Optimizing MT5 on a VPS: Squeeze Every Drop of Performance

Your VPS is not a desktop PC. Nobody is sitting in front of it admiring smooth window animations or browsing the news feed. Every CPU cycle, every megabyte of RAM, and every millisecond of network latency either contributes to your trading or it does not. This guide is about stripping away everything that does not.

We run thousands of MT5 terminals across our infrastructure. We have seen every configuration mistake there is, and we have spent years tuning setups for traders running everything from a single conservative swing strategy to twelve aggressive scalping terminals on one box. What follows is everything we have learned, distilled into the settings and practices that actually matter.

1. Windows Server Optimization

Windows Server ships configured for general-purpose workloads. A trading VPS is not a general-purpose workload. Start here before you even open MT5.

Disable Visual Effects

Open System Properties > Advanced > Performance Settings (or run SystemPropertiesPerformance.exe) and select Adjust for best performance. This disables Aero glass, window animations, smooth scrolling, font smoothing, and every other visual nicety that costs GPU and CPU time for zero trading benefit. On a headless VPS you will never notice the difference visually, but you will notice it in your resource usage.

From an elevated command prompt, you can also disable animations directly via the registry:

reg add "HKCU\Control Panel\Desktop\WindowMetrics" /v MinAnimate /t REG_SZ /d 0 /f
reg add "HKCU\Software\Microsoft\Windows\CurrentVersion\Explorer\VisualEffects" /v VisualFXSetting /t REG_DWORD /d 2 /f

Disable Unnecessary Services

Open services.msc and set the following to Disabled:

Print Spooler (Spooler) — you are not printing from your VPS.
Windows Search (WSearch) — the indexing service hammers disk I/O. You do not need full-text search on a trading server.
SysMain (formerly Superfetch) — preloads applications into memory based on usage patterns. On a VPS where the workload is constant and predictable, this wastes RAM and causes random I/O spikes.
Windows Update (set to manual, not disabled) — you want to control when updates install, not have them restart your terminal mid-session. Schedule updates for weekends when markets are closed.
Themes — disables desktop composition. Saves a small but measurable amount of memory.

From PowerShell, you can disable these in one shot:

$services = @('Spooler','WSearch','SysMain','Themes')
foreach ($svc in $services) {
    Stop-Service -Name $svc -Force -ErrorAction SilentlyContinue
    Set-Service -Name $svc -StartupType Disabled
}

Power Plan and Pagefile

Set the power plan to High Performance so the CPU does not downclock:

powercfg /setactive 8c5e7fda-e8bf-4a96-9a85-a6e23a8c635c

Disable the screen saver entirely — it serves no purpose and can consume CPU when it activates:

reg add "HKCU\Control Panel\Desktop" /v ScreenSaveActive /t REG_SZ /d 0 /f

For the pagefile, set it to a fixed size rather than system-managed. A good rule is 1.5x your physical RAM. On a 4GB VPS, set it to 6144MB. A fixed pagefile avoids the overhead of dynamic resizing and prevents fragmentation. Configure this in System Properties > Advanced > Performance Settings > Advanced > Virtual Memory.

2. MT5 Terminal Tuning

A fresh MT5 installation is configured for retail desktop use. It downloads history for every symbol, streams news, and holds far more data in memory than any algo strategy needs.

Limit Chart History

Go to Tools > Options > Charts and set Max Bars in Chart to the minimum your strategy actually requires. The default is often 100,000 or unlimited. If your strategy only looks back 500 bars, set it to 1000 (some buffer for indicator warm-up). Each chart holds its bar data in memory, and on a 1-minute timeframe, 100,000 bars is roughly 70 days of data sitting in RAM for no reason.

On an M1 chart, reducing max bars from 100,000 to 5,000 can save 30-80MB of RAM per chart. Across ten charts, that is 300-800MB recovered.

