Getting to the 'Core' of Processors

With advancements in technologies, microprocessors have moved on from days when a new processor version only meant an increase in clock speeds to now when they provide better multi-tasking performance

Friday, October 03, 2008

Print

Comment

Email

Digg

Del.icio.us

Reddit

Twitter

The advancements in technologies such as very large scale integration (VLSI), ultra large scale integration (ULSI) and wafer scale integration (WSI) have made it possible to integrate millions of transistors on a single chip with great precision. The microprocessor as we know is an integrated circuit on which all the components of a computer are present on a single chip. These are capable of controlling devices ranging from microwave ovens, computers, cell phones to robotic arms of space ships and have become ubiquitous in the fields of computing, communications, manufacturing, etc. Allowing more circuitry to be packed on each chip, processors have consistently been migrating to smaller feature sizes, resulting in becoming more efficient in terms of both performance and energy consumption.

Server class processors
Intel and AMD both have been battling it out in the enterprise consumer market of servers, blade servers and workstations with their Xeon and Opteron brand of dual processor and multi processor configurations. Designed for better performance than their desktop counterparts, these processors have more cache with better multiprocessing capabilities and have been maintained over several generations of x86, x86-64 bit processors.

In 2005, Intel and AMD released their first dual core Xeon and Opteron processors respectively, and since then both consumer and server markets have constantly seen rise in popularity of multi-core processors, effectively increasing the processing power on each socket two to four times, and as the cost of motherboards increase dramatically with the increase in number of sockets, building a more powerful machine has become a little more economical with multi-core processors.

AMD's Opteron
First released in April 2003 with the sledgehammer core, Opteron is the server line of x86-64 processors from AMD and were initially made on 135 nm process. Opterons had gained popularity when it comes to multiprocessor machines, where the CPUs communicated using the Direct Connect architecture over HyperTransport, a bidirectional serial/parallel high bandwidth, low latency point to point link. Transparent to the programmer, each CPU can access the main memory of another processor. The Direct Connect is the I/O architecture which is also used in AMD's Athlon X2 and Phenom processors.

In May 2005 AMD introduced their first multi-core Opteron CPUs having two separate processor cores on each die. With multicore Opterons one socket could delver the performance of two processors and two sockets could deliver of four.

The second generation Opteron processors had three categories, the 1000 series (1processor/2 cores), 2000 series (2 processors/4 cores) and the 8000 series (4 processors/8cores). Based on code named Barcelona core design, the third generation Quad core Opterons based on a 65 nm process were launched in 2007 AMD is the latest in the line server class processors. AMD claimed that the new K10 architecture based processors had better power and thermal management and incorporated a number of improvements, mainly in memory prefetching, speculative loads, SIMD execution and branch prediction yielding performance improvements over K8 architecture Opterons. Featuring their Powernow and Coolcore technologies, AMD claims that the Quad core Opterons are the most power efficient server processors they have ever produced. These processors also supported the NX bit and AMDs virtualization technology AMD-V.

Intel's 45 nm quad-core MCM: Two dual core dies are placed next to each other to form a four core configuration.

Future
A core codenamed Shanghai should be seen in the future Opteron processors which will be based on a 45nm process. Also rumours are that AMD is working on their 6 core Sao Paolo and12 core Magny Cours processors which will be based on the MCM technique. Further, the server line of processors will incorporate a code named Buldozer processor with 4 or more cores on a 32 nm process, each supporting SSE5 aimed at better HPC and cryptographic computations.

Intel's Xeon
The Xeon branded 5200 series codenamed Woodcrest, based on the Intel's core-micro architecture was the server and workstation version of the Intel Core 2 processor. The fastest processor in this category operated at 3.0 GHz, claiming better performance and also less energy consumption than previous processors. In Jan 2007, Intel launched its quad-core, core 2 quad, as the 3200 series which comprised of two separate dual core dies placed next to each other in one CPU package and was targeted for the blades. The 3300 series was similar to 3200 but was manufactured using 45 nm process and featured XD bit and virtualization technology.

True to Intel's tick-tock release cycle of processors where a tick means a refresh of the current architecture and tock means a brand new architecture, the clock ticked and the Harpertown Xeons were released in late 2007. This family of processors consisted of dual die quad core processors manufactured on a 45 nm process and featured 1333 to 1600 MHz front side bus with lesser TDPs rated between 50W to 150W depending upon the model.

