According to AbbreviationFinder web dictionary, PCI Express is a short name for Peripheral Component Interconnect Express, basically a standard that allows various types of devices, such as video cards and SSDs, to be connected to a computer. This standard is also known as PCIe, or PCI-E.
But how exactly does PCI Express work? How does PCIe differ from USB ports, for example? What is PCI Express x1, x4, x8 and x16? I, Emerson Alecrim, explain everything you need to know about this technology from now on, follow me.
PCI Express: definition
In order to understand PCI Express, we must look to a relatively distant past, more precisely, to the 1990s. It was there that the PCI bus (Peripheral Component Interconnect ) emerged. For several years, this has been the most widely used standard for connecting video cards, network cards, modems and other devices to PCs and even servers.
By default, PCI connections worked with data transfer rates of up to 132 MB/s (megabytes per second). But technology is constantly evolving, you know. Over time, certain devices, especially video cards, started to demand higher speeds.
Variations of the PCI managed to offer more data transfer capacity, but it was still not enough. That’s when AGP (Accelerated Graphics Port ) came into being. This bus was introduced in the market with speeds that theoretically went from 266 MB/s (AGP x1) to 2,128 MB/s (AGP x8).
Much better, isn’t it? So much so that the AGP standard was on the market for a few years. But, guess what: over time, AGP specifications started to not be enough for the advances that were emerging in graphics applications. In addition, this standard was developed specifically for video cards. Other types of devices also evolve, right?
In the search for a solution to these limitations, the industry has developed – and continues to develop, hence the emergence of new versions – a technology called 3GIO that was later named PCI Express or, to facilitate, PCIe.
PCIe has a theoretical initial speed of 250 MB/s (see, initial) and can exceed 30 GB/s (gigabytes per second). You will already understand how.
How PCI Express came about
The Intel is a major precursor of technological innovations. In 2001, during a specific event, the company showed the need to create a technology capable of replacing the PCI and AGP standards : it was the aforementioned 3GIO (Third Generation I/O ).
Later that year, a group of companies formed the PCI-SIG consortium (including IBM, AMD Microsoft and Nvidia, in addition to Intel itself, of course) to address the issue. The first 3GIO specifications were approved there.
Among the requirements raised in these specifications are: support for the PCI bus, possibility of using more than one lane (you will already understand what that is), support for other types of connections, better power management and more protection against errors.
In April 2002, PCI-SIG approved a more complete set of specifications. It was at this time that 3GIO technology changed its name to PCI Express.
In July 2002, the consortium approved the final specifications of the standard and, as a result, the PCI Express 1.0 bus officially appeared, which was later revised and became PCI Express 1.1.
In November 2003, the first devices with PCI Express technology started to be developed and, in 2004, these products started to reach the market.
PCI Express x1, x4, x8 and x16
One of the most notable features of PCI Express is its ability to work with lanes, that is, lanes or data transmission paths. To make it easier to understand, think for a moment on a highway: the more lanes it has, the more cars it can drive at a time. The principle here looks like this.
If a device uses only one way, we say it is PCIe x1. If it uses four lanes for data transmissions, it is PCIe x4. You already understood the rest: eight lanes correspond to PCIe x8; 16 lanes, to PCI x16.
Why so many variations? For cost control, essentially. A device that does not have to deal with large volumes of simultaneous data – such as a simple network card – will work well with PCI Express x1, a format that is cheaper to implement.
Now a demanding GPU needs a lot of data. That is why the vast majority of video cards are based on PCI Express x16. As you already know, the more lanes or lanes, the more data is transferred at the same time. These are the theoretical maximum rates:
- PCIe x1: 250 MB/s
- PCIe x4: 1,000 MB/s (or 1 GB/s)
- PCIe x8: 2,000 MB/s (or 2 GB/s)
- PCIe x16: 4,000 MB/s (or 4 GB/s)
These numbers refer to PCI Express 1.0 and 1.1. The technology achieves higher data transfer rates. You will understand more in the topics about the standard versions. But first, let’s take a look at the PCIe connectors.
PCI Express connectors (slots)
Unlike USB or HDMI, PCI Express is an internal connection standard, which means that you will not find a specific port for this technology on the back of your desktop or on the side of your notebook. It is necessary to open the computer to find the slots (slots) for PCIe devices on the motherboard.
These slots have different sizes, which vary depending on the types of PCI Express that the motherboard supports. The more lanes or lanes there are, the bigger the connector. Because of this, a PCIe x1 is short; a PCIe x16 is already elongated – and extremely common, after all, this is the standard most used in video cards:
- PCI Express x1: 25 mm in length; 18 pins
- PCI Express x4: 39 mm in length; 32 pins
- PCI Express x8: 56 mm in length; 32 pins
- PCI Express x16: 89 mm in length; 82 pins
Two important observations: the first is that these slots are not always mandatory. Many M.2 SSD models, for example, have a completely different connector, but can communicate with the motherboard via the PCI Express interface.
