If you look at AMD processors in the market, you’ll notice they’re cheaper than the competition. Despite the lower pricing, they offer incredible performance at high levels of efficiency. So, what’s the catch? How can they offer their central processing units (CPUs) at such a low price?

AMD processors are cheaper than their competition because they switched from a monolithic chip design to a chiplet architecture. This reduced the die size, making producing chips with fewer defects easier and reducing production costs. It also allowed them to reuse low-performing chips.

An AMD processor is placed with the pins facing upward.

1. AMD Switched to a Chiplet Design Architecture

If you’ve followed the tech news over the last few years, you would have heard everyone use the word ‘chiplets’ when referring to AMD.

This type of chip design played a significant role in propelling AMD back into the CPU industry.

AMD and Intel followed the conventional monolithic design when manufacturing desktop and mobile CPUs.

In other words, they produced CPUs from a single silicon wafer. Given the market’s competitive nature, both companies sought ways to outdo each other.

The Problem With Monolithic CPU Design Architectures

This topic is highly technical, and we’ll do our best to break it down. One way to increase performance and claim the top spot is to fit as many transistors as possible into the chip. Manufacturers could do this by decreasing the size of the transistor significantly.

The other manufacturing technique is to increase the number of cores in the CPU die for better performance.

However, this technique doesn’t make it any easier for manufacturers. For instance, they need to increase the die size to fit more cores into a monolithic chip design.

Larger die sizes are expensive to produce. Another problem with this solution is that it increases the number of defects in the manufacturing process.

Also, keeping the CPU cool under load becomes challenging, which affects performance and failure rates.

This followed the observations of Gordon Moore, who theorized that the number of transistors in the chip would double every two years.

It reached a point where the transistors were several thousand times smaller than a single strand of human hair. However, the CPU manufacturers still used the monolithic design, making a chip from a single silicon wafer.

This caused many problems, which is why giants like Intel faced a period of stagnation.

Why? For starters, the miniature size of the transistors increased the defect rate when manufacturing chips. The chip had to be discarded as it was done on a single silicon wafer.

One thing to remember regarding defective chips is that this cost has to be eaten by the companies who place the order, i.e., AMD and Intel, which increases production expenses.

AMD’s Answer to Monolithic CPU Designs

Instead of sticking with the monolithic design, AMD decided to try something different.

They chose to give the chiplet design a shot. This was a big gamble for the company as they were struggling financially.

Fortunately, the design change worked off, changing how the industry looked at AMD forever. It allowed them to break down a single wafer into multiple smaller chips or chiplets.

To understand this topic, we need to look at the internals of the CPU.

Every CPU has the following components:

CPU Cores.
Input/Output Control.
Graphics Processing Unit (GPU).
Display Control.
Memory Controller.

Rather than trying to produce everything on a single wafer, they separated each component into chiplets.

Any error in the manufacturing process won’t render the chip useless as the defective part can be swapped for a new one easily.

This helped AMD bring down costs as it reduced the number of silicon wafers they needed to discard due to manufacturing defects.

Also, the chiplet design lowered the die size, making it easier to manufacture.

Intel was the top dog in the CPU market for a long time. As AMD needed an edge over their competitor, the switch to chiplet design gave them everything they wanted.

This allowed them to offer excellent performance at a lower cost, irrespective of the product stack.

2. AMD Moved Away From Manufacturing

A lab technician is inspecting a silicone wafer for a semiconductor.

Another factor that helped AMD reduce the cost of producing processors was its decision to stop manufacturing CPUs.

This process is extremely expensive and can significantly impact a company’s resources if done incorrectly.

For example, the machines required to manufacture chips cost hundreds of millions of dollars.

AMD would not only have to make a substantial initial investment, but they’ll also have to spend billions of dollars in prototyping and testing for several years before they have a final product.

Another reason why CPU manufacturing is expensive is because of the precision required to produce processors.

Given that AMD was facing severe financial trouble, sticking to this practice would have been costly.

Instead, they partnered with Taiwan Semiconductor Manufacturing Company Limited (TSMC).

These semiconductor design and manufacturing giants had the technology, tools, and expertise to produce processors.

A Partnership That Changed the Tide

Combined with the chiplet design, it set the stage for AMD to take the CPU market by storm.

Generally, if you use the monolithic design strategy, you’ll need to use the latest manufacturing processes to get the best performance.

However, AMD didn’t have to stick with this for all the components in the chip.

Chiplets allowed them to use the latest manufacturing process for the CPU cores, which are crucial in determining its performance.

As the chiplet design separated each component, they could use the previous manufacturing process for other parts, like the input and output controller.

The design may be older, but it is reliable and cheaper than using the latest technology. This further helped the company bring down the cost of producing CPUs.

Another way this helped AMD was the ability to focus on improving their processors’ performance without a significant cost increase.