Disable the News Feed

Tools > Options > Community — uncheck everything. Then Tools > Options > Server — uncheck Enable news. The news feed creates a persistent connection and a constant trickle of data that your EA does not read. Kill it.

Clean Up Market Watch

Right-click in Market Watch and select Hide All. Then manually add back only the symbols your EAs actually trade. Every symbol in Market Watch maintains a live price feed. If your broker offers 500 symbols and you trade 5, you are processing 99% garbage data. This is one of the single biggest performance wins most traders overlook.

Close Unused Charts

Every open chart is an active data consumer. If you have charts open “for monitoring” that no EA is attached to, close them. Use a separate lightweight tool or your broker’s web platform if you want to watch prices.

Use Appropriate Timeframes

An M1 chart processes 60x more ticks per hour than an H1 chart. If your strategy runs on H4, do not attach it to an M1 chart just because you want to see more detail. The EA executes on the timeframe of the chart it is attached to. Match them correctly.

3. Memory Management

MT5 is a 64-bit application, so it does not hit the 2GB per-process wall that plagued MT4. That said, it can still consume surprising amounts of memory if unchecked.

Real-World Memory Numbers

From our measurements across production terminals:

MT5 base process, no charts: ~80-120MB
Each chart with default history: +50-150MB depending on timeframe and max bars
Each custom indicator: +10-50MB depending on complexity and buffer sizes
Each running EA: +5-30MB depending on whether it maintains internal data structures
A typical terminal with 5 charts, 2 indicators each, 1 EA: 300-500MB
A heavy terminal with 15 charts, multiple indicators, complex EAs: 800MB-1.5GB

Planning Your RAM

For a single terminal running a straightforward strategy on 3-5 charts, 2GB of VPS RAM is workable but tight once you account for Windows overhead (~800MB-1.2GB). We recommend 4GB as the baseline for comfortable single-terminal operation.

For multiple terminals, add 400-600MB per additional lean terminal. A 3-terminal setup with moderate chart counts wants 6-8GB. If you are running anything memory-intensive — large tick databases, machine learning indicators, or EAs that cache extensive history — you need more.

Monitoring

Open Task Manager > Details, right-click the column headers, and add Working Set (Memory) and Peak Working Set. The peak tells you the high-water mark. If your peak is within 80% of your physical RAM, you are in the danger zone — the system will start paging to disk, and your tick processing latency will spike.

For ongoing monitoring, use Performance Monitor (perfmon.exe). Add counters for Process > Working Set - Private for each terminal64.exe instance, and Memory > Available MBytes for the system. Set up an alert if available memory drops below 500MB.

4. CPU Optimization

MT5 is more CPU-efficient than MT4, particularly because it can distribute indicator calculations across cores. But certain operations will still max out a core.

What Burns CPU

Custom indicators on every tick. If your indicator recalculates its entire buffer on each tick rather than only the current bar, it will hammer one core. This is the number one CPU issue we see. Poorly written indicators are the single biggest performance killer.
Tick processing in EAs. An EA attached to a high-tick-volume symbol (like EURUSD or NAS100 during US session) using OnTick() to run heavy logic will spike CPU. Offload non-urgent calculations to OnTimer() if your strategy allows it.
History downloads and optimization. The built-in strategy tester is multi-threaded and will consume every core you have. Never run optimizations on a live trading VPS.

Affinity and Priority for Multiple Terminals

If you are running multiple terminals, you can pin each one to specific CPU cores to prevent them from competing. Open Task Manager, go to Details, right-click a terminal64.exe process, and select Set Affinity. On a 4-core VPS with 2 terminals, assign cores 0-1 to the first terminal and cores 2-3 to the second.