The second is that these slot sizes are standardized, that is, later versions of PCIe also use them. Speaking of which, let’s go to them.
PCI Express 2.0
So far we have only talked about PCI Express 1.0 and 1.1 (we will call them 1.x). But in 2007, the PCI-SIG group introduced PCI Express 2.0. The main feature of this new version is the greater bandwidth of each lane: up to 500 MB/s. This means that a PCIe 2.0 connection with x16 can reach up to 8 GB/s in data transfer (16 x 500 MB/s).
Like versions 1.x, PCI Express 2.0 is based on 8b/10b encoding, which indicates that each group of 8 bits is encoded in a 10-bit signal on each transmission.
In addition, PCIe 2.0 can work with 5 GT/s (gigatransfers per second) – a gigatransfer corresponds to basically a billion data transfers per second. For comparison, PCIe 1.x works with up to 2.5 GT/s.
PCI Express 3.0
The PCI Express 3.0 appeared in 2010. By 2019, when this text was written, this pattern was still widely used by the industry, after all, it provides up to double the performance of PCIe 2.0.
This version works with 8 GT/s. The bandwidth for each lane is 1 GB/s. This means that a PCIe 3.0 device with x16 can handle up to 16 GB/s.
In part, the improved performance of PCI Express is due to the use of 128b/130b encoding instead of 8b/10b. This more generous coding, so to speak, is one of the factors that allow PCI Express 3.0 to be faster.
PCI Express 4.0
The specifications for PCI Express 4.0 were finalized in 2017. Yes, the gap over the previous generation was seven years. It’s a long time, but PCI-SIG struggled to double the performance compared to PCIe 3.0 and maintain backward compatibility.
But the efforts paid off: PCIe 4.0 is capable of working with 16 GT/s and 2 GB/s per track. As a result, a PCI Express 4.0 x16 connection has a theoretical speed of up to 32 GB/s.
Here, the encoding is also 128b/130b.
PCI Express 5.0
The PCI Express 5.0 was officially announced in June 2019, a period of about a year and a half compared to PCI Express 4.0. One reason is that the two versions are technically very similar, so much so that PCIe 5.0 is seen as an extension of PCIe 4.0.
It may even be an extension, but the logic of doubling the speed has been maintained: PCIe 5.0 connections can work with 32 GT/s and up to 4 GB/s per lane, that is, 64 GB/s at x16. The 128b/130b encoding was maintained.
It is worth noting that PCI Express 5.0 is compatible with devices based on previous versions of the technology, as well as these are compatible with predecessors – one of the requirements of each new version is to be backwards compatible.
Find out more in the text that explains the differences between PCI Express 4.0 and PCI Express 5.0.
PCI Express 6.0
It may sound like a joke, but it’s not: PCI-SIG confirmed the development of PCI Express 6.0 in June 2019, just a few days after the official announcement of PCIe 5.0.
PCI Express 6.0 is expected to be officially announced in 2021, however. The new version should also have more bandwidth: up to 128 GB/s (twice the width of PCI Express 5.0). If this feature is confirmed, the technology will have the following speeds:
- PCIe 6.0 x1: 8 GB/s
- PCIe 6.0 x4: 32 GB/s
- PCIe 6.0 x8: 64 GB/s
- PCIe 6.0 x16: 128 GB/s
What is most surprising is that plans for PCIe 6.0 were revealed when the first devices based on PCI Express 4.0 were just starting to hit the market.
If there are no changes or setbacks along the way, PCI Express 6.0 will be launched sometime in 2021, but the first devices based on the technology will not hit the market until the following year, probably.
Conclusion: PCIe speeds
Finally, I must make two considerations. The first concerns the maximum speeds of PCI Express. This is a standard capable of working in full duplex mode : transmission is done in a bidirectional way, that is, data is sent and received at the same time.
Well, the speeds covered in this text are related to the maximum speeds in only one direction (only sending or receiving data only), but you can find manufacturers and specialized vehicles relatively easily using measurements that consider simultaneous data sending and receiving (bidirectional).
Whenever the approach is bidirectional, PCI Express speeds will appear doubled, note:
- PCIe 1.0: 8 GB/s
- PCIe 2.0: 16 GB/s
- PCIe 3.0: 32 GB/s
- PCIe 4.0: 64 GB/s
- PCie 5.0: 128 GB/s
- PCie 6.0: 256 GB/s (expected)
Another important consideration is that, in this text, bandwidth and speed appear as synonyms, but for the sake of convenience: commercially, this approach is very frequent.
However, conceptually, these two metrics are not the same. Bandwidth actually indicates the amount of data that can be transmitted over a given period of time over a connection, but not necessarily the speed at which it occurs.
But since bandwidth and speed are directly related metrics, it is common for both to be treated as the same thing.