Skip to content
ServerSpotter

Understanding CPU, RAM, storage, and bandwidth specs

8 min read

Why server specs matter (and what they really mean)

When you compare hosting plans—VPS, dedicated servers, or cloud instances—you’ll see the same four spec categories repeated: CPU, RAM, storage, and bandwidth. They look straightforward, but hosting providers often present them in ways that are easy to misinterpret.

The goal of this guide is to translate those specs into plain language so you can answer: Will this server feel fast for my workload, and will it stay stable under load?

---

CPU: cores, threads, and what performance actually looks like

What a CPU does in a server

The CPU runs computations: generating pages, executing application code, handling database queries, compressing images, encrypting traffic (TLS), and more. CPU limits often show up as:
  • Slow page generation under load
  • High response times during traffic spikes
  • Background jobs taking too long (queues, cron tasks)

Cores vs threads (vCPU) in plain terms

  • Core: a physical processing unit inside a CPU.
  • Thread: a “virtual lane” within a core that allows it to work on multiple tasks more efficiently.
  • vCPU: what hosts usually sell you (often equals one thread, but not always).
A common mental model:
  • More cores/threads = more tasks can run at the same time.
  • Higher per-core speed = each task finishes faster (especially important for single-threaded workloads).
Example:
  • A WordPress site with lots of uncached PHP work benefits from stronger per-core performance.
  • A CI server running multiple builds in parallel benefits from more cores/threads.

CPU frequency (GHz) isn’t the whole story

Hosts sometimes highlight “3.5 GHz CPU” as a selling point. Frequency helps, but real performance depends on:
  • CPU generation/architecture (newer chips do more work per cycle)
  • Cache size
  • Sustained boost behavior (not just peak)
  • Whether your CPU is shared with noisy neighbors (on some VPS plans)

Dedicated vs shared CPU

Many beginner hosting plans provide shared CPU. You might see:
  • “2 vCPU” but performance varies widely based on neighbors
  • “Fair usage” policies or hidden throttling
What to look for in specs:
  • “Dedicated vCPU” or “CPU pinning” (more consistent)
  • Clear CPU model (e.g., AMD EPYC 7xx2/7xx3/9xx4, Intel Xeon Gold)
  • Stated CPU limits in measurable units (like “200% CPU” in containers can mean 2 full cores)
---

RAM: capacity, speed, and DDR4 vs DDR5

What RAM does

RAM is short-term working memory. It holds:
  • Application runtime state
  • Cached data (object cache, filesystem cache)
  • Database working set (indexes, frequently accessed rows)
  • Buffers for web servers, queues, etc.
When you run out of RAM, performance can collapse because the system starts swapping to disk (much slower), or processes get killed.

How much RAM do you need? (practical examples)

Very rough starting points:
  • Small static site / reverse proxy: 512 MB–1 GB
  • WordPress / small app: 1–4 GB (more if traffic or plugins are heavy)
  • Database server (MySQL/Postgres): often 4–16+ GB depending on dataset and concurrency
  • Docker host running multiple services: 4–16+ GB quickly

DDR4 vs DDR5: what it means for hosting

  • DDR4: common, good performance, widely available
  • DDR5: higher bandwidth and improved efficiency, often paired with newer CPUs
In real hosting scenarios, DDR5 can help memory-heavy workloads, but for many beginner workloads, RAM capacity and CPU quality matter more than DDR generation.

ECC RAM (bonus concept you’ll see on dedicated servers)

ECC (Error-Correcting Code) RAM detects and corrects some memory errors. It’s common in serious server hardware and beneficial for stability, especially for databases. Many VPS plans don’t specify it.

---

Storage: NVMe vs SSD vs HDD (and why IOPS matter)

Storage affects “snappiness” more than many expect

Storage isn’t just “how much space.” It’s also how fast your server can read/write data—especially small random reads/writes common in databases and app servers.

Key metrics:

  • Throughput: how fast large files transfer (MB/s or GB/s)
  • Latency: delay before data starts moving (lower is better)
  • IOPS: operations per second (critical for many small reads/writes)

HDD (spinning disks)

  • Cheapest per GB
  • High latency, low IOPS
  • Fine for backups, archives, and large sequential workloads
  • Poor for databases and busy websites

SSD (SATA SSD)

  • Much faster than HDD
  • Good general performance
  • Often limited by SATA interface (compared to NVMe)

NVMe SSD

  • Uses PCIe, much higher throughput and typically lower latency
  • Best for databases, high-traffic sites, and anything I/O-heavy
  • Often what you want if the price difference is reasonable
Practical example: If you host a database-backed site, NVMe can reduce query latency and improve response times even if CPU/RAM are unchanged.

Local vs network-attached storage (important hidden detail)

Some cloud platforms use network storage behind the scenes. It can be reliable and scalable, but performance depends on the network and storage tier.

