The processor, or central processing unit (CPU), is the brain of the computer. It does the majority of the processing work and is a key factor in the overall performance of a system. Over the years, CPUs have steadily improved, and as a computer technician, you’re expected to know some basics about them.
There are two primary manufacturers of computers used in computers: Intel and Advanced Micro Devices (AMD).
Intel. Intel is the largest seller of CPUs, selling about 80 percent to 85 percent of all CPUs. It manufactures other products as well, including chipsets, motherboards, memory, and SSDs.
AMD. AMD is the only significant competition to Intel for CPUs, and it sells about 10 percent to 15 percent of all CPUs. It also manufactures other products, including graphics processors, chipsets, and motherboards.
It’s possible to purchase a new CPU and install it in a motherboard as part of an upgrade. An important question to ask is, “What should I buy?” When shopping, you’ll see names like the following:
Intel Core i7-960 Processor 3.2 GHz 8 MB Cache Socket LGA 1366
Phenom II X4 965 AM3 3.4 GHz 512KB 45 NM
Will either of these fit in your motherboard? You might not know right now, but by the end of this chapter, you’ll have the information to answer that question.
32-bit vs. 64-bit
CPUs are identified as either 32-bit or 64-bit. Similarly, operating systems and many applications are referred to as either 32-bit or 64-bit. Key points to remember include the following:
Windows operating systems come in both 32-bit and 64-bit versions.
A 64-bit CPU is required to run a 64-bit operating system.
A 64-bit operating system is required for 64-bit applications.
A 64-bit CPU will also run 32-bit software.
The numbers 32 and 64 refer to the address bus discussed in Chapter 2. As a reminder, the address bus is used to address memory locations. A 32-bit CPU supports a 32-bit address bus and can address 232 memory locations, or 4 GB of RAM. A 64-bit CPU supports a 64-bit address bus and can address 264 memory locations, or about 17 EB.
Operating systems and applications have gotten more sophisticated over the years. Developers have programmed extra features and capabilities, but all of these extras consume additional RAM. For many users, 4 GB of RAM simply isn’t enough.
Due to the demand, developers such as Microsoft have created 64-bit versions of their operating systems. However, these 64-bit operating systems can run only on 64-bit CPUs. If you want to directly address more than 4 GB of RAM, you need both a 64-bit CPU and a 64-bit operating system.
32-bit and x86. You often see 32-bit operating systems and software referred to as x86. This is a reference to the long line of Intel CPUs that ended in 86 and can run 32-bit software. AMD processors have different names but are also known to be x86-compatible.
64-bit. Intel refers to its 64-bit processors as Intel 64, and AMD calls its 64-bit processors AMD64. Software makers often refer to 64-bit compatible software as x64.
Most CPUs today have multiple cores within them. Each core is a fully functioning processor. With multiple cores, the CPU can divide tasks among each core. The result is a faster system.
Operating systems view the multiple cores as individual CPUs. For example, a single eight-core processor will appear in Task Manager as though it is eight separate processors, as shown in Figure 3-6.
Figure 3-6. Task Manager showing eight cores of a single CPU.
It’s worth noting that Figure 3-6 is the same view you’d see if you had an Intel four-core processor with hyper-threading enabled. Hyper-threading is described later in this chapter.
A key point to remember is that even when a CPU has multiple cores, it is still a single chip that plugs into the motherboard. Motherboards are available that accept multiple CPUs, but they are more common on servers than on desktop systems. Most desktop systems have a single CPU, and it’s common to see CPUs with multiple cores.
Hyper-Threading Technology (HT) is used on some Intel CPUs to double the number of instruction sets the CPU can process at a time. Within a CPU, a thread is an ordered group of instructions that produce a result. When hyper-threading is used, a single CPU can process two threads at a time.
This is not physically the same as a multiple-core CPU. However, just as a dual-core CPU simulates two physical CPUs, a single-core CPU with hyper-threading simulates two physical CPUs. Operating systems can’t tell the difference.
Intel makes use of both hyper-threading and multiple cores on some of its CPUs. For example, Figure 3-7 shows a screen shot of the System Information tool in Windows 7. It identifies the processor as an Intel Core i7 CPU with four cores and eight logical processors. Each core is using hyper-threading, and the operating system interprets it as eight CPUs.
Figure 3-7. Msinfo32 showing that hyper-threading is enabled.
Many computer components and software applications use some type of cache. As a simple example, web browsers use a browser cache. When you go to a website, information is transmitted over the Internet and displayed in your web browser, and it is also stored in the browser cache. If you go to the website again, data can be retrieved from the browser cache rather than downloaded from the Internet again. The browser uses different techniques to ensure that it displays current data, but if that data is on your drive, it is displayed much more quickly than it would be if it had to be downloaded again.
