Note Taking: How Memory Works Student Name:_______________________________________________________ Date:___________
Directions: Use this resource to take notes over the next three class periods. Five of these questions will be on the quiz at the end of this lesson. On the quiz, you will need to write two or three sentences to answer each question.
1. How is memory different from permanent storage?
2. Explain the difference between RAM and ROM.
3. What is cache?
4. How does cache compare with DRAM?
5. Explain the differences between SRAM, DRAM, flash memory, and virtual memory.
6. How is disk cache different from virtual memory?
This presentation compares storage to different types of memory, including random-access memory (RAM), read-only memory (ROM), cache, and virtual memory. It also explores basic troubleshooting procedures.
Also called main memory, RAM is high-speed temporary memory. RAM holds currently used data so that the processor and the operating system can access it quickly. RAM allows both “reads” and “writes,” which means that data in RAM can be displayed and modified.
The more RAM a computer has, the better the system will operate. However, you have to check to see what the maximum amount of RAM is for any given system. There is an upper limit on the RAM size defined by the processor address space. For example, the upper limit for a 32-bit processor is 4GB.
RAM accesses data in electronic blocks that the system can access in any order (thus the word random in the name). Before RAM, memory was linear access memory, which worked like an audio tape cassette. Trying to find data was similar to forwarding an audio tape to a song you wanted to hear.
Memory modules were developed to change the way data was accessed. Memory modules use DRAM technology to store and read data. A DRAM module stores bits of data in capacitors, which are electronic devices that can hold a charge. When we refer to main memory, or the large quantity of memory in a computer, we are referring to DRAM.
The term storage is used to refer to data storage on a hard drive, solid-state drive, CD-ROM, or other medium that holds data permanently. Accessing data from storage is slower than accessing data that is in memory.
The hard drive is the most common secondary storage device on a computer, but solid-state drives are becoming more common. A solid-state drive (SSD) is a data storage device that uses an integrated circuit assembly to store data. An SSD does not use a moving mechanical component, which distinguishes it from a magnetic disk like a hard drive. SSDs are much faster than traditional hard drives but the cost per gigabyte is higher.
The components in your computer—the processor, hard drive, memory, and operating system—work together as part of the whole system.
Unlike RAM, which reads and writes data, read-only memory (ROM) can only be read. The memory on ROM is permanent, or nonvolatile. Data remains stored in ROM even when the computer is powered off.
The ROM chip holds the basic input/output system (BIOS). The BIOS ROM chip stores 256KB to 512KB of instructions; newer BIOS versions can store even more.
ROM chips are manufactured with their data in place. There are several variations of ROM chips, including the following:
• Blank programmable read-only memory (PROM): these chips are not written to when they are manufactured and instead come blank. Once they are written to, the chips cannot be erased.
• Erasable, programmable, read-only memory (EPROM): A programmable ROM chip that retains its contents until it is exposed to ultraviolet light, which enables the memory to be reprogrammed. EPROM chips are rapidly becoming obsolete. They are being replaced by EEPROM and flash memory.
• Electrically erasable, programmable, read-only memory (EEPROM): A programmable ROM chip that can be erased and reprogrammed one byte at a time by exposing it to an electrical charge.
A special type of EEPROM chip that can be erased and reprogrammed in blocks rather than bytes is called flash memory. A BIOS that is stored on a flash memory chip is called a flash BIOS. EEPROM chips are known to have a limited number of erasures before they fail. Modern EEPROM chips are estimated to allow for as many as 100,000 erasures.
Cache, also known as static RAM (SRAM), is much faster than main memory (DRAM). It is the fastest and most expensive memory in the PC. It is also not something a user can simply “upgrade,” as it is built in to the processor and the motherboard.
Cache acts as a buffer between a fast device and a slow one. Over time, processor speeds have increased more quickly than speeds of other components. Because main memory could not send data to the processor fast enough, cache was added to the processor. Cache holds data in transistors, so it can send data to the processor very quickly. Cache minimizes the number of clock cycles that the processor must wait for data.
