Read any server specification sheet and you meet the same three words, often used as though they were interchangeable: cores, threads, and vCPUs. They are not interchangeable, and the difference between them decides how much work a machine can genuinely do. A server advertised with 40 threads does not have 40 cores, and a virtual machine sold with 8 vCPUs does not hold 8 physical cores. Understanding CPU cores, and the terms that surround them, is what lets you read a spec sheet and see the real capacity rather than the marketing figure.
This guide explains each term in plain language, shows how they relate, and clears up the confusion that leads people to overestimate what a processor can handle. It stays on the concepts, not the buying decision. By the end, a line like “20 cores, 40 threads” will tell you exactly what it means.
📖 New to server hardware?
The processor is one part of a larger picture. Read What Is a Dedicated Server?, a plain introduction to how the hardware fits together and why exclusive access to it matters.
What a CPU Core Actually Is
A core is a complete, independent processing unit inside a dedicated server‘s processor. It reads instructions and executes them, and it does this on its own, in parallel with the other cores on the same chip. A processor with 10 cores has 10 of these units, and it can genuinely do 10 things at the same time.
This is the most honest number on a spec sheet, because a physical core is real hardware. When a processor is described as having 10 cores, there are 10 physical execution engines etched into the silicon. Nothing is being counted twice, and nothing is shared. If a workload can be split across all of them, each core carries its full share.
For decades, processors had a single core, and the way to make them faster was to raise the clock speed, the number of cycles per second. That approach hit physical limits, chiefly heat, and the industry turned instead to adding more cores. A modern server processor like the Intel Xeon Gold 6138 carries 20 physical cores on one chip, where a processor from twenty years earlier had one.
Threads: Where the Numbers Start to Mislead
Here is where the confusion begins, and where a spec sheet can quietly overstate what you are getting.
A thread, in processor terms, is a single stream of instructions that a core works through. Traditionally, one core handled one thread at a time. Intel introduced a technology, called hyper-threading, that lets a single physical core present itself to the operating system as two logical processors, each able to accept its own thread. This is why a 20-core Xeon reports 40 threads: each physical core is showing up as two.
The critical point is what those two threads share. As Intel’s own description of the technology makes clear, they do not each get a separate execution engine. They share the single core’s execution units, taking turns to use the parts that would otherwise sit idle. When one thread is waiting, for data from memory, say, the other can slip in and use the resources that would have gone unused. That is the entire mechanism, and it explains both the benefit and its ceiling.
The benefit is real but modest. According to figures Intel has published, hyper-threading delivers up to around 30 percent more performance on a single-socket system, and up to about 15 percent on a dual-socket one. That is a useful gain. It is not the 100 percent that doubling the thread count might suggest, and treating a 40-thread processor as though it were a 40-core one overstates its real capacity by a wide margin.
Physical cores vs the performance you actually gain
Two threads on one core add roughly 30 percent, not 100 percent. Two physical cores double the work. Figures are indicative and vary by workload.
📖 How this shows up under load
Thread counts change how you read server load figures. Read What Is Server Load and Why Websites Slow Down, on how load average relates to the number of cores actually available.
vCPUs: The Virtual Layer on Top
The third term, vCPU, adds one more layer of abstraction, and it is the one most likely to mislead.
A vCPU, or virtual CPU, is the unit of processing a hypervisor hands to a virtual machine. When a provider sells a virtual server with 4 vCPUs, those vCPUs are not physical cores set aside for you. In the most common arrangement, each vCPU maps to a single thread of a physical core, not to a whole core. So 4 vCPUs typically correspond to the compute of roughly two physical cores, not four.
There is a further catch, and it matters more than the mapping. On shared virtual infrastructure, the physical cores beneath the vCPUs are frequently oversubscribed, meaning the provider sells more vCPUs than there are threads to back them at full load. When several tenants are busy at once, their vCPUs contend for the same physical hardware. A vCPU, in other words, is a claim on processing time, not a guarantee of it. This is the root of the <a href=”https://swify.io/blog/glossary/#noisy-neighbour”>noisy neighbour</a> problem, where another tenant’s workload degrades yours.
This is precisely where a dedicated server differs. On bare metal, there is no hypervisor and no vCPU abstraction. You hold the physical cores directly, and their full capacity is yours at every moment, because no one else is sharing them.
Putting the Three Together
The clearest way to hold these terms in mind is to see them as three layers, each one further from the physical hardware than the last.
