PC - Motherboard (II)

Choosing The Right Chipset


The gateway between a processor and other parts of the computer is a set of interface controllers generically called the chipset. Traditional chipsets include a Northbridge, with a memory controller and either a PCI Express or an AGP interface for graphics, and a Southbridge, containing the standard PCI controller and various peripheral/communications buses for networking, audio and other components.

Example of an intel chipset. 440BX Chipset was very popular in motherboards for iPentium II

Though single-component chipsets have been around for many years, AMD's CPU-integrated Athlon 64 memory controllers removed one task from the Northbridge, making it easier for companies to integrate Northbridge and Southbridge components into one part.


The Northbridge

Traditional Northbridge designs include a memory controller linked directly to the CPU through the Front Side Bus. Early chipsets used a common CPU and memory bus frequency, so that the entire pathway was often called the Front Side Bus. Later chipsets allowed separate CPU and memory bus frequencies, limiting the acronym "FSB" to the pathway between CPU and Northbridge. AMD later removed the memory bus entirely from the chipset in its Athlon 64, with separate Northbridge and RAM pathways on the CPU replacing the Front Side Bus.

What remains common to all Northbridge designs is an AGP or PCI Express controller and a Southbridge interface (internal on single-component chipsets). Some Northbridge designs also incorporate a graphics processor, using either AGP or PCI Express interfaces internally.

Nvidia chipsets normally offer higher performance than chipsets of SiS or Via

Intel Quad Data Rate Northbridge Technology

Intel's quad-pumped quad data rate bus transfers data four times per clock cycle, so that 100, 133, 200 and 266 MHz clock rates provide effective 400, 533, 800 and 1066 MHz data rates. Because CPU Front Side Bus data rates are often twice as high as those of DRAM data rates, the performance solution has been to double the memory bus width from 64 to 128 bits by placing two modules on parallel pathways, which involves a technology called dual channel mode.

As an example, two DDR400 (PC3200) memory modules in dual-channel mode have the same bandwidth as Intel's FSB800 CPU bus, and both operate at an actual clock rate of 200 MHz. The same can be said of matching two DDR2-533 modules in dual-channel mode to Intel's FSB1066 Front Side Bus.

We recommend choosing a dual-channel supporting chipset if you want to opt for such a solution. Current PCI Express Northbridge technology supports DDR2-SDRAM in dual-channel mode.

This is a scheme of a Via chipset for AMD Athlon 64 where we can see most of the features supported by current motherboards

HyperTransport Interconnect Technologies (S754, S939, AM2)


With the memory bus removed from the Northbridge, AMD chipsets are able to mix-and-match older and newer technologies far more easily. AGP chipsets originally destined for use with Socket 754 have also been made available for Socket 939 buyers, PCI Express chipsets designed for Socket 939 have been used in Socket 754 motherboards, and AM2 motherboards using the previous Socket 939 generation of chipsets are available.


The Southbridge


The Southbridge contains most peripheral, multimedia, and communications busses, including the PCI controller (Peripheral Components Interconnect), ATA controller (for hard drives and optical drives), USB controller (Universal Serial Bus for external devices), network controller interface, audio controller interface, and often even a modem interface. Most chipsets from any given time period offer similar performance, but reviews can inform the buyer of potential inadequacies, such as which recent Southbridge designs had the worst Serial ATA performance, and which had sub-par USB performance.

Another feature of current motherboards is RAID mode for Serial ATA controllers, which allows the user to run up to four drives in a secure or high-performance array.

Gigabit networking is now mature, and most chipsets include dedicated links to a Gigabit PHY (physical chip, which takes care of the physical connection). High-end chipsets typically feature two links for two gigabit connections

Read more »

PC - Motherboard (I)

The support for a desired CPU is key to choose the motherboard. But a motherboard is far more than simply the device a CPU plugs into. What we must avoid buying a motherboard is an inappropriate component selection that may limit the performance of a high-priced system, making the added expense a waste of money. That is to say, we have to avoid bottlenecks.

There are three basic steps in order to choose our future motherboard:

A) Choosing parts that fit and work well together. For that we have to consider: motherboard size, socket type and chipset features.