When the CPU needs to access new data, the operating system first checks to see whether the data is in the cache before it looks in the RAM or the hard drive.
Some cache is designed to predict what data will be requested next so that data can be sent to the cache before it is actually needed. When a processor requests data, chances are good that the next request will be for data stored nearby. So, when the processor requests data from memory, memory sends the requested data plus the nearby data. All the data is stored in processor cache. That way, subsequent data requests can be filled from the cache rather than from the slower memory.
Note that both SRAM and DRAM are forms of RAM (random-access memory). SRAM is faster but more expensive, so computers have a substantially smaller amount of SRAM than DRAM.
Level 1 (L1) cache is located on the chip with the processor, and it operates at nearly the same speed as the processor. While the CPU is busy crunching data, the L1 cache tells the CPU controller where to find the data the CPU might need. L1 cache should have at least a 90% hit ratio; that means that 90% of the time, the L1 cache successfully finds the data the processor needs next. When the CPU looks for an item in L1 and it is not found, it is known as a miss and the CPU looks instead at the L2 cache.
In older systems, Level 2 (L2) cache and possibly a Level 3 (L3) cache are located on the motherboard and connected to the processor through the system bus. These L2 and L3 caches are slower than L1 cache but faster than memory. In newer high-end processors, L2 cache, and sometimes even L3 cache, are on the processor rather than the motherboard.
Multilevel caches generally operate by checking the smallest (L1) cache first; if it hits, the processor proceeds at high speed. If the smaller cache misses, the next larger cache (L2) is checked, and so on, before main memory is checked. If the item is not in main memory either, then it will be held in virtual memory, stored on the swap space portion of the hard disk.
Although cache is fast, it is more expensive than DRAM. DRAM stores a bit with a single transistor/capacitor pair, but cache requires six to eight transistors to store the same bit. The more transistors there are, the more expensive the chip is to produce. Think of DRAM as consisting of capacitors requiring a refresh, whereas the faster SRAM is built with active, current-carrying transistors that don’t require a refresh, and can provide faster access.
Since the transistors heat up the cache, heat sinks are required to dissipate the heat. This also increases the cost.
Disk cache, also referred to as the disk buffer, is the memory in a hard drive that acts as a buffer between the computer and the physical platter in the hard drive. Disk cache speeds up the time it takes to read from or write to a hard drive because it holds data that has recently been read from the disk. Sometimes it also holds “pre-fetch” information, which is nearby data likely to be requested in the near future.
However, this holding area can be risky. If the power fails and the computer crashes, the system might not have time to copy the information in the cache back to the hard drive. Any changes you might have made would be lost. This doesn’t happen often, because the system usually updates the disk cache frequently enough that only a small amount of data would be lost, if any. This type of cache is called write-back cache. Write-through cache caches data only for read operations; write operations are always sent directly to the hard drive.
Up to 32MB of disk cache is usually included as part of a hard drive. A microcontroller in the hard drive controls the disk cache.
Virtual memory is a reserved space on the hard disk known as swap space, and it is used to store overflow data from DRAM. When running programs will not fit in DRAM (either because the programs are too large or you are running too many programs to fit), only portions of the programs and their data are placed in DRAM. The remainder is placed in virtual memory.
The address space of a 32-bit processor allows 4GB of RAM in practice, but even if the 4GB is available, it can’t be allocated to a single program. So each program uses a smaller portion of this memory and the rest is stored on disk. This allows all programs running in the system to think they have 4GB of memory.
As the CPU moves through the program and the data, if an item is needed that is not in DRAM, it must be retrieved from swap space. This process of moving items from swap space to memory is called swapping.
While virtual memory allows us to extend the size of DRAM, it comes at a cost. The disk drive operates at a speed up to 1 million times slower than DRAM, so using virtual memory slows down the processor drastically. The more virtual memory is used, the slower processing becomes.