The CPU has cache that it uses for fast access to data. If the CPU expects to use some type of information again, it keeps that information in cache. A significant difference between the web browser cache and the CPU cache is that the CPU cache is RAM and the web browser cache is stored as a file on a hard drive.
CPU Cache Types
The two primary types of cache used by CPUs are:
L1 cache. This is the fastest, and it’s located closest to the CPU. A multiple-core CPU has a separate L1 cache located on each CPU core.
L2 cache. L2 cache is a little slower than L1 cache, and it is shared by all cores of the CPU. In older systems, L2 cache was stored on the motherboard, but today it is much more common for L2 cache to be part of the CPU.
Figure 3-8 shows the relationship of the CPUs to cache and RAM installed on the motherboard. In the diagram, the CPU is a two-core CPU, and you can see that the L1 cache is included on each core and that L2 cache is shared by each of the cores. When the CPU needs data, it will check the L1 cache first, the L2 cache next, and then the L3 cache if it exists. If the data isn’t in cache, the CPU retrieves it from RAM.
Figure 3-8. CPU and cache.
Many newer CPUs include L1 cache for each core, L2 cache for each core, and a single shared L3 cache—all on the same CPU chip.
Without cache, the CPU would have to store data in the motherboard RAM. The CPU cache is SRAM, which is much faster than the dynamic RAM used on the motherboard. Also, the motherboard RAM is physically farther away, adding more delays.
CPU Cache Size
The size of the CPU cache is small compared to the overall amount of memory in a system. For example, you might see cache sizes as low as 8 KB or as large as 20 MB. In contrast, most personal computers have 1 GB of RAM or more. The cache can be listed as just a total of all L1, L2, or L3 cache, or you might see it listed individually.
L1 is smallest. L1 is sometimes stated as two numbers, such as 32 KB + 32 KB, to indicate it is using one cache for frequently used instructions and another cache for data. Sizes of 32 KB or 64 KB are common.
L2 is larger than L1. When a CPU has separate L2 cache for each core, it is often identified as the amount per core. For example, a two-core CPU with 4 MB total L2 cache can be expressed as 2 x 2 MB, or just 2 MB per core. Sizes of 256 KB, 512 KB, and 1,024 KB are common.
L3 is larger than L2. Sizes between 2 MB and 8 MB are common.
The speed of a CPU is based on the speed of the crystal and the multiplier. For example, if the crystal speed is 100 MHz and the multiplier is 20, the CPU has a speed of 2 GHz (20 x 100). The faster the speed, the faster the CPU.
You commonly see the speed of the processor listed as only the multiplied speed. For example, in Figure 3-7 you can see that the processor is an Intel Core 7 CPU 870 and the clock is listed as 2.93 GHz. The system is using a 133.333-MHz clock (commonly listed as 133 MHz) and a 22-times multiplier.
Processors are rated based on the maximum speed they can handle, and more expensive processors can handle faster speeds. You can increase the speed by increasing the clock frequency, increasing the multiplier, or both. Most motherboards have this preselected, but it is sometimes possible to manipulate the clock or the multiplier to overclock the system. In some systems, the BIOS includes a Cell menu that enables you to increase the base frequency and increase the CPU Ratio (multiplier).
Chapter 2 mentions the front side bus (FSB) and how it provides a direct connection between the CPU and the north bridge portion of the chipset. In the past, CPU speeds were stated as the FSB speed. Today, many CPUs have taken over the functionality of the north bridge. The CPU still needs to communicate with the chipset, and there are a few different ways this is done, including the following:
Intel Direct Media Interface (DMI). The DMI can use multiple lanes, similar to Peripheral Component Interconnect Express (PCIe).
Intel’s QuickPath Interconnect (QPI). Each core in a processor has a separate two-way 20-lane QPI link to the chipset.
HyperTransport. AMD uses HyperTransport with the FSB to increase the speed.
You still see CPUs advertised with a speed that you can use for comparisons. For example, one CPU might have a speed of 2.8 GHz and another might have a speed of 3.4 GHz. It’s safe to assume that the 3.4-GHz CPU is faster, but the speed isn’t always tied to the FSB.
Chapter 2 introduced virtualization concepts and instructions on how to enable virtualization in BIOS. As a reminder, virtualization software allows you to run multiple virtual machines (VMs) as guests within a single physical host computer. The CPU needs to support virtualization, and it usually needs to be enabled in BIOS. On many AMD-based systems, virtualization is enabled by default and cannot be disabled.