For automated affinity on startup, use PowerShell:

# Set affinity for a process — value is a bitmask (3 = cores 0,1; 12 = cores 2,3)
$proc = Get-Process -Name terminal64 | Select-Object -First 1
$proc.ProcessorAffinity = 3

Set process priority to Above Normal (not Realtime — that can starve system processes):

$proc.PriorityClass = 'AboveNormal'

When to Upgrade

If your CPU is consistently above 70% during active market hours, your tick processing is falling behind. You are likely missing ticks, which means your indicators are working with incomplete data and your entry signals may be delayed. This is the point where you need more cores or faster cores, not a workaround.

5. Network and Latency

Location Matters

The single most impactful optimization for execution speed is placing your VPS geographically close to your broker’s trade server. If your broker’s server is in London (LD4/LD5 Equinix), a VPS in New York adds 70-80ms of round-trip latency that no software tuning can eliminate. A VPS in London brings that to 1-5ms.

We offer VPS locations specifically chosen for proximity to major broker data centers. If you are unsure where your broker’s server is, open MT5, go to File > Open an Account, and look at the server ping times — or simply ask your broker.

Measuring Latency

From your VPS command prompt:

ping your.broker.server.com -n 20

What the numbers mean for different strategy types:

Under 5ms: Excellent. Suitable for any strategy including high-frequency scalping.
5-20ms: Good. Fine for most algorithmic strategies, including aggressive scalping.
20-50ms: Acceptable for intraday strategies. Scalpers trading on tight targets will notice slippage.
50-100ms: Only suitable for swing or position strategies. Do not scalp on this latency.
Over 100ms: Relocate your VPS. This latency will cause measurable slippage on any active strategy.

What “Low Latency” Actually Means

Marketing materials love the phrase “low latency” without defining it. In practice, for retail algorithmic trading, latency below 10ms to your broker is genuinely low. Latency below 1ms requires co-location (your server in the same data center as the broker’s matching engine), which is a different product category entirely.

The impact of latency compounds with trade frequency. If you take 2 trades per day on H4, the difference between 5ms and 50ms is irrelevant. If you take 200 trades per day scalping 3-pip moves, that same difference is the gap between profitability and death by slippage.

6. Multiple Terminal Setup

Running several MT5 instances on one VPS is common and perfectly viable if done correctly.

Portable Mode

Install each MT5 instance in a separate directory and launch with the /portable flag. This forces the terminal to store all data (configuration, history, logs) in its own installation directory rather than the shared %APPDATA% location. This prevents terminals from interfering with each other.

"C:\MT5_Terminal1\terminal64.exe" /portable
"C:\MT5_Terminal2\terminal64.exe" /portable
"C:\MT5_Terminal3\terminal64.exe" /portable

Create separate shortcuts or batch files for each. If you need them to start automatically on boot, create scheduled tasks triggered At startup with a 30-second delay between each (to avoid all terminals hammering the disk simultaneously during initialization).

$action = New-ScheduledTaskAction -Execute "C:\MT5_Terminal1\terminal64.exe" -Argument "/portable"
$trigger = New-ScheduledTaskTrigger -AtStartup
$settings = New-ScheduledTaskSettingsSet -AllowStartIfOnBattery -DontStopIfGoingOnBatteries
Register-ScheduledTask -Action $action -Trigger $trigger -Settings $settings -TaskName "MT5_Terminal1" -User "SYSTEM" -RunLevel Highest

Resource Guidelines Per Terminal Count

Terminals	Minimum RAM	Recommended RAM	Minimum Cores
1	2 GB	4 GB	2
2	4 GB	6 GB	2
3	6 GB	8 GB	4
4	8 GB	12 GB	4
5+	12 GB	16 GB	6

These assume moderate chart counts (5-8 per terminal) and reasonably efficient EAs. Heavy setups with dozens of charts and complex indicators per terminal will need more.

7. Monitoring and Maintenance

A VPS that you set up once and never check is a VPS that will eventually cost you money.