Another problem with virtual memory is that in some rare cases, items are swapped in and out of DRAM and virtual memory often, resulting in a situation known as disk thrashing. Thrashing may continue for seconds or longer, in which time processing essentially comes to a halt.
If you are running programs that require a lot of memory or are running many applications simultaneously, you might notice poor performance. If you do, check the activity light on the hard drive. If it is flickering when the computer is not actively reading or writing to the hard drive, the operating system is probably running out of physical memory and accessing the slower virtual memory on the hard drive. In fact, nearly every time you see the hard disk drive light flicker when you are not opening or saving files, the system is swapping between DRAM and virtual memory.
Some applications generate error messages when available memory runs low, but many do not. To resolve this problem, shut down some applications or add more memory to the PC.
Bad memory often causes sudden and unpredictable crashes. If you recently upgraded the system memory and suddenly the system seems unstable, suspect a bad memory module. To test the module, remove and reseat it. Swap memory modules between slots and banks. Check whether the problem follows a specific memory module as you move it or whether it stays with a bank or slot.
To prevent problems caused by corrosion, be sure to match the module connector materials with the system memory connector materials (gold plated to gold plated and tin-lead plated to tin-lead plated).
Be sure that all memory modules in the same bank are the same size and speed. Also be sure that the memory is as fast as the minimum required system speed. To diagnose, reduce the memory to the minimum size and speed to perform a successful power-on self-test (POST).
Many components in a computer system require their own memory in order to store data. For example, the video card typically has its own dedicated memory as well as a dedicated microprocessor called a GPU (graphics processing unit).
Knowing which type of memory works best in different parts of the PC helps you achieve top performance in your system.
Reading: Types of RAM An elephant never forgets, but does a computer? We think of our computers as having unlimited capacity for storing and remembering things: our papers for class, pictures of our vacation, and family holiday videos. But the truth is that most of a computer’s internal memory is actually quite short and needs constant help to hold on to the information we give it. Let’s take a closer look at the different types of RAM and how they work.
To start, all RAM is stored in memory chips. In the first personal computers, memory chips were attached directly to the motherboard. But as people used computers more, their memory requirements increased, and the motherboard grew overcrowded. As a result, memory chips moved to cards called memory modules.
Each memory module is lined on one or both sides with memory chips. Electrical traces lead from the memory chips to gold pins at the bottom of the memory module. These pins plug into slots on the motherboard, enabling the motherboard to hold significantly more memory.
Types of RAMThere are two basic types of RAM:
Dynamic RAM (DRAM): RAM that is dynamic is constantly being rewritten.o Advantages: Cheapest o Disadvantages: Must be refreshed every 8 nanoseconds or data will be losto Use: Most common type, used in main memory
Static RAM (SRAM): RAM that is static does not need to be refreshed. o Advantages: 10 to 25 times faster and more reliable o Disadvantages: More expensive to produceo Use: Only in cache memory and registers where greater speed is needed
While both types of RAM are volatile, SRAM loses its data only if the computer loses power, whereas DRAM must be refreshed nearly constantly to preserve data.
RAM EvolutionOver time, RAM has grown as follows:
30-pin SIMMs using four memory modules (8 bits per stick to equal 32 bits) 72-pin SIMMs using two sticks (16 bits per stick to equal 32 bits) 168-pin DIMMs needing only one stick (64 bits per stick) 184-pin DIMMs (SDRAM synchronized to the front-side bus) 240-pin DIMMs (DDR, DDR2, or DDR3)
Types of Memory ModulesRAM is available in two form factors, also called sticks:
Single inline memory module (SIMM) cards: SIMMs have redundant electrical contacts on both sides of the module.o Pins: 72o Data path: 32-bito Each SIMM handles 16 bits, but most processors such as Intel Pentium processors require at
least a 32-bit path to memory. So, you must install SIMMs in matching pairs. Note: SIMM modules are found only in older PCs.