Most Intel and AMD CPUs include native support for virtualization. The exception is laptop computers, which sometimes include CPUs that do not support it. Intel refers to its virtualization support as VT-x, and AMD calls its support AMD-V. If you want to verify that a CPU or motherboard supports virtualization, look for those terms.
Graphics is one of the areas of a computer that has been increasing as quickly as the CPU area, and the two are starting to merge. Early computers could display only letters on a screen 80 characters wide. Today, it’s common to watch high-quality video streaming from a website or to play games with computer-generated graphics and amazingly realistic scenery.
The following list describes the progression of graphics capabilities on computers:
Onboard graphics. Graphics capability was built into the chipset. This was often very basic but met most needs.
Expansion cards. You could install a graphics card with a dedicated graphics processing unit (GPU) and plug it into an available expansion slot. Instead of the CPU doing the graphics calculations, the GPU would do them. Peripheral Component Interconnect (PCI) cards were an early version.
Dedicated graphics slots. Accelerated Graphics Port (AGP) provided a single dedicated graphics slot that worked separately from PCI. AGP did not compete with PCI, so it provided better performance. Later, PCIe allowed graphics cards to use their own dedicated lanes, and it replaced AGP.
Direct access graphics. The CPU interacted with the AGP slot via the chipset. Newer CPUs bypass the chipset and interact directly with a dedicated PCIe slot used for graphics. This is common in many systems today.
Integrated graphics processing unit (GPU). A recent trend in newer CPUs is to include an integrated GPU on the CPU. GPUs can provide high-quality graphics without the additional cost of a graphics card. However, these are not as powerful as a dedicated card.
AMD refers to some chips with a GPU as an accelerated processing unit (APU) instead of a CPU. APUs can include a GPU or other specialized capability, and the AMD Fusion is an example.
There is a dizzying number of different processors. You’re not expected to know the characteristics of each individual CPU, but you should be able to recognize the names and know the manufacturers. The objectives specifically list the CPU socket types you should know, but for the sockets to make sense, you need to have a little bit of knowledge about the CPU versions.
Intel and AMD use code names related to the manufacturing process and then create different processor families with the process. The manufacturing process is stated as a measurement and refers to the distance between certain components within the chip. Many current CPUs have processes of 65 nanometers (nm), 45 nm, 32 nm, and 22 nm. A nanometer is one billionth of a meter and is often used to express atomic scale dimensions, such as the width of an atom or the width of a group of molecules. In this case, smaller is better.
The following are recent Intel and AMD code names:
Core—65-nm and 45-nm process
Sandy Bridge—32-nm process
Ivy Bridge—22-nm process
K8—65-nm, 90-nm, and 130-nm processes
K9—processors were never released
Table 3-3 shows a list of common Intel code names and some of their related CPUs. You can see that the Core i3, i5, and i7 family names are frequently repeated.
Table 3-3. Intel Code Names and Processors
CPU Family names
Core 2 Duo, Core 2 Quad, Core 2 Extreme
Intel Pentium, Core i3, Core i5, Core i7, Xeon
Celeron, Pentium, Core i3, Core i5, Core i7
Core i5, Core i7, Xeon
The Core i3, i5, and i7 series represents a Good, Better, Best philosophy, with the i3 versions representing the basic version and the i7 versions providing the most power. The number (such as i3 or i5) doesn’t refer to the number of cores.
It’s also important to realize that there are significant differences between a Nehalem Core i5 and an Ivy Bridge Core i5. The Ivy Bridge versions have smaller processes and are more powerful.
Table 3-4 shows a list of common AMD code names and their related CPUs. The primary AMD CPUs that you find in desktop computers are Sempron, Athlon, and Phenom.
Table 3-4. AMD Code Names and Processors
CPU Family names
Opteron, Athlon 64, Athlon 64 FX, Athlon 64 X2, Sempron, Turion 64, Turion 64 X2
Opteron, Phenom, Athlon, Athlon X2, Sempron
Phenom II, Athlon II, Sempron, Turion II
FX (Zambezi), Interlagos Opteron
CPU Socket Types
A CPU plugs into a socket on the motherboard. There was a time when just about every motherboard had the same socket type, but that certainly isn’t the case today. Instead, there are a wide variety of different socket types for different types of CPUs. If you ever need to replace a CPU, it’s important to recognize that there are different types of sockets. The following sections talk about some sockets used by Intel and AMD, with information about how they are installed.
Zero Insertion Force
It’s important that each of the pins on a CPU has a good connection to the motherboard. In early versions of CPUs, this was accomplished by creating a tight connection between the pins and the socket. This required technicians to use some force to plug the CPU into the socket. Unfortunately, it was easy to bend one or more pins, and bent pins would often break, making the CPU unusable.