A core is real, physical, and honest. It is a complete execution engine, and its count is the truest measure of parallel capacity. A thread is a logical stream; with hyper-threading, one core presents two, but the two share the core’s real resources and add perhaps 30 percent, not 100. A vCPU is a virtual allocation on top of a thread, one further step removed, and on shared infrastructure it may not even be guaranteed.
The practical reading of a spec sheet follows directly. Physical cores are the number to trust. Threads are useful but should never be read as cores. And vCPUs are the figure to treat with the most caution, because their real value depends entirely on whether the underlying hardware is dedicated or shared.
The three layers, from hardware to abstraction
| Term | What it is | How much to trust it |
|---|---|---|
| Core | A physical execution engine | Fully. The honest number. |
| Thread | A logical stream on a core | Partly. Adds ~30%, not 100%. |
| vCPU | A virtual slice of a thread | Cautiously. May be oversubscribed. |
Each layer sits one step further from the physical silicon than the one above it.
Physical cores, held directly by you
Swify dedicated servers run on Intel Xeon Gold processors with no virtualisation layer and no shared vCPUs. Full root access, enterprise SSD and NVMe storage, and 1Gbps unmetered bandwidth from a Netherlands data centre. From €120/month.
→ Explore Swify Dedicated ServersFrequently Asked Questions
What is the difference between a CPU core and a thread?
A CPU core is a physical, independent processing unit that executes instructions on its own. A thread is a single stream of instructions a core works through. With Intel’s hyper-threading, one physical core can present two threads to the operating system, which is why a 20-core processor often reports 40 threads. The two threads share the core’s execution resources rather than each having their own.
The result is that two threads on one core add roughly 30 percent more performance, not double, because they share the same physical hardware. Read How to Choose the Right CPU for Your Dedicated Server for how core and thread counts factor into a real workload.
Is a vCPU the same as a physical core?
No. A vCPU, or virtual CPU, is a unit of processing a hypervisor assigns to a virtual machine, and it usually maps to a single thread of a physical core rather than to a whole core. So a virtual server with 4 vCPUs typically holds the compute of around two physical cores, not four. On shared infrastructure, vCPUs can also be oversubscribed, meaning more are sold than the hardware can back at full load.
A vCPU is therefore a claim on processing time, not a guaranteed slice of dedicated hardware. On a bare-metal dedicated server, there is no vCPU layer at all. Read Dedicated Server vs VPS for how dedicated and virtual resources compare.
How many cores do I need for a server?
It depends entirely on the workload. Cores handle parallel work, so applications that serve many simultaneous requests, such as busy web servers or databases, benefit from more of them. Work that runs as a single stream gains little from extra cores and more from higher clock speed per core. The honest measure to plan around is physical cores, not thread or vCPU counts, which can overstate real capacity.
Matching core count to the actual work is the key, rather than simply buying the largest number. Read How to Choose the Right CPU for Your Dedicated Server for a practical framework.
Why does my server show more processors than it has cores?
This is hyper-threading at work. When it is enabled, each physical core presents itself to the operating system as two logical processors, so a system monitoring tool on an 8-core machine will display 16 logical processors. The operating system schedules work across all 16, but they are backed by only 8 physical cores sharing their execution resources in pairs.
This is why load and usage figures need care to interpret: the logical processor count is not the physical core count. Read What Causes High CPU Usage on a Server? for how to read these figures correctly.
Does more cores always mean better performance?
No. More cores help only when the workload can be split across them. An application built to run as a single thread will not run faster on a 40-core processor than on a 10-core one with the same per-core speed, because it can only ever use one core at a time. For that kind of work, clock speed and per-core performance matter more than the total count.
The right processor matches the shape of the workload, whether that favours many cores or fast individual ones. Read Why Dedicated Servers Deliver Superior Performance for how hardware choices translate into real-world speed.
What are cores and threads on a dedicated server versus a VPS?
On a dedicated server, you hold physical cores and their threads directly, with no virtualisation layer between you and the hardware, so their full capacity is always yours. On a VPS, you receive vCPUs, which are virtual allocations mapped onto the threads of physical cores shared with other tenants. The same underlying hardware is divided among several customers.
This is why a dedicated server delivers consistent, predictable processing, while a VPS can vary with the activity of its neighbours. Read Bare Metal vs Virtualized Servers for the full comparison.