Dual inline memory module (DIMM) cards: DIMMs have separate (non-redundant) electrical contacts on each side of the module.o Pins: 168, 184 (DDR SDRAM) or 240 (DDR2 and DDR3 SDRAM)o Data path: 64-bito Each DIMM handles 64 bits, so they only need to be installed one at a time. However, to take
advantage of dual-channel memory (discussed later), you must install DIMMs in pairs also.
Registered Memory vs. Unbuffered MemoryMemory modules can be divided in one additional way: registered and unbuffered.
Registered Memory: Registers buffer the signals that the memory controller sends by temporarily holding address and command data for one clock cycle before it is transferred. o Advantages: Reduces the electrical load; more reliable high-speed data access to high-
density memoryo Disadvantages: More expensive; decreased performanceo Use: High-density systems including applications that use more than 4GB of memory; servers
or high-performance workstations (can only be used on motherboards that can support it) Unbuffered Memory: The memory controller directly addresses each memory chip on all
modules in the system.o Advantages: Cheaper; better performanceo Disadvantages: High electrical loadso Use: Almost all system memory in today’s PCs is unbuffered memory.
Installing MemoryThe actual installation of RAM is pretty straightforward. The memory module is placed in the appropriate slot on the motherboard. However, based on the type of memory module you use, there are a few key rules to follow to ensure your RAM functions properly:
You can install a single DIMM. However, installing pairs of DIMMs in computers that support dual-channel DDR SDRAM increases performance.
When different DIMM sizes are mixed on the motherboard, you must put the largest DIMM in the first bank.
Each bank of memory for a SIMM has two sockets. You must fill the first bank before filling the next.
Each bank must be filled with RAM modules that have the same access time and size.
Before upgrading memory, learn which combinations of modules are supported in the memory slots on the motherboard and whether there are restrictions. You can find more information on specific memory types in user manuals, in the motherboard manual, or on manufacturers’ websites.
RAM Basics Let’s take a look now at how RAM actually works.
Have you ever heard of a memory palace? This is a technique originally developed in ancient Greece and Rome, and used by the top memory experts as they compete in memory competitions. In its simplest form, you associate each item you have to remember with items in a place very familiar to you (like your house).
Computer memory works in pretty much the same way. The computer takes the data you give it—the bits—and translates it into code the computer understands: electrical charges. Each data bit is placed in its own capacitor, an electronic component that holds a charge in the form of an electrostatic field. The capacitor then “remembers” the bit by translating it: an electric charge is equivalent to a binary 1. A capacitor without a charge is equivalent to a binary 0.
Much like your memory palace, the DRAM has a physical structure to help you find what you are looking for. Rather than a house (or a palace), the DRAM is composed of what is sometimes called a memory matrix, a grid of rows and columns. Each row is numbered, as is each column. At the intersection of each row and column sits a capacitor. So, the address of each capacitor—and its bit—is composed of a column address and a row address.
To find a specific bit of data in the DRAM, a signal, called a column address strobe (CAS), is sent by the processor or memory controller to identify and activate the capacitor’s column. At the same time, a row address strobe (RAS) is sent to identify and activate the capacitor’s row. The combination of the right column and row lead the processor to the unique capacitor and bit.
This process is used to both write and read data. When writing, RAS and CAS either charge or discharge the capacitor at that intersection, setting it to a 1 or a 0. When reading, a sensor measures the charge of the capacitor at that intersection. If it is charged, the DIMM sends a 1 to the memory controller. If it is not charged, the DIMM sends a 0 to the memory controller.
Remember, though, that DRAM is volatile. Each read is a destructive read; the act of reading the contents of a DRAM address causes that electrical charge to dissipate. This is why DRAM must be recharged thousands of times every minute. Essentially, while the DIMM is reading, it simultaneously refreshes the capacitor by either resetting the charge back (for 1) or leaving it untouched (for 0).
It is important to note that all DRAM is not created equal. DRAM varies in speed based on three factors:
CAS latency (CL): The number of clock cycles it takes to access a column or row of data. The lower the number, the better. For example, a CL of 2 means two clock cycles to read or write data; a CL of 3 would be slower.
