximum of 16GB of memory to the Java process.

Testing Environment Details#

Client Environment: Minecraft 1.21.4 (Fabric version) with Spark installed for performance analysis.

Server Stress Test Content:

• Single-player maximum view distance creative mode flying map
• Single-player long-distance teleportation
• Million-boat mob farm

Comparison Group: To better assess server performance, a Xiaomi gaming laptop (i5-7300HQ, 16GB memory, priced around 1200 on a certain platform) will be used for comparative testing.

Test Results#

Map Loading Performance Comparison#

The following images show the performance comparison of E5-2696 V3 and i5-7300HQ in map loading.

E5-2696 V3:

i5-7300HQ:

TPS (Ticks Per Second)：

• Definition: TPS is the number of game logic ticks processed by the Minecraft server per second. By default, the target TPS for a Minecraft server is 20.
• Importance: TPS is a key metric for measuring server performance. If the server's TPS stabilizes around 20, players will have a smoother gaming experience. If TPS drops below 20, players will experience lag and delays.
• Statistical Metrics:
• 1m (1 Minute Average TPS): The average TPS value over the past minute, reflecting the server's performance over a short time.
• 5m (5 Minute Average TPS): The average TPS value over the past five minutes, reflecting the server's performance over a medium time, smoothing out short-term fluctuations.
• 15m (15 Minute Average TPS): The average TPS value over the past fifteen minutes, reflecting the server's performance over a longer time, more stable and reliable.

MSPT (Milliseconds Per Tick)：

• Definition: MSPT is the average number of milliseconds required for the server to process one game logic tick. MSPT can more intuitively reflect the performance pressure on the server.
• Calculation Method: MSPT = 1000 milliseconds / TPS
• Importance: The lower the MSPT, the better the server's performance. Ideally, MSPT should be below 50 milliseconds.
• Statistical Metrics:
• min (Minimum): The minimum value of MSPT during the statistical period, reflecting the server's performance in optimal conditions.
• med (Median): The median value of MSPT during the statistical period, which can eliminate the influence of extreme values and more accurately reflect the server's average performance.
• 95%ile (95th Percentile): During the statistical period, 95% of the ticks have an MSPT below this value, reflecting the server's performance in most cases.
• max (Maximum): The maximum value of MSPT during the statistical period, reflecting the server's performance in the worst conditions.

The map running test lasted about 2 minutes, so we focus on the average TPS within 1 minute. The test data shows that the average TPS of both CPUs can stabilize at 20. In terms of MSPT, the 95%ile value for E5-2696 V3 is 17.9ms, while the 95%ile value for i5-7300HQ is 6.62ms; however, the MSPT Max value for E5-2696 V3 is 280ms, lower than i5-7300HQ's 410ms. But in actual testing, the map loading smoothness of E5-2696 V3 is significantly better than that of i5-7300HQ. The i5-7300HQ has a problem where the map loading speed lags behind the creative flying speed, requiring frequent pauses to wait for chunks to load, while E5-2696 V3's map loading speed always exceeds the creative flying speed, avoiding situations where it flies into unloaded chunks.

Long-Distance Teleportation Performance Comparison#

The following images show the performance comparison of E5-2696 V3 and i5-7300HQ after long-distance teleportation.

E5-2696 V3:

i5-7300HQ:

In long-distance teleportation, the performance of i5-7300HQ is poor, with chunk loading taking a very long time. The TPS of i5-7300HQ cannot stabilize at 20, dropping to 16.72, while the MSPT Max value skyrockets to 5770ms, and the server also shows an overload warning. In contrast, the performance of E5-2696 V3 is relatively stable, with map loading speed still quite fast, TPS can remain stable at 20, and the MSPT Max value is 696ms, with a 95%ile value of 13.1ms (even better than the previous map running test). However, a strange phenomenon occurred in this test where the MSPT of i5-7300HQ actually decreased as the TPS decreased, which is abnormal.

Million-Boat Mob Farm Performance Comparison#

The following images show the performance comparison of E5-2696 V3 and i5-7300HQ when running a million-boat mob farm.

E5-2696 V3:

i5-7300HQ:

In this test, the i5-7300HQ performed relatively well, managing to barely stabilize TPS around 20, with a 95%ile value of 53ms, which is quite smooth. In contrast, E5-2696 V3 showed a decline under the assault of the million-boat mob farm, with the server's TPS dropping to a pitiful 13.81, and the 95%ile value also significantly rising to 91.4ms.

Power Consumption Test#

To accurately assess the actual power consumption of E5-2696 V3 and i5-7300HQ, this test used a smart socket to record their electricity usage. The following images show the power consumption comparison between E5-2696 V3 and i5-7300HQ. To visually demonstrate the power consumption differences of E5-2696 V3 under different loads, we display its idle and high-load power consumption graphs side by side.

E5-2696 V3 (Idle)	E5-2696 V3 (High Load)

i5-7300HQ:

i5-7300HQ (Idle)	i5-7300HQ (High Load)

Test Results:

• E5-2696 V3: Idle power consumption is about 90W, and it can reach around 140W under high load.
• i5-7300HQ: Idle power consumption is about 10W, and it does not exceed 50W under high load.

To visually demonstrate the impact of power consumption on costs, we will estimate the electricity costs for E5-2696 V3 and i5-7300HQ. Considering that small servers are mostly in idle state, we will use idle power consumption for calculations, not considering regional electricity price differences, tiered pricing, and peak-valley pricing, uniformly calculated at 0.75 yuan/kWh (generally, it is not this high; this value is on the high side).

Electricity Cost Estimate (based on 0.75 yuan/kWh):

• E5-2696 V3 (Idle):

  • Daily electricity consumption: 2.16 kWh
  • Daily electricity cost: 1.62 yuan
  • Annual electricity consumption: 788.4 kWh
  • Annual electricity cost: 591.3 yuan

• i5-7300HQ (Idle):

  • Daily electricity consumption: 0.24 kWh
  • Daily electricity cost: 0.18 yuan
  • Annual electricity consumption: 87.6 kWh
  • Annual electricity cost: 65.7 yuan

Conclusion#

I think this matter is ultimately subjective. Under low-cost conditions, a server built from second-hand parts is still usable, although this server may not outperform a second-hand laptop in some aspects (the laptop's built-in battery is equivalent to having a UPS).
However, it should be noted that E5-2696 V3 is not the most suitable second-hand part for running a server; actually, comparing E5-2666 V3 with i5-7300HQ would be more reasonable. But given the significant price increase of 2666, I suggest waiting and observing.
Finally, I hope this article can be helpful to you.