Detecting Frozen Terminals

MT5 terminals can freeze — no crash, no error, just a silent stop processing ticks. Create a simple monitoring script that checks whether the terminal process is responsive and whether the log file is still being written to:

# Check if terminal log was updated in the last 5 minutes
$logPath = "C:\MT5_Terminal1\Logs"
$latest = Get-ChildItem $logPath -Filter "*.log" | Sort-Object LastWriteTime -Descending | Select-Object -First 1
if ($latest.LastWriteTime -lt (Get-Date).AddMinutes(-5)) {
    # Terminal may be frozen — restart it
    Stop-Process -Name terminal64 -Force
    Start-Sleep -Seconds 5
    Start-Process "C:\MT5_Terminal1\terminal64.exe" -ArgumentList "/portable"
}

Schedule this script to run every 5 minutes via Task Scheduler during market hours.

Scheduled Restarts

Memory leaks in EAs and indicators are more common than most developers admit. A weekly restart of each terminal during a market-closed period (Saturday or Sunday) prevents slow memory creep from eventually crashing the process. Automate it with a scheduled task that kills the terminal, clears the terminal’s memory-mapped files, and restarts it.

Disk Space Management

MT5 accumulates logs, journal entries, and historical data over time. The Logs directory in particular can grow to gigabytes if left unchecked. Set up a scheduled task to purge log files older than 30 days:

Get-ChildItem "C:\MT5_Terminal1\Logs" -Filter "*.log" |
    Where-Object { $_.LastWriteTime -lt (Get-Date).AddDays(-30) } |
    Remove-Item -Force

Also monitor overall disk usage. When a VPS disk fills up, MT5 cannot write to its journal and will behave unpredictably. Keep at least 2GB free at all times.

8. When to Scale Up

There is a point where tuning cannot compensate for insufficient hardware. Here is how to recognize it.

Signs Your VPS Is Underpowered

Missed ticks. If your EA logs show gaps in tick data during active sessions, your CPU or network cannot keep up with the tick rate.
Delayed order execution. If the time between your EA calling OrderSend() and receiving a response is consistently above 200ms (and your network latency is fine), the system is bottlenecked — likely CPU or disk I/O from excessive paging.
CPU sustained above 80%. Open Task Manager during the busiest trading session (London/New York overlap). If CPU usage stays above 80% for extended periods, you are on borrowed time.
Available RAM below 500MB. At this level, Windows is actively paging, and your tick processing is running through the disk instead of memory.
Disk queue length above 2. Check in Performance Monitor under PhysicalDisk > Current Disk Queue Length. Sustained values above 2 indicate I/O bottleneck, usually from insufficient RAM causing excessive paging.

More RAM vs More CPU

High memory usage, low CPU: Add RAM. This is the most common scenario — too many charts, too much history, or too many terminals for the available memory.
High CPU, low memory usage: Add cores or get faster cores. You likely have CPU-intensive indicators or very high tick-rate symbols.
Both high: You have outgrown this VPS tier. Scale up across the board.

When to Split Across Multiple VPS Instances

If you are running 4+ terminals and your total resource consumption exceeds what a single VPS can comfortably provide, splitting across two smaller VPS instances often outperforms one large one. Each VPS runs independently, so a crash or freeze on one does not take out everything. This is simple redundancy and it has saved more traders than any single optimization trick.

The rule of thumb: if losing your VPS for one hour would cost you more than a month of a second VPS, get the second VPS.

The Bottom Line

VPS optimization is not glamorous. Nobody posts about it on trading forums. But the difference between a well-tuned VPS and a default installation is the difference between your EA executing at the price it intended and executing 2 pips worse on every single trade. Over hundreds of trades, that adds up to real money.

Start with the Windows optimizations — they take ten minutes and the payoff is immediate. Then strip down your MT5 terminals to only what they need. Set up monitoring so you know when something goes wrong before it costs you. And be honest about when your hardware is no longer enough.

If you want us to do all of this for you, every FXVPS plan includes a pre-optimized Windows Server image with these settings already applied. But now you know exactly what we did and why.