Front-side bus speed: The faster the memory controller and the processor can communicate, the faster the overall retrieval.
Refresh rate in nanoseconds: A lower refresh rate would allow more energy to go to the actual reading of the capacitor.
BanksNow that you know how memory is stored and read, let’s take a look at how the processor and memory controller communicate with the RAM.
On the motherboard, memory chips are divided into independent sections called banks. The memory controller or the processor sends requests to a drive over the bus. Generally, a bank can perform only one memory access at a time. So, if a computer had only one memory bank, memory access would be a slow process. This is why most computers have two or more memory banks, allowing them two—or more—independent accesses simultaneously. For instance, one access might come from the CPU while another comes from the disk drive.
A CPU (for example) can also send requests to multiple memory banks at a time. While only one memory bank at a time can send instructions back over the bus, they can all look up their contents simultaneously. Because memory accesses are much slower than bus transfers, this form of “pre-fetching” of instructions can speed up processor execution.
BusesBecause accessing memory is one of the primary functions of a computer, it needs a serious bus, known as the system bus. The system bus is composed of three parts:
Front-side bus: This is the first step in the communication process. All the processors use this bus to connect to the memory controller (the north bridge). Because this bus is shared, the memory controller acts as a traffic controller, deciding the order in which signals pass onto the memory bus.
Memory bus: The memory controller uses this bus to pass the signals from the processors onto the memory modules, via the gold pins on them. Instructions are then sent back to the memory controller over this bus, and then over the front-side bus to the processors.
Back-side bus: Because the memory in a processor’s off-chip cache needs to be accessed frequently, it has its own dedicated bus. This bus is significantly faster than the front-side bus, with a speed that matches the processor.
Dual-Channel MemoryIf the system bus were a network of roads, the front-side bus would be a four-lane road, and the memory bus a single lane. This could lead to a serious slowdown in system traffic, as signals have to wait their turn to be sent down the single-lane road.
So how do you speed things up? Add another lane. If DRAM has two or more DIMMs, dual-channel memory can separate it onto two different buses. Two different memory controllers, then, can access them in parallel. Each DIMM must be accessed separately still, but instead of one lane of 64 bits of data coming through, we now have two lanes, doubling the throughput.
Making Sure It’s Right: ParityWhen this entire process goes right, RAM is a wonderful improvement to a computer. But, as you can imagine, with millions of bits flying back and forth over multiple buses and the electrical charges that record them constantly fading, there is a lot of room for error. Even one missing bit can stop a computer’s functioning in its tracks. So what do you do when the DIMM is not working properly?
Some computers use error-detecting technology to find corrupted data and, in some cases, fix it. But how do you find the error in the first place?
Computers use a system called parity checking, which looks for errors in data transfer by checking whether data has been lost or overwritten. Many PCs perform a parity check on memory every time a byte of data is read.
Here is how it works: Every byte of storage gets an extra bit called a parity bit. A parity check adds up all of the 1s in a byte, plus the parity bit. When a system is constructed, the computer architect decides if parity will be odd or even. There is no real difference between odd and even parity; it just determines what the memory controller is looking for. In even parity, we always expect the number of 1 bits to be an even amount; in odd, an odd amount. So, whether a 1 bit or a 0 bit is added to the byte is determined by how many 1s are already in the byte. For example, in even parity, the byte 11110000 would have a parity bit of 0 added because there are already an even number of 1 bits (4 of them). In odd parity, that same byte would have a parity bit of 1, so that the total number of 1s would equal 5 (an odd number).
Now, let’s say a system has even parity. If the memory controller detects that the byte has an odd number of 1s, it will know there is an error in that byte. It then recalculates the parity bit and compares the new parity bit with the stored parity bit. If the new parity bit does not match the stored parity bit, the memory transfer stops.
Parity checking only works if both the sender and receiver agree to use it, and if both sides are configured with the same parity sense (odd or even). Otherwise, communication is impossible.
