CPU cores and threads: why more does not mean that much faster

Based on: official documentation from Intel, AMD, Apple, and Blender, checked in September 2026. Blender's requirements can change between versions.

A CPU spec sheet lists two numbers side by side, like "8 cores, 16 threads". Bigger numbers look better, but in practice the effect depends a lot on what kind of work you do.

• A core is a worker inside the CPU that actually does the calculating.
• Threads (logical processors) are the number of "slots the operating system can hand work to". When one worker alternates between two tasks, there are more threads than cores.
• Work that can be split into small pieces, like rendering or video encoding, gets faster with more cores. Work that runs in a single line, on the other hand, depends more on the speed of a single core than on the number of cores.

This guide explains what cores and threads mean, what Hyper-Threading (SMT) does, why speed does not rise as much as the numbers, CPUs that mix big and small cores, the numbers to look at for 3D work, and how to check your own CPU. If you first want to know how a CPU and a GPU differ, read CPU vs GPU first.

Cores and threads in one line each

The CPU vs GPU guide compared the CPU to "a small number of skilled workers". This guide steps one step further into that workshop. One skilled worker is one core.

The two terms divide like this.

The word "thread" is used in two ways here. A program splits its work into several software threads (streams of work that run in order), and several cores process these threads side by side. The thread count on a CPU spec sheet means "how many of these streams it can take at the same time". When reading a spec sheet, you only need to remember the second meaning.

Put a "4-core, 8-thread" CPU into this analogy. There are four skilled workers, but each moves between two workbenches, so eight tasks are laid out at once. That does not mean the workers have grown to eight. So the CPU does not get faster in proportion to the thread count. The next section explains why.

Hyper-Threading and SMT: how one core does the work of two

CPUs with more threads than cores usually have this technology. Only the name differs by company.

• Intel Hyper-Threading: a hardware technology that lets one core run two or more threads. When it is on, one physical core looks to the operating system like two logical cores.
• AMD SMT (Simultaneous Multi-Threading): in its most common form, one core runs two threads together.

The secret of one core doing two jobs is in its idle gaps. While calculating, a core often sits idle for a moment, waiting for the data it needs to arrive from memory. AMD's article on its server CPUs explains that such stalls happen often, and that SMT fills the empty time by letting another thread use the core's resources during those gaps. Intel likewise explains that Hyper-Threading raises throughput by using the time a core would otherwise spend waiting idle.

In worker terms, it is like doing a bit of the task on the next workbench while waiting for parts to arrive at the first one. The skilled worker did not become twice as fast. The waiting time got shorter.

So it helps to know these points.

1. The two threads share the resources of one core. It is not the same as getting two cores.
2. Not all work benefits. AMD explains that some high-performance computing tasks prefer to have a core's resources to themselves, while many tasks gain a little from SMT.
3. Not every CPU has it. Hyper-Threading is only on some Intel CPUs. That is why you will also see products whose spec sheets list the same number of cores and threads.

Three reasons speed does not rise as much as the numbers

It is common for more cores and threads to bring less speed than the numbers suggest, or for the speed you feel to stay the same. There are three main reasons.

1. Work that cannot be split

Even with eight skilled workers, if the work is one continuous chain, only one of them works. For example, work where "you must calculate A before B, and finish B before C" cannot skip steps, so it cannot be split.

3D rendering, on the other hand, splits well. The picture can be divided into many pieces, with each core calculating one piece. That is why Blender's Cycles render engine, when rendering on the CPU, automatically uses as many threads as there are logical processors.

2. Parts that cannot be split

Even work that splits well contains parts that must be done in a single line, such as the preparation stage, the stage that gathers results, and the stage that saves the file. No number of extra cores makes these parts faster.

Picture moving boxes in a warehouse. Moving boxes finishes faster with more workers. But if the warehouse has only one door and only one person can pass at a time, calling in more workers only makes the line at the door longer. The more cores you add, the larger the share of total time this "narrow door" part takes up, so doubling the cores does not cut the time exactly in half.

3. Single-core speed drops when many cores run together

A CPU runs within limits of power and heat. Take AMD's Precision Boost 2 as an example: boost (a feature that raises speed automatically) is decided by the number of cores working, temperature, the type of work, and more. Light work gets the highest boost, and heavy work that uses many cores for a long time gets less.

In the worker analogy, the workshop as a whole has a fixed amount of power. When only one person works, that person can run the machine at full force, but when everyone works together, each has to use a little less. Why the "maximum speed" on the spec sheet does not always appear, and how heat lowers speed, is covered in detail in Heat and thermal throttling.

CPUs that mix big and small cores

Some recent CPUs have cores of different sizes and characters. Intel CPUs, for example, have two kinds.

• P-cores (performance cores): large cores built so that a single core reaches the highest speed.
• E-cores (efficient cores): small enough that several fit in the space of one P-core, and built to do more work on the same power.