Manufacturers came up with a great idea to eliminate the problem—zero insertion force (ZIF) sockets. A ZIF socket has a locking lever. You can place a CPU into a socket without any force other than gravity, and after the CPU is in place, you lock the lever to secure it. This lever ensures that the pins are making a solid connection to the motherboard.
Figure 3-9 shows a ZIF socket with the lever raised. The CPU is removed and standing up on the left. You can see that there are some areas on the CPU where there aren’t any pins. These provide a key, and they match up to areas on the socket where there aren’t any pin holes.
Figure 3-9. Processor and ZIF socket.
PGA vs. LGA
The socket shown in Figure 3-9 is a pin grid array (PGA) type of socket. It includes holes into which the pins can be plugged. A newer type of socket is a land grid array (LGA) socket. Instead of the processor having pins and plugging into a socket with holes, the socket has small pins, and the CPU has small pins created as bumps or pads. When the CPU is installed, the pins and bumps line up, making the connection.
When using an LGA socket, the CPU sits on top of the socket but is locked in place with a flip-top case. Figure 3-10 shows an example of a flip-top case used with an Intel processor.
This socket has a hinged top and a lever that locks the case when it’s closed. You unlock the lever, open the case, and remove the CPU. When installing a new CPU, ensure that the keys line up, place the CPU in the case, close the top, and lock it with the lever. Remember to use ESD protection when handling the CPU.
Figure 3-10. Removing processor from a flip-top case. Diagram provided by Intel. [Copyright © Intel Corporation. All rights reserved. Used by permission.]
Another type of array you might run across is ball grid array (BGA). In a BGA chip, the pins on the CPU are replaced with balls of solder. The chip is mounted in the socket and then heated, often in an oven, to melt the solder. Manufacturers can fit more pins on a BGA CPU, and they are sometimes used in mobile devices.
Intel CPU Sockets
The following list describes recent Intel sockets:
LGA 775. 775 pins. Also called Socket T. Replaced Socket 478.
LGA 1366. 1,366 pins. Also called Socket B and designed to replace LGA 755 in high-end desktop computers.
LGA 2011. 2,011 pins and released in 2011. Also called Socket R. It replaces LGA 1366 sockets in high-end desktop systems.
LGA 1156. 1,156 pins. Also called Socket H or Socket H1.
LGA 1155. 1,155 pins. Also called Socket H2 and replaces LGA 1156 in basic desktop systems. LGA 1,156 CPUs will work in LGA 1155, but the BIOS may need to be upgraded.
Table 3-5 lists the common Intel sockets along with some CPUs used with them, busses they support, and supported DDR channels.
Table 3-5. Intel Sockets and Related CPUs
CPUs, Busses, DDR Channels
LGA 775 (Socket T)
Pentium 4, Pentium D, Core 2 Duo, Core 2 Quad, Celeron, Xeon Front side bus, single channel DDR2 and DDR3 RAM
LGA 1366 (Socket B)
Core i7, Xeon, Celeron QPI, triple channel DDR3 RAM
LGA 2011 (Socket R)
Core i7, Xeon QPI, DMI, quad channel DDR3 RAM
LGA 1156 (Socket H or H1)
Core i3, Core i5, Core i7, Celeron, Pentium, Xeon DMI, dual channel DDR3 RAM
LGA 1155 (Socket H2)
Core i3, Core i5, Core i7, Celeron, Pentium DMI, dual channel DDR3 RAM
AMD CPU Sockets
The following list describes recent AMD sockets:
Socket 940. 940 pins (PGA).
Socket AM2. 940 pins (PGA). Not compatible with Socket 940.
Socket AM2+. 940 pins (PGA). Replaces AM2. CPUs that can fit in AM2 can also fit in AM2+.
Socket AM3. 941 pins (PGA). Replaces AM2+. Supports DDR3. CPUs designed for AM3 will also work in AM2+ sockets, but CPUs designed for AM2+ might not work in AM3 sockets.
Socket AM3+. 942 pins (PGA). Replaces AM3. CPUs that can fit in AM3 can also fit in AM3+.
Socket FM1. 905 pins (PGA). Used for accelerated processing units (APUs).
Socket F. 1,207 pins (LGA). Used on servers and replaced by Socket C32 and Socket G34.
Table 3-6 lists the common AMD sockets along with some CPUs used with them, busses they support, and supported DDR channels.
