Advertisment

Understanding Server CPUs

author-image
PCQ Bureau
New Update


Advertisment

The first chip (CPU) was developed in 1971. Intel developed the first computer chip, the Intel 4004, and the first mainstream chip, the Intel 8080. Before Intel developed these microprocessors, computers existed, but they would occupy an entire building, and could only do one task at a time. Today the processors are adept at multi-tasking. They have shrunk in size and can easily be held on one's palm.

One of the most exciting components in a computer is the CPU (Central Processing Unit), which is essentially the brain of a computer. It receives information in the form of data packets that need to be processed. It interprets and executes instructions and data contained in software programs as well as the commands of its user. The more generic term 'processor' is often used to refer to a CPU as well. The power and performance of its CPU are the greatest determinants of a computer's overall performance.

Advertisment

Vendors:

The CPU battlefield is dominated by two companies, Intel and AMD--and their processors are forever pushing the performance of each other.

Advertisment

Speeds:

CPU speeds are measured either in MHz (megahertz) or in Gigahertz, where Hertz is a unit of frequency (1/s). Logically, the more frequency a computer has, the quicker is it capable of processing tasks. That number represents how many times the internal clock of a CPU ticks in cycles per second. The clock inside a 2.5 GHz CPU ticks 2.5 billion times each second. However, it's also important to note that the number of cores, the better a processor shall run. A dual-core, 1.5 GHz processor will run better and faster than a 2.8 GHz single-core processor. In essence, you should buy a processor that has the most cores with the greatest individual core speed.

Manufacturing process:

Advertisment

Processors have gotten quicker and smaller as generations have progressed. These days the 32 nm manufacturing process is the standard. Intel and AMD both have processors which are based on the 32 nm manufacturing process .

Cores:

Modern PC processors are multicore. These cores within a CPU can work as independent entities or be used for parallel processing to increase overall efficiency of the system for the workload. Each core is as fully functional as the others. Each has its own cache, but can communicate with other CPU cores as needed. Cores can also share information in a cache socket.

Advertisment

Multithreading:

Processors these days are able to provide parallel processing support. For one physical core on a processor, we find 2 logical cores which work in parallel. What this does is that it increases the number of cores available to the workload which speeds up the entire process. Anyways, the CPUs per server have been increasing, and multithreading ads to the specs available to the admin which he could dedicate to a particular workload. Virtualized workloads can also make use of these multithreaded CPUs to provide a spec package to the admin.

Cache:

Advertisment

Fast access to memory is important for a PC processor, and that's what your cache is for. Waiting for data to come in from system memory can make your CPU inefficient. The memory hierarchy is a way to show where memory bottlenecks exist. The CPU registers hold the data actually being used in calculations and comprise the fastest memory. The next fastest are Level 1 and 2 caches. These are small amounts of fast memory built right into the CPU core. Level 1 (L1) cache is faster but smaller than Level 2 (L2) which is faster but smaller than Level 3 (L3). Even the L3 cache is much faster than your computer's main memory. If some data is not in the cache, the CPU must wait until that data is retrieved from memory. In the slowest case, the data isnâ??t even in main memory, but must be fetched from the hard drive. However, while waiting for that data to be retrieved from system memory, the CPU can turn to a different task to keep the process running.

Advertisment

Bandwidth:

CPUs are part of a larger platform. They need to communicate with the outside world. Data buses (circuits that carry data) communicate with chipsets built onto the motherboard. The chips route data to the outside world which communicate with discrete graphics cards and other expansion cards, and USB ports, which communicate with external storage, keyboards, mice, etc. The speed of this communication is called bandwidth. It's split into two types: memory and input/output (I/O). This block of circuits is called the memory controller.

Socket:

The interconnect slot between the CPU and the printed circuit board is known as the socket. Different generation processors and even processors within a generation can be connected to the motherboard using only a particular socket. These sockets can only support processors compatible with their own spec and a forcefeed does not work. This is another reason why one cannot simply replace a processor with that of a different generation. The socket compatibility has to be kept in mind.

Memory compatibility and support:

One cannot simply change a processor and keep the rest of the platform same across generations. The CPU needs to be compatible with the overall circuitry like the socket, the motherboard and eventually has to be compatible with the memory type (DDR/DDR2/DDR3) and frequency as well. This more or less means that a CPU cannot be considered a changeable entity. A change of generation in a CPU will only be facilitated by an overhaul of the entire system in most cases.

Click on the image to enlarge

Click on the image to enlarge

Advertisment