With this technique, they didn’t need to increase the density of transistors in the chip. This made it easier to manufacture the CPUs due to the simplicity of the design.

Also, fewer transistors mean the number of manufacturing defects reduces, further reducing costs.

3. AMD Reuses Older Technologies

There are a lot of benefits when it comes to using a modular design for manufacturing chips. It allows technology to trickle down, which benefits all their products.

For instance, if they’re using the latest technology for their server lineup, they have the luxury of using it in the next generation of consumer processors.

This isn’t possible with the monolithic chip design as several changes have to be made, which isn’t worthwhile.

With the chiplet design, they can make more use of the yields, which helps reduce costs in the long run. Also, it allows them to sell more products to meet consumers’ ever-growing demands.

The cost of using older technology goes down significantly as time goes on. This subsidizes using the latest fabrication processes, technology, and tools.

4. AMD Doesn’t Have To Discard Low-Performing Chips

Irrespective of how good the manufacturing process is, there’s always the chance of defects. Sure, certain errors can make chips unusable, but that’s not always the case.

Sometimes, manufacturing defects can result in chips that can’t utilize all their CPU cores.

For example, the chip only has 4 functioning cores, even though the manufacturing process produced one with 8 cores.

If AMD followed the monolithic design, they might have to eat the cost of this mistake if they can’t find a way to use this CPU.

With the modular design, all is not lost if this type of defect occurs in the manufacturing process.

AMD can harvest and use all the other components in other processors across the product stack.

Think of the chiplets as LEGO building blocks. You don’t have to throw away the entire set when one piece breaks. Instead, you can still reuse the remaining blocks in other sets.

This is a simplified way of how AMD can reuse low-performing chips.

5. AMD Allows Overclocking of All Processors

Overclocking is an easy technique to get more performance out of your CPU.

When choosing clock speeds and voltage for processors, manufacturers will stick with what they find stable across the whole lineup.

You should know every CPU will have slight differences, even if they’re the same product.

It means that you may be able to push your processor beyond the manufacturer’s specifications. How far you can boost the performance depends on your luck.

For example, you may be able to push it so far that it is close to the CPU in the next tier. This means you can get a lower-performance CPU and tweak it.

For Intel processors, you can only overclock those that are unlocked. If the processor falls into the locked category, there’s not a lot you can do.

However, AMD allows you to overclock every chip in its lineup. This means you get better frames per second (FPS) per dollar. Gamers usually use this metric to determine the value of processors.

If you want to overclock AMD CPUs, you need a motherboard to support it.

Generally, budget-oriented variants don’t support it as their voltage regulator module (VRM) won’t be able to handle the extra voltage for overclocking.

You’ll have to get the mid-range or high-end motherboards to use this feature.

You also don’t need to know the ins and outs of overclocking to make the most of it on AMD CPUs.

Thanks to their software for processors and GPUs, you can do this in one click. It also allows you to fine-tune the overclock if you’re an advanced user.

6. AMD Provides Long-Term Support for CPU Sockets

The CPU socket is where you install the processor on the motherboard.

How long manufacturers continue using the same socket determines whether you can swap out the old CPU for a new one from another generation of processors.

Intel usually supports two generations of CPUs before switching to a new socket design. This means you’ll have to get a new motherboard to upgrade to their latest offering.

On the other hand, AMD has been offering excellent support. Until now, they used the same socket design for five generations of CPUs.

They can focus on optimizing their technology rather than changing the processors’ die size and pin layout, which increases manufacturing costs.

Not only does this help them get more out of newer generations of CPUs, but it also reduces manufacturing overheads.

7. AMD Uses Multi-Die Design for Processors

As mentioned earlier, one way to increase the performance of CPUs is to fit more cores in the die. But, this method won’t work with monolithic processors due to design limitations.

AMD’s modularity approach allows them to work around this problem. This is why you see AMD offers CPUs with more than 16 cores and do not suffer from the drawbacks that come with monolithic design architecture.

With the chiplet design, AMD can produce more chips per silicon wafer. They can then stack multiple dies to increase the core count significantly and interconnect them. This minimizes latency, improving performance significantly.

Here’s an example of how this can work in the real world. Let’s say AMD is designing a CPU with 64 cores. With the chiplet architecture, they can break it down to 16 chiplets with four cores each.

This would be next to impossible if they stick with the conventional CPU manufacturing technique.

Not only will the design be complex, but it will also involve shrinking the transistors to an even smaller size.

As a result, there will be a considerable increase in the number of defects, making the whole manufacturing process expensive.

With the multi-die design, they can reduce their costs significantly.

One downside to this approach is that the various chiplets need to communicate with each other quickly.

Any delay can cause performance issues, which every manufacturer wants to avoid. Thankfully, AMD’s Infinity Architecture prevents this from being a problem.