Table 3-6. AMD Sockets and Related CPUs
CPUs, Busses, DDR Channels
Opteron and Athlon 64 FX
FSB with HyperTransport version 1, single channel DDR2 RAM
Athlon 64, Athlon 64 X2, Athlon FX, Sempron, Phenom, Opteron
FSB with HyperTransport version 2, single channel DDR2 RAM
Athlon 64, Athlon 64 X2, Athlon II, Sempron, Phenom, Phenom II, Opteron
FSB with HyperTransport version 3, single channel DDR2 RAM
Phenom II, Athlon II, Sempron, Opteron
FSB with HyperTransport version 3, single channel DDR2 and dual channel DDR3 RAM
Phenom II, Athlon II, Sempron, Opteron
FSB with HyperTransport version 3, dual channel DDR3 RAM
Fusion and Athlon II APUs
FSB with HyperTransport version 3, dual channel DDR3 RAM
Opteron, Athlon 64 FX
FSB with HyperTransport version 3, single channel DDR2 RAM
Earlier in this chapter, I listed two CPUs using common marketing names. To tie some of this together, here are the two CPUs with an explanation of the names. I’m hoping these names make a lot more sense at this point.
Intel Core i7-960 Processor 3.2 GHz 8 MB Cache Socket LGA 1366. This name indicates that it is an Intel processor in the Core i7 family with a model number of 960 and a 3.2-GHz multiplied clock. The 8-MB cache phrase refers to the total amount of cache. Last, LGA 1366 indicates the type of socket into which the processor will plug.
Phenom II X4 965 AM3 3.4 GHz 512 KB 45 NM. This indicates that it is an AMD Phenom II processor with a model number of 960. X4 indicates that the processor has four cores, and AM3 indicates the socket type. The 3.4-GHz clock speed is the internal speed of the processor. Cache size is indicated by 512 KB, and in this case, it indicates the L2 cache size for each of the cores. The process is 45 nm.
CPUs have millions—and sometimes billions—of miniaturized transistors within them, all connected with extremely small wires. If these transistors or wires get too hot, they can easily break, rendering the CPU useless. Manufacturers spend a lot of time designing these chips, and one of their goals is to keep temperatures within acceptable limits. However, most of the cooling occurs externally.
Heat Sinks, Fans, and Thermal Paste
Common methods of cooling a CPU include using a heat sink, a fan, and thermal paste. Take a look at Figure 3-11 as you read about how these components work together.
Figure 3-11. CPU with heat sink and attached fan.
Heat sink. A heat sink is a piece of metal that draws heat from the CPU and dissipates it into the air. Heat sinks have multiple fins to increase the surface area and to allow air to easily flow through them. The fins are usually flared to allow more air through.
Fan. A fan is attached to the heat sink to increase the airflow around the fins. These are called CPU fans. They aren’t attached to the CPU but usually plug into the motherboard close to the CPU. Many CPU fans have variable speeds and spin faster when the CPU gets hotter.
Thermal paste. Heat sinks commonly have clamps to secure them to the motherboard and provide a better connection with the CPU. However, there are microscopic gaps in the metal on both the CPU and the heat sink, so it isn’t possible to get 100 percent contact between the components. Thermal paste is used to improve this connection. This paste fills these microscopic gaps and also helps draw heat from the CPU into the heat sink.
If you are replacing a CPU, you’ll need to clean off the old thermal paste from the heat sink. Some vendors sell specialized cleaning compounds to remove old paste, but you can often use cotton swabs and isopropyl alcohol to remove it.
After installing the new CPU into the socket and locking the ZIF arm, place a dab of the paste in the center of the CPU. When you attach the heat sink and clamp it down, the pressure will spread the paste evenly between the heat sink and the CPU. Be careful not to apply too much paste; you need only enough to fill the microscopic gaps between the CPU and the heat sink.
An advanced method of keeping a system cool is using a liquid-based cooling system. Liquid-based cooling systems use water (most commonly) or some other liquid that is pumped through the cooling system.
For example, Figure 3-12 shows a basic diagram of a liquid-based cooling system. A specialized heat sink is attached to the CPU, using thermal paste just like a standard heat sink. However, this heat sink has channels so that the liquid can flow through it. Tubing is connected from the pump to the heat sink, and the pump constantly pumps the liquid through the heat sink.
Figure 3-12. Liquid-cooled heat sink.
One of the biggest challenges with a liquid-based cooling system is ensuring that the tubing connections do not leak. This is one place where you don’t want to skimp on quality. The liquid is usually water, and if it leaks, it could easily destroy the system.
Liquid-based cooling systems are most common among gamers and hobbyists. These people often overclock the processors to get more power out of them, but overclocking generates more heat. Overclocking is sometimes possible by changing jumpers on the motherboard or by manipulating BIOS settings, but manufacturers discourage the practice.