A hardware feature called Thread Director assigns "the right work to the right core". Heavy work that runs in a single line goes to the P-cores, and light background tasks like messengers or antivirus go to the E-cores. Note that Thread Director is only on 12th-generation and later Intel Core processors that combine P-cores and E-cores, and it needs operating system support.

E-cores do not only do background work. According to Intel, E-cores work together with the P-cores to speed up work that uses many cores, such as video rendering. In work that splits into small pieces, like CPU rendering, big and small workers all work together.

So with this kind of CPU, it is hard to compare it with other CPUs by core count alone. Even "one core" means different abilities for a big core and a small core. That is why makers list the count of each core type separately on their official spec pages. Apple also lists CPU cores in its Mac tech specs split into "X performance cores and Y efficiency cores".

The numbers to look at for 3D work

Separating the roles first avoids confusion. As shown in CPU vs GPU, the GPU draws the Blender work view (the viewport). Before that, the CPU calculates the scene (for example, calculations that move and deform objects, and running scripts) and runs the overall flow of the program. So whether the view turns smoothly depends heavily on the graphics card, and the cores and threads in this guide are about the calculations the CPU handles.

The number that matters depends on the type of work. The table below shows general tendencies based on the principles above, and real speed depends on the program and the scene.

Blender's official requirements are a CPU with at least 4 cores, 8 cores recommended. These numbers can change between versions, so check the official page again before installing. If you plan to render mainly on the graphics card, it is worth also reading Choosing a graphics card for Blender rendering.

When you want to do other things while rendering

In Cycles, the Performance ‣ Threads setting on the Render tab of the Properties editor uses all threads, matching the number of logical processors, when left at the default Auto-Detect. If you want to use the computer while rendering, change it to Fixed and enter a number smaller than the number of logical processors. This leaves a few workbenches empty so other work has room to come in. The setting applies only to CPU rendering.

Checking your CPU's cores and threads

Windows

1. Press Ctrl+Alt+Delete and choose Task Manager.
2. Click the Performance tab.
3. Click CPU on the left.
4. In the information at the bottom, the Cores number is the physical core count, and the Logical processors number is the thread count.

If the two numbers are the same, Hyper-Threading (SMT) is absent or off. If there are more logical processors, it is on. This number is exactly the thread count Blender's Auto-Detect uses.

Mac

Check in two steps.

1. Choose About This Mac from the Apple menu to see the Mac model name and the chip name. For more detail, hold the Option key and choose Apple menu ‣ System Information.
2. Open Apple's official tech specs page for that model. Under the chip item, the CPU core count is split like "X performance cores, Y efficiency cores".

Apple's tech specs do not list a separate thread count.

Correcting common misunderstandings

• "16 threads is as fast as 16 cores." Threads fill a core's idle gaps, which is not the same as doubling the cores.
• "More cores make every program faster." If a program does not split up its work, most cores sit idle. For work that runs in a single line, single-core speed matters more.
• "If the work view stutters, you need more CPU cores." The GPU draws the viewport. If the view is slow, look at the graphics card and VRAM first, and look at the CPU when the scene calculations are heavy.
• "E-cores are for background work, so they are useless for rendering." E-cores speed up work that uses many cores, like rendering, together with the P-cores.

Summary

The numbers on the spec sheet only tell you "how much work this CPU can take at once". How much faster your work gets depends on whether the work you do splits well. If you are planning a whole computer setup, also read Choosing RAM capacity and Laptop vs desktop for Blender and 3D work.

References

• Intel, What Is Hyper-Threading? (physical cores, logical cores, and software threads) — https://www.intel.com/content/www/us/en/gaming/resources/hyper-threading.html
• Intel, How to Check the Number of Cores and Threads in Your Intel Processor (checking in Task Manager) — https://www.intel.com/content/www/us/en/support/articles/000029254/processors.html
• AMD, Simultaneous Multithreading: Driving Performance and Efficiency on AMD EPYC CPUs — https://www.amd.com/en/blogs/2025/simultaneous-multithreading-driving-performance-a.html
• Intel, How Intel Core Processors Work (P-cores, E-cores, Thread Director) — https://www.intel.com/content/www/us/en/gaming/resources/how-hybrid-design-works.html
• AMD, Precision Boost 2 Performance Enhancement — https://www.amd.com/en/resources/support-articles/faqs/CPU-PB2.html
• Apple Support, View system information about your Mac — https://support.apple.com/en-us/guide/mac-help/syspr35536/mac
• Apple Support, Identify your Mac model name and serial number — https://support.apple.com/en-us/102767
• Apple Support, an example MacBook Pro tech specs page (performance and efficiency cores under the chip item) — https://support.apple.com/en-us/121553
• Blender, System Requirements — https://www.blender.org/download/requirements/
• Blender 5.2 LTS Manual, Cycles Performance — https://docs.blender.org/manual/en/5.2/render/cycles/render_settings/performance.html