How to Choose a Bare Metal Server for Automated Workloads

Choose a Bare Metal Server that Fits Your Workload

“What’s your budget?” and “What automation are you running?” are two important questions you should ask first. For example, running 50 scrapers in parallel needs CPU cores and RAM, while training a machine-learning model requires a powerful GPU.

Does your automation need to run constantly or in bursts? Constant workloads get the most value from bare metal servers. But if you just need to automate sales reports, a cheaper VPS server is enough. 

Make Sure You’re Allowed to Run Your Automation

The “unlimited” part of VPS plans almost always comes with an asterisk. Many providers throttle you if you’re maxing out the CPU 24/7. Some cap average use at as little as 30% over a week. With bare metal, you don’t have to wonder whether you’re getting dedicated cores. 

Ask the provider its acceptable use policy (AUP). Can you run the CPU at 100$/7? Are scraping, scanning, or bulk automations allowed? How providers handle complaints matters too.

Let’s say they allow scraping. But one of the sites you scraped filed a complaint. Some providers suspend entire servers over a single report. Others give you notice and time to fix the issue. 

Remember that the IP addresses belong to the provider. If your automation gets their IPs blacklisted, it affects their other customers too. They have every reason to act fast. Pick a provider whose policies match how you work.

Speed per Core vs. Number of Cores

Newer CPUs, like the AMD EPYC 9454P or Intel Xeon Gold 6430, are popular mid-to-high-end server chips. Per-core speeds are fast. You’ll need that for automating tasks in sequence.

For parallel workloads, like running dozens of bots, some of the best bare metal value comes from older dual Xeons, which offer better price-to-performance. 

Don’t Skimp on Memory and Storage

Get ECC memory if possible. It detects and corrects memory errors. When deciding how much RAM you might need, use this simple formula: RAM needed = RAM per task × peak tasks × 1.3. For example, 50 automated browsers at 300 MB each need around 20 GB, so choose 32 GB.

NVMe drives are standard for servers. As a rough guide for size:

• Bots and agents that mostly use API calls get by on 250-500 GB.
• Scrapers and crawlers that store massive amounts of data might need 1 to 2 TB
• Converting videos and other media requires at least 2 TB of storage.

Check the Network, Bandwidth, and IP Addresses

Automated workloads send and receive a lot of data. How much bandwidth you need depends on what you’re using it for. As a baseline, choose a server that gives you at least 1 Gbps of bandwidth. Some providers offer 10 Gbps or more. But always ask if it’s unmetered or capped. Find out what overages cost, too.

It’s also important to check how many IPv4 addresses are included and what extras cost. Additional IPs usually run about $2–$4 per address each month, sometimes with a one-time setup fee. That adds up fast: 30 extra IPs could cost $60–$120 per month on top of your server. 

Consider Hardware Failures

Even well-maintained hardware breaks eventually. Drives, RAM, and power supplies fail. What matters is how fast the provider can fix it for you. So, ask about the following:

• Remote access (IPMI): IPMI lets you reboot, reinstall, or troubleshoot the server even when it’s frozen. It’s a must-have for experienced users.
• Replacement process: Many providers replace parts fast. But it’s your job to report it. So, set up alerts.
• Setup time: Some bare metal servers are ready in minutes. Others need to be set up manually and take hours or days.
• Support: Does the provider offer 24/7 support?

Compare Providers on More Than Price

There’s a poll on LowEndTalk comparing two popular bare metal providers—Hetzner and OVH. More than 70% of the votes went to Hetzner. Price wasn’t the main reason. It was simpler support, no port caps, and fair IP pricing. 

But these might not be the features you specifically need. If you’ve got a shortlist of providers, you can judge your decision on the following:

• Annual cost of running the server: Add up server costs, setup fees, extra IPs, backups, and bandwidth for a full year. Introductory deals and add-ons can change which provider offers the best price-to-performance ratio. 
• Real-world performance: Ask for a trial. Most don’t offer one but have money-back guarantees. Use that period to test automations, then monitor performance. 
• Support response time: How fast can a provider get back to you? Do they have 24/7 support? How do they handle urgent issues or hardware failures?
• Recent reviews: Search hosting forums for the provider’s name plus words like “down,” “suspended,” or “support” from the last 6–12 months. 

Be realistic about the total cost. Moving from the cloud to bare metal saves you money in the long term. But you also have to take on monitoring, maintenance, and backups. You might need to factor in the cost of hiring someone to manage the server. 

Key Takeaways

The best bare metal server for automated workloads depends on your needs and the types of automation you want to run. To recap:

• Know how many automations you’ll run, how often they run, and the resources used.
• Confirm that your provider allows the type of automation you want to run
• Go for newer chips if you want faster per-core performance.
• Older, high-core-count chips have better price-to-performance for parallel workloads. 
• Check port speed, bandwidth limits, port caps, IP pricing, and server location.

Special thanks to Imran Ishaq for contributing to this article.