B) Getting the best performance. Memory configuration and graphics support is what we have to consider concerning performance

c) Finally, ultimate functionality: Onboard devices and/or additional card slots have to be always considered.

The following picture shows some of this elements:

This is a Socket 775 motherboard from Foxconn (image obtained from www.tomshardware.com)

Choosing The Right Processor Socket

Current possibilities for sockets are especially:

LGA 1366
Intel Core i7. Integrated memory controller, DDR3 support and three memory channels

LGA 775
Intel Core 2 Quad&Duo, Pentium 4, Pentium D, Celeron. DDR2 support and two memory channels (dual channel)

Socket AM2
AMD Phenom, Athlon 64 & 64 X2 & 64 FX, Sempron. Integrated memory controller, DDR2 support and two memory channels (dual channel)

Socket 939
AMD Athlon 64 & 64 X2 & 64 FX. Integrated memory controller, DDR support and two memory channels (dual channel)


Memory

Memory type and configuration limits are normally thought to be those of the memory controller, but motherboard's slot configuration can further limit choices. For example, several smaller motherboards provide only two memory slots, both wired to a single memory channel, making a dual-channel capable memory controller pointless. Certainly, it's best to have at least four DIMM slots.

In the following posts I will keep explaining more about points A), B) and c)

Read more »

PC - RAM Memory (III)

SUMMARIZING

After reading the main features and parameters of RAM memory I have explained in the previous two posts, you are able to decide which RAM configuration you need for your personal computer.

But to summarize, here are my final recommendations for RAM memory purchases, which will remain valid even for the next generation platforms:

• I recommend to buy Kingston RAM memory modules.
  

• 2 modules of 1 GB or 2 modules of 2 GB of DDR2-800 or DDR2-1066 memory. It's not worth spending more money on DDR3 memory yet.
  

• Always get as few memory modules as possible, but equip all memory channels of a system to get maximum performance. The ideal configuration consists of two DIMMs today, and three DIMMs with Intel’s upcoming Core i7 systems.
  

• Look for quick timings, latencies for example (smaller numbers are better), but it's not worth spending substantial money on slightly faster modules.
  

• Spend additional budget on a faster processor, a faster graphics card or a faster drive instead of fast memory, unless you need overclocking memory

Read more »

PC - Memory DDR2 or DDR3?

What kind of memory: DDR2 or DDR3?


DDR3 advantages

One major factor favoring the purchase of DDR3 memory is that Intel is slowly moving all of its chipsets in that direction. Motherboard manufacturers expect big spenders to be the earliest adopters of new technology, so the majority of ultra-expensive motherboards will likely support only this latest memory standard while it gradually works its way into lower-cost markets.

DDR3 RAM memory module

While most PC builders won't "need" anything faster than mid-priced DDR2 for a while, DDR3 holds two key benefits over the technology it replaces: First, its maximum chip density has been extended to 8 Gb, allowing a 16-chip module to support a maximum 16 GB capacity. Second, its default voltage has been reduced to 1.50 volts from DDR2's 1.80 volts, resulting in a 30% power consumption decrease per clock speed.

However, the latest technology always comes at a significant penalty in value and DDR2 is more than sufficient for most systems, so why do we move to DDR3? Intel is likely preparing the desktop market for something big, specifically a move of the memory controller from the chipset to the CPU itself with its Intel i7 new processor. As with AMD's current products, this design eliminates the bandwidth limitations of a FSB and allows future processors to receive data as fast as it can be translated.

DDR2 and other RAM memory modules used currently

DDR3 drawbacks


It's often argued that DDR2 memory isn't fast enough for today's processors, as Intel's current fastest Front Side Bus (FSB) uses a 1333 MHz data rate. But this newer FSB doesn't require a "1333 MHz" memory to perform. As I explained in a previous post, Dual-Channel technology has stuck around so that today's FSB-1333 is easily fed by two DDR2-667 (PC2-5300) modules in dual-channel mode.


Another argument for buying DDR3 RAM memory could be for running memory "synchronously" to the CPU's FSB. But DDR3-1333 isn't synchronous to FSB-1333. Intel's FSB uses Quad Data Rate technology while the memory is only Double Data Rate. FSB-1333 runs at a 333 MHz clock rate, which is the same clock rate as DDR2-667.


Conclusion

So, in spite of DDR3 benefits, I think that it's not worth buying DDR3 memory yet. In my opinion, its benefits don't justify its price until new motherboards and processors compatible with DDR3 memory are cheaper and more developed (with a faster BUS for example). In the next post I will try to explain the key parameters I would take into account to decide which motherboard to buy

Read more »

PC - RAM Memory

After explaining a bit the microprocessor, here I will try to explain the key parameters I would consider to decide the RAM memory configuration for my Personal Computer:


Manufacturer

I strongly recommend to buy Kingston memory.


Number of modules

It's always advisable to minimize the number of memory modules. If we have four modules, for example, this may force the BIOS to work with relaxed timings and can cause memory compatibility issues and that the memory will work slower. The best choice is two modules to take advantage of dual channel.

Different RAM types. From top to bottom DIP, SIPP, SIMM 30 pin, SIMM 72 pin, DIMM (168-pin), DDR DIMM (184-pin).

Quantity of memory

Although 8 GB memory kits (which consist of four 2 GB DIMMs) are affordable, these only make sense if you have applications that really take advantage of the increased memory capacity. But these kits might give you the same headache as 3 GB kits, because all of them include four memory modules. This may force the BIOS to work with relaxed timings and can cause memory compatibility issues. A memory capacity of 2 GB is enough. If you think that you will need more memory, the best choice is to buy 2 modules of 2 GB each (a kit of 4 GB).

Dual Channel

Although Dual channel is a concept related to motherboards rather than memory, it's worth explaining here its importance. Dual channel can double the memory bandwith provided that the memory modules are completely identical.


Frequency

The more high is the frequency, the more fast the memory can work. But the frequency of the memory doesn't have to be much higher than the frequency of the bus. In this situation, the frequency of the bus is a bottleneck and the extra money we have spent on this memory will be wasted. It's extremely important to match the frequency of the RAM memory with the frequency of the bus.

For example, Intel's current fastest Front Side Bus (FSB) uses a 1333 MHz data rate. But this newer FSB doesn't require a "1333 MHz" memory to perform. Dual-Channel technology has stuck around so that today's FSB-1333 is easily fed by two DDR2-667 (PC2-5300) modules in dual-channel mode. In this example the RAM memory matches the frequency of the bus. Besides a DDR2-667 (PC2-5300)memory is synchronous to FSB-1333 because both runs at a 333 MHz clock rate.

So a DDR3-1333 memory isn't synchronous to FSB-1333 as Intel's FSB uses Quad Data Rate technology while the memory is only Double Data Rate. FSB-1333 runs at a 333 MHz clock rate, which is the same clock rate as DDR2-667.

Bandwith

The bandwith of a memory can be defined as:

Bandwith (Mbytes/seg) =Frequency of the memory (MHz) x 8 bytes

The easy method to convert data

rate to bandwidth is to multiply by eight. Thus, DDR-400 is called PC-3200; DDR2-800 is called PC2-6400 and DDR3-1600 is called PC2-12800.

The math behind this conversion factor is simple: PC memory modules based on SDRAM technology use a 64-bit connection; there are eight bits in a byte and 64 bits equal eight bytes. For example, DDR2-800 transfers 800 megabits per pathway per second; its 64 pathways provide one eight-byte transfer per cycle and 800 times eight is 6400. For example, today's faster memory is DDR3-1333 which has a peak bandwidth of 10666 MHz. This memory is commonly labeled as either PC3-10600 or PC3-10666.

Latency

RAM memory is structured in a table of rows and columns. The latency is the time elapsed, measured in clock cycles, between the request of reading a data and the moment in which this data is available. Values of latency are usually CL2'5, CL3, CL4, CL5.

Speed Vs. Latency

There's a myth that every new memory format brings with it a latency penalty. The myth is perpetuated by the method upon which latency labels are based: Clock cycles.

Consider the latency ratings of the three most recent memory formats: Upper-midrange DDR-333 was rated at CAS 2; similar-market DDR2-667 was rated at CAS 4 and today's middle DDR3-1333 is often rated at CAS 8. Most people would be shocked to learn that these vastly different rated timings result in the same actual response time, which is specifically 12 nanoseconds.

The problem perceived by many less-informed buyers is that faster memory responds more slowly, but it's obvious from these examples that this simply isn't often the case. The real problem isn't that response times are getting slower, but instead that they've failed to get quicker! When we see astronomical "speeds," we hope that our entire systems will become "more responsive" as a result. Yet, memory latencies are one place where things really haven't changed much. For example, with DDR400 and dual channel the theoretical bandwith would be 6400MB/s and with DDR2 800, the theoretical bandwith would reach 12800MB/s. But latencies can't be reduced, so the bandwith improves with DDR2 800 but only up tu 8GB/s more or less.

Conclusion

So I strongly recommend 2 identical Kingston modules of 1 GB each. If you needed more capacity, I would recommend 2 identical Kingston modules of 2 GB each. In the next post I will explain a bit the differences between DDR2 and DDR3 and whether it's worth choosing DDR3.

Read more »

PC - Microprocessor

The Microprocessor

Next you can find the main features of a microprocessor to be considered in order to choose the chip which matches better your needs.

Manufacturer

Currently there are two main manufacturers: Intel and AMD (Advanced Micro Devices). Intel processors were better until the appearance of the AMD Athlon in 2003. This AMD chip performs better than its competitor: Intel Pentium IV. But now Intel has recovered the leadership with his Intel Core 2 Duo. So there isn't a manufacturer absolutely better than the other one. For example, the better mainstream processor now is the Intel Core 2 Duo (two cores) or Intel Core 2 Quad (four cores).

AMD processors

But The Athlon 64 family had been superior to Intel's Pentium 4 or Pentium D offerings most of the time from the AMD64 launch in 2003 and Intel's introduction of the Core microarchitecture in 2006. While the Pentium 4 and Pentium D were aimed at reaching high clock speeds, their power requirements grew faster than the performance benefits.

As a consequence, AMD's approach of delivering more performance per clock was clearly more successful. Intel realized that a reasonable balance between processing cores and clock speed, based on the available manufacturing technologies, provides a much better path to more performance at reasonable power requirements. This eventually helped the company to get the crown back with Core 2 Duo.

Intel microprocessors

Technology

The main component of a microprocessor is the transistor. AMD phenom processor has more than 700 million of this semiconductor elements, most of them belonging to cache memory, so reducing the size of transistors the memory cache size can be increased. Nowadays Intel i7 processor technology is 45nm, where 45 nanometers is the length of a part of the transistor.

Architecture

When it's said that a processor has an architecture of 64 bits, then this processor can get 64 bits from the RAM memory simultaneously through the data bus. The more bits can get the processor simultaneously, the more fast can it work. Therefore a 64 bits processor with the suitable software can perform better than a 32 bits processor working at the same clock frequency. Another features concerning the architecture of a processor is whether it has memory cache and how much levels of this kind of memory it has,if it has also a memory controller included inside the processor and the number of cores it has.

Number of Cores

Until three years ago the microprocessors only had one core. The design and architecture of the chip was improved continuously (reducing voltage and transistors size and increasing memory cache size) so that the chip could work each time at a higher frequency and, this way, improve its performance.

But frequencies couldn't increase more without a big increase in power consumption and difficulties concerning power dissipation appeared (energy efficience). So manufacturers decided to decrease frequency of microprocessors and to use parallelism technology to keep improving perfomance of processors at lower frequencies. So manufacturers began to design chips with two o more processors inside, called cores.

AMD X2 3600 processor

Memory Cache. Layers

The microprocessor process the data obtained from the memory much faster than the time needed for the memory and the Bus to supply the next data. So the processor is often idle waiting for more data and its performance get worse.

For this reason a small but incredibly fast memory is put inside the microprocessor. This memory is called Cache and it normally contains the data the processor needs more often. This way the processor doesn't remain idle too much time. Even more than one level of memory cache are put in the processor to improve its performance.

The processors which don't have memory cache are also cheaper because the processor is easier to manufacture. For example, Intel sells Pentium processor with the tag Celeron. The processors called Celeron have much less memory cache, so they perform worse but are cheaper.

The main reason of the performance of the Core 2 Duo processor is its large and shared level 2 memory cache for both cores.