You might see marketing terms like:

  • ��NVMe storage” (local) vs “NVMe-backed” (might still be networked)
  • “Premium SSD” vs “Standard SSD” (tiered IOPS limits)
What to look for:
  • IOPS limits (e.g., “up to 3000 IOPS”)
  • Whether storage is local or network-attached
  • RAID level on dedicated servers (RAID 1/10 improves redundancy; RAID 0 improves speed but increases risk)
---

Bandwidth vs speed: monthly transfer, port speed, and real throughput

Bandwidth (hosting plan) usually means monthly data transfer

Hosts often say “10 TB bandwidth” or “unmetered bandwidth.” Typically this is monthly transfer allowance.

  • 10 TB/month = you can send/receive up to ~10 terabytes in a month before fees/throttling.
  • “Unmetered” usually means no fixed cap, but still subject to fair use or a maximum port speed.

Speed usually means port speed (network link)

You’ll also see:
  • 1 Gbps port
  • 10 Gbps port
  • 100 Mbps port
This is your maximum instantaneous network rate.

Key relationship:

  • Monthly bandwidth = how much total data you can move
  • Port speed = how fast you can move data at any moment

Example: does 1 Gbps mean my downloads are always fast?

Not necessarily. Real-world throughput can be lower due to:
  • Network congestion
  • Routing distance to your users
  • Provider quality/peering
  • Your server’s CPU (encryption, compression)
  • Application limits (single-thread downloads, disk read speed)

Converting bandwidth to intuition

A 1 Gbps port running flat out 24/7 can theoretically transfer around 300+ TB per month. So if a plan offers 1 Gbps but only 10 TB/month, the cap is the monthly transfer, not the port.

---

How to read hosting specs (a practical checklist)

1) Identify what “CPU” actually means

Look for:
  • Number of vCPU
  • Whether CPU is dedicated
  • CPU model or generation (if stated)
  • Any “burstable” notes (you may get spikes but not sustained performance)
If it only says “2 vCPU” with no context, assume it’s shared and benchmark-dependent.

2) Check RAM capacity first, then generation

  • Is it 2 GB or 8 GB? That matters a lot.
  • DDR4 vs DDR5 matters less than having enough RAM to avoid swapping.

3) Treat storage as performance + size

  • Prefer NVMe for most application servers
  • Look for IOPS or storage tier info if you’re running databases
  • Understand backups: storage redundancy (RAID) is not the same as backups

4) Separate monthly transfer from port speed

A spec line might say:
  • “1 Gbps uplink, 10 TB traffic”
That means you can burst quickly but you cannot transfer unlimited data.

---

Common marketing tricks (and how not to get fooled)

“Up to” performance claims

“Up to 3.5 GHz” or “up to 10 Gbps” can be technically true but not sustained. Ask:
  • Is that guaranteed or best-case?
  • Is it shared?
  • Are there fair-use clauses?

Counting threads as cores

Some providers advertise “8 cores” when they mean “8 threads/vCPU.” Threads can help, but they are not identical to physical cores.

“Unlimited bandwidth”

Usually means unmetered monthly transfer but limited by port speed and fair use. Always check the acceptable use policy.

“NVMe” with tiny fine print

Sometimes only the cache layer is NVMe, or only certain nodes use NVMe. Look for:
  • Guaranteed storage type
  • Storage tier/IOPS limits
  • Reviews/benchmarks from users

Overselling and contention

On shared platforms, your neighbor’s workload can affect you. Clues include:
  • Very low pricing for high specs
  • No mention of dedicated CPU resources
  • Vague “enterprise hardware” statements without specifics
---

Putting it together: picking specs for common beginner scenarios

A small website or blog (WordPress, Ghost, small CMS)

  • CPU: 1–2 vCPU (dedicated if possible)
  • RAM: 2–4 GB
  • Storage: NVMe preferred, 20–80 GB depending on media
  • Bandwidth: depends on traffic; 1–2 TB/month is plenty for many small sites

A small app + database on one server

  • CPU: 2–4 vCPU
  • RAM: 4–8 GB (databases like memory)
  • Storage: NVMe strongly recommended
  • Bandwidth: consider API usage and file downloads

File hosting or media-heavy site

  • CPU: modest
  • RAM: modest
  • Storage: capacity matters, but SSD/NVMe helps with many concurrent users
  • Bandwidth: this is often the limiting factor—watch monthly transfer and port speed
---

Bottom line: what matters most for beginners

  • CPU: cores/threads determine concurrency; CPU model and dedicated access determine consistency.
  • RAM: running out causes severe slowdowns; capacity matters more than DDR4 vs DDR5 for most beginners.
  • Storage: NVMe usually improves real-world responsiveness; don’t ignore IOPS and latency.
  • Bandwidth: distinguish monthly transfer (TB/month) from port speed (Gbps); “unlimited” rarely means truly unlimited.
If you share what you’re hosting (site/app type, expected traffic, database size, and whether you use Docker), you can map these specs to a sensible starting configuration without overpaying.