ECCAnother way to find an error is to use an error checking and correction (ECC) system. Whereas parity checking can find an error, an ECC system can find and repair an error.
Instead of adding a parity bit to each byte of data, an ECC system adds 8 check bits to every 64 bits of data, creating a 72-bit checksum. Memory modules with ECC capabilities have nine memory chips on each side, instead of eight chips as on non-ECC memory modules. The 72-bit checksum is stored on the separate ninth chip.
The memory controller sends the 72-bit checksum to the memory modules. If an error occurs, the processor requests the data, and the DIMMs send the checksum to the memory controller. The memory controller determines which bit is faulty and determines whether the error was only in one bit or across several bits.
The most common type of ECC memory performs single-bit error correction. If a single-bit error has occurred, the memory controller corrects the error and sends the corrected 64 bits of data to the processor. It can also detect two-bit and some multiple-bit errors but can’t correct them. If a multiple-bit error has occurred, the computer stops.
Technically, the ECC memory bus is 72 bits wide to allow for the entire checksum to be transferred, while non-ECC memory is only 64 bits wide. However, only 64 of those bits are counted for bandwidth and applications (the other 8 bits are all check bits), so effectively the bandwidth of ECC and non-ECC memory is the same.
ECC memory is not essential for most PCs but is very important in workstations and servers, and works best with registered memory. In order to take advantage of its correction capabilities, the memory controller on the motherboard must support ECC memory.
Directions: Research the computers listed in this chart on the Internet. Kingston.com is a good site for this. You can also try the website of the computer manufacturer.
For each computer model, fill in details about the type of RAM supported and configuration options such as the number of memory slots, the maximum amount of RAM, operating system requirements, or other requirements that might impact your choice of RAM. The first row has been completed for you.
Computer Model Number and Type
Type of RAM Supported Configuration Options
Dell Inspiron 17R notebook computer
Shared single- or dual-channel DDR3 SDRAM
Supports Windows 8 64-bit OS
6GB maximum memory; 6GB standard
Up to 1TB SATA hard drive (5400 rpm) or up to 750GB SATA hard drive (7200 rpm)
Lenovo IdeaCentre K430 ATX series desktop computer
Oracle Netra Blade X3-2B Family blade server
Generic desktop clone with a Gigabyte Intel desktop Z77 ATX motherboard
Directions: Using the computer assigned to you, complete each task listed below by following the instructions for your particular operating system. Then write your observations about what you change and what you discover in the space provided.
Task Windows Vista Windows 7 Windows 8 or 10
View the virtual memory settings on your computer
Start > Control Panel > System > Advanced System SettingsOn the Advanced tab, click the Settings button in the Performance box. In Performance Options, click the Advanced tab. The Virtual Memory box shows the paging file size.
Windows icon > Control Panel > System > Advanced System SettingsOn the Advanced tab, click the Settings button in the Performance box. In Performance Options, click the Advanced tab. The Virtual Memory box shows the paging file size.
My Computer > Properties > Advanced System SettingsOn the Advanced tab, click the Settings button in the Performance box. In Performance Options, click the Advanced tab. The Virtual Memory box shows the paging file size.
Observations
Check how much memory your computer is using
Right-click the taskbar, then click Task Manager > Performance tab
Right-click the taskbar, then click Task Manager > Performance tab
Right-click the taskbar, then click Task Manager > Performance tab
Start > All Programs > Accessories > System Tools > System Information > System Summary > BIOS Version/Date
Windows icon > All Programs > Accessories > System Tools > System Information > System Summary > BIOS Version/Date
Search for System Information on the Start screen, and then select System Summary > BIOS Version/Date
Observations
Access the BIOS
Procedures vary depending on manufacturer. Usually, you must press a key (such as F2, F12, Delete, or Escape) or a key combination immediately after you turn on your computer but before the OS starts. For more information, check your user’s manual or visit the computer manufacturer’s website.