With Intel becoming an undisputed leader in the quad core arena, the last in this segment was the 7300 series code named Tigerton, consisted of two dual core Core2 architecture silicon chips on a single ceramic module. Claiming greater processing capabilities, the Tigerton was based on the Intel's Caneland (Clarksboro) platform.

Intel's 45 nm dual core die

Intel's latest offering
This month Intel has launched its new server class processor codenamed Dunnington. Dunnington is a 6 core (yes you got it right, 6 cores) processor and a first of its kind. It means that on a four socket motherboard installing four Dunnington processors would give you a staggering 24 cores. It features a single-die six core design and is based on Intel's 45 nm Penryn process. It is the first Xeon processor to feature 16 MB of L3 cache on board which helps improve performance by reducing latency in accessing frequently used data.

Desktop processors
A couple of years back, dual or multi-core processors changed the way desktop processors would function. Intel and AMD both have offerings in this segment under their Core2 and Athlon X2 brands respectively. Introduced in mid 2006, the brand Core2 refers to Intel's consumer range of 64-bit dual core and quad core microprocessors based on the x86-64 instruction set. It is based on Intel's core micro architecture and is an advanced version of the dual core Yonah microprocessor. The brand Core2 consists of Duo (double core), Quad (four cores) and Extreme (dual and quad cores for serious enthusiasts), on the other hand AMD has desktop class processors called Phenom is based on the K10 architecture and consists of triple core versions (Toliman) belonging to the Phenom 8000 series and the Quad core processors (Agena) that fall in the Phenom 9000 series.

A quad core processor is similar to the dual core with the basic difference that it has four cores instead of two. Intel's multi-chip module (MCM) or the quad core version had two separate dual core dies placed next to each other to provide a quad core configuration; on the other hand AMD claims their Phenom X4 series to be 'true' quad cores as they have all four cores on a single piece of silicon wafer unlike Intel. Though the clock speeds of these processors are reduced, the main advantage of core architecture lies in more efficient decoding stages, execution units, increased cache memory and reduced power consumption.

The core 2 branded processors from Intel include dual core processors for desktops named Conroe, Allendale and Wolfdale, whereas the Quad core for desktops computers consists of Kentsfield and Yorkfield processors. Most of these processors support features such as Virtualization Technology, ED bit and SSE3. The Core2 Extreme has Conroe XE (dual core) and Kentsfield XE and Yorkfield XE (Quad core) processors. The Yorkfield XE is also the first desktop processor from Intel with 45 nm technology and high k metal gates and features SSSE 4.1 instruction set.

Intel's future processors
Nehalem is the codename given to Intel's future micro architecture which is being developed under the Core i7 brand name, comprising a new series of quad core processors. Due for release in late 2008, processors based on this new micro architecture shall target high-end desktop and mobile platforms. They are the successor to the Core architecture and will be based on the 45nm process. It appears from Intel's spring IDF 2008 that Nehalem is going to incorporate one of the most significant changes to the current microprocessor architecture. Intel claims that Nehalem represents the next step in processor energy efficiency, performance and dynamic scalability. Dynamic scalability in a processor would mean managed cores, threads, interfaces, cache for energy efficient power on demand. These processors shall also feature Intel's Quick path technology which allows the concept of scalable shared memory between different cores with integrated memory controllers. Intel shall also produce processors based on Nehalem on a 32 nm process later.

AMD's future processors
AMD is expected to launch its new line of 45 nm processors code named Deneb FX for Phenom FX, Deneb for Quad core Phenoms, Heka for triple core and Regor for Athlon X2 with DDR3 compatibility and probably larger shared L3 cache by the end of 2008.

Fusion is a future next gen microprocessor that is being developed by AMD and ATI. It is supposed to combine the general processor execution and 3D geometry processing and other functions that are performed by a GPU into as single package and is expected to be first seen in 2009.

'Atom'ic energy
With increase in popularity of ultra mobile personal computers, Intel has come up with a low power x86 and x86-64 family of microprocessors called Atom. Atom is Intel's smallest processor ever made, designed for a 45 nm CMOS process and is intended to target portable and low power applications. It's a low cost processor with lower clock speeds but powerful enough to run the new wave of mobile internet devices (MID) and simple entry level low cost laptops. With devices known as netbooks and nettops gaining popularity, Atom is an ideal option due to its low power consumption, which results in improved battery life. Though it's based on an entirely new architecture, Atom was developed to fulfill the targeted performance with low power consumption while maintaining compatibility with Intel's Core micro architecture instruction set. It also features multiple threads for better performance and increased system responsiveness. