The expansion slots on a motherboard provide us with almost unlimited options for customizing our computers. In this video, you’ll learn about the PCI bus, the PCI Express bus, and riser cards.
<< Previous Video: Motherboard Form Factors Next: Motherboard Connectors >>
You can extend the functionality of a motherboard by adding adapter cards. In this video, we’ll show you the addition of these adapter cards into the motherboard expansion slots. There are many different components on a motherboard.
This motherboard has some CPU slots. There are slots for memory. There’s options for expansion on the motherboard expansion slots. And all of these devices need to communicate between each other.
We do that through a series of buses on the motherboard. There might be a data bus that sends information from your CPU to your memory. There may be a different computer bus that is connecting your memory to the expansion slots on your motherboard.
This allows you to extend the functionality of your motherboard. You could add new memory to memory slots. And those new memory modules will now communicate across the memory bus. You could also add other options, such as an additional video card or other interfaces, using the expansion slots on the motherboard.
Those expansion slots use an expansion bus to connect those components to the rest of the motherboard. Older expansion slots that supported parallel communication allowed us to send information across the bus, and the wider the bus allowed us to send more information. As we’ve transitioned from a parallel bus to a serial bus, we no longer describe the size of the bus as the width or the number of bits we can communicate. Instead, we describe it in the total amount of bandwidth we’re able to communicate across that bus.
You often see specifications of a motherboard that describe a particular bus running at a particular clock rate, or particular speed. For example, a 1 megahertz clock rate– or 1 MHz– means that you can send 1 million cycles, or 1 million pieces of information, in 1 second. A clock rate that’s running at 1 gigahertz is 1,000 megahertz, or 1 billion cycles per second.
To confuse things further, the clock rate that you’ll see in these specifications don’t necessarily equal the transfer rate of the data across that bus. For example, a DDR3 memory module can transfer 64 times the amount of data than its stated memory clock speed. Fortunately, expansion buses are very standardized. You won’t have to worry about the clock speeds of these expansion buses, and instead, you can look at the size and type of the interface to know exactly what bandwidths will be available.
The two most common interfaces that you’ll see on an expansion bus are the conventional PCI and the PCI Express. Let’s first look at the conventional PCI expansion bus. This PCI stands for Peripheral Component Interconnect.
We don’t commonly refer to it with that name. Instead, we just refer to it as the PCI bus. PCI interface cards communicate over a parallel connection, and they will support a 32-bit parallel communication or a 64-bit parallel communication. There have been updates and modifications to the PCI bus interface over the years, and the total amount of speeds that you’ll find on these buses will depend on what version of PCI happens to be in use on your motherboard.
For example, there are standards of the PCI interface that support 133 megabytes per second. That’s using a 32-bit card running at a 33 megahertz clock rate. The higher end of PCI would be a 533 megabyte per second interface, and that’s a 64-bit interface that’s running at 66 megahertz.
Here’s a block diagram of the 32-bit PCI parallel bus interface. This parallel bus means that we’re sending all 32 bits at the same time across this bus from one side to the other. You can see the expansion slots are connected to the bus on one end, and usually, on the other end, is what we call the south bridge, or the input/output controller hub. A 64-bit PCI bus looks very similar, but you can see we have a much wider bus that we’re sending those 64 bits across simultaneously.
Here’s a close-up of a motherboard that has PCI slots on it. The shorter slots are the 32-bit slots, and the larger slots are the 64-bit slots. One thing you’ll also notice is that these slots have particular notches, or keys, that will determine the type of card that will connect to these different interfaces.
The devices that we’re plugging into these expansion slots need to be powered. And in most cases, these devices are getting that power directly from the PCI slot itself. The PCI slot will provide power at 3.3 volts, or it will provide power at 5 volts. The device that you’re plugging in may use one or either one of those power sources, and the type of power it’s going to use will depend on the notches that have been made available at the bottom of the interface itself.
If the notch is closest to the back of the computer, that specifies that this device can support 3.3 volts of power. This particular card also has a notch farther to the back, which means that it also can support 5 volts of power from that PCI interface. A 64-bit card is going to be a bit longer, as we saw with the motherboard. It still has notches for 3.3 volts and 5 volts– if that card supports it– and you’ll notice there is a third notch available closer to the back that signifies that this is a 64-bit interface card.
If we look at our motherboard again, we see that we do have a slot available for a 64-bit card. In fact, the motherboard itself has written, on the motherboard, 64 PCI number 3 and number 4. We can use either one of those 64-bit slots to install our interface card.