Frequency

The more high is the frequency, the more fast can work the processor. AMD processors always work at a lower frequency than Intel processors but some AMD chips can perform better because they can execute more instructions per clock cycle than Intel chips thanks to a more advanced architecture. Frequency is an important parameter to compare two processors provided they have a similar memory cache size and cores because a higher frequency can't compensate for a small memory cache. For example, a Pentium IV Celeron at 3.0 GHz performs worse than a Pentium IV 2.6 GHz.

Socket

In personal computers (PC), the microprocessor is inserted in a specific socket in a motherboard. The motherboard, through this socket, connects the microprocessor to the memory and to the peripherals. Through the peripherals, the processor can interact with the external world. Also in the motherboard some chips of RAM memory are inserted. The data between the RAM memory and the microprocessor is transferred by a bus, which is an electrical path between computer components.

Sockets Intel:

478 (Pentium IV and celeron)
LGA (Land Grid Array) 775 (from Pentium IV to Core 2 Duo)
Socket B (LGA 1366, incorporates the integrated memory controller and Intel QuickPath Interconnect)

Sockets AMD:

754 (Sempron i Athlon)
939 (Athlon 64 with dual channel)
AM2 (ATHLON 64 X2, supports DDR2 RAM memory)
AM3 (will support DDR3 RAM memory)

Read more »

PC - Introduction

With this post I start a serial with the main aim of explaining how a Personal Computer (PC) works and which are the most important features which have to be considered to buy a suitable PC for our needs.


Introduction

A computer is composed of several physical components, the hardware. The most important ones are the microprocessor, the memory and the motherboard (which controls the in/out devices and the memory and links them to the processor). The in/out devices are called peripherals. Examples of peripherals are the hard disk, the DVD unit, the graphic card, the sound card, the ethernet card, the modem and so on.

Elements of a current Personal Computer: Case, Screen, Keyboard and mouse

The microprocessor is the brain of the system and executes the instructions loaded in the memory, the software. This memory, called RAM (Random Access Memory), is really fast but can't store any information if the computer is off. On the other side, a hard disk is a massive memory (much more than the RAM but much slower) which can store the information indefinitely.

Computers only understand binary information. This language only has two symbols: o and 1, called bits. 0 means no electrical signal at all. 1 means some electrical signal. With eight bits we create a "word", called a byte. Each word represents a letter, a symbol or a number. So, when we type an 'a' with the keyboard, eight bits are generated, stored in the memory and sent to the processor. The amount of information a memory can store is precisely measured in bytes or Megabytes (a megabyte is a million of bytes).

There is a kind of software quite important: The Operating System (OS). The OS is loaded in the RAM memory when the computer is turned on. The OS has to manage the peripherals and tells the processor what to do every moment and where the data it need is.

The data needed by the processor is introduced through an IN device, like a keyboard, a mouse, a hard disk, a DVD unit. The results are shown in an out device, like a screen, a speaker...

The microprocessor, the RAM memory and the peripherals are all connected to the motherboard, and all of them are linked each other through a BUS. Through the data bus the information travels from one device to another one. The address bus is used by the processor to tell the memory where the next instruction or data it needs is stored.

An exploded view of a modern personal computer and peripherals:

1. Scanner, 2. CPU (Microprocessor), 3. Primary storage (RAM), 4. Expansion cards (graphics cards, etc), 5. Power supply, 6. Optical disc drive, 7. Secondary storage (Hard disk), 8. Motherboard, 9. Speakers, 10. Monitor, 11. System software, 12. Application software, 13. Keyboard, 14. Mouse, 15. External hard disk, 16. Printer

The components of a computer work syncronized by a clock. A clock tells each component when it has to send some information, when it has to get some information, when it has to process some information, when it has to show a result and so on. In short, the clock let the components talk and understand each other. A clock generates a lot of electrical impulses per second, and always the same number of impulses, so it has a frequency. The frequency is measured in Herz (Hz) or Megaherz (MHz). For example, if a clock generates a million of electrical impulses per second, it has a frequency of 1 MHz. The higher is the frequency, the faster the computer can work.

In the next post I will try to explain deeper the most important parts of a microprocessor and the features we have to consider to buy the processor more suitable for us.

Read more »