An expansion card should slide into a slot on a motherboard without too much force. You don’t want to push down on the motherboard and bend any part of those motherboard components. I would normally put the card onto the slot and make sure that all of the notches on the slot are keyed appropriate for the interface that I’m plugging into.
If I’m comfortable that this particular card matches the slot that I’m pushing it into, then I can begin pushing down on the card itself. You want to push down on the edges of the card and be careful not to touch any of the components on the card itself. Here’s a PCI card that’s now installed into the 64-bit slot. Looks like the right side is actually up a little bit higher than I’d like, but because this is not in an actual case– we’re not screwing down the top of this card to the case itself. Once you do that, you’ll find that the card sits very flat and completely inside of that PCI slot.
On most modern motherboards, you’re not going to see the older PCI interface. Instead, you’ll see the newer PCI Express. You may see this abbreviated as PCIe. PCI Express has effectively replaced all of the older expansion slot types– not just PCI, but even PCI-X and AGP, which is the older, accelerated graphics port.
One big change to PCI Express over PCI is that PCI Express communicates over a serial connection. You’ll sometimes see these referred to as serial lanes of communication. And another advantage to PCI Express is that not all of the devices share these lanes, which means we have a much more efficient communication if we have more than one interface card in our computer.
We also have options on the size of the different lanes that are available for these expansion slots on our motherboard, and you’ll see them described as 1, 2, 4, 8, 16, or 32 full-duplex serial lanes. Written out, you’ll see these referenced as x1, x2, x4, x8, x16, and x32. The x is actually pronounced “by,” so it’s also common to hear this referred to as “by 1,” “by 2,” “by 4,” et cetera.
Because this is a serial connection, the block diagram for PCI Express looks very different than the one we saw for PCI. We don’t have the large, wide parallel bus any longer. Instead, there’s a serial connection. So if we have a by one lane, that means we have one lane in one direction and one lane in the other direction. If we had a larger expansion bus, like a four-lane expansion, then we would have four separate serial connections between the different PCI Express components.
PCI Express has also had a number of different versions through the years, and those different versions have brought different speeds. You’ll need to refer to your motherboard to see what version of PCI Express it supports. For example, PCI Express version 1 can support a per lane throughput in each direction of 250 megabytes per second, and the most common version of PCI Express today is version 4.0, which supports 2 gigabytes per second of throughput per lane in each direction.
There will, of course, be newer versions of PCI Express released. For example, in 2019, we expect to see version 5.0, which should take the total speeds up to 4 gigabytes per second in a per lane throughput in each direction for PCI Express. Here’s an ATX motherboard that supports both PCI interfaces and PCI Express interfaces on the same device. You’ll commonly see this on larger motherboards that want to provide a migration path between the older PCI slots and the PCI Express slots.
Here’s a closer look, and you can see the older PCI slots on the left, and the faster PCI Express slots on the right. You can also see that the PCI Express slots may be different sizes, depending on the total lanes that are supported for each interface slot. It’s easy to pick out the one-lane PCI slots, and then you have these other two slots that appear to be the same physical size, but if you look closely, there are different number of copper contacts inside of this interface.
This is a four-lane slot on the left. You can almost see the writing on the motherboard that says PCIe x4, and then you have the PCIe x16 for the 16-lane slot on the right. And selling a PCI Express card is almost identical to installing one in a PCI interface.
One characteristic to these larger PCI Express cards is that there’s often a hook near the back that allows you to anchor the card on the far side and anchor the card with a screw on the case itself. This becomes important. If you need to remove this card, you’ll need to remove the screw, and you’ll also need to remove the latch that is connecting this card to the motherboard on the other side.
Once the card’s installed, you usually have this hook in the back that is fastening it on one side, and the screw that’s fastening it to the case on the other. One challenge you have with interface cards in a data center is that a number of these data center servers are these thin, small, rack-mounted devices. And you can see the size of the adapter card is much taller than the size of the device itself.
The challenge, of course, is that our motherboard is sitting horizontally inside of that server, and we would normally plug our interface card in vertically. Instead, we want to find some way to turn the interface card so that it can fit inside of this same form factor. And to do that, we need to use a riser card.
A riser card plugs in vertically to that motherboard, but notice that it changes the direction of the interfaces. This means that we can plug in those larger interface cards into a server even when the server is that narrow size that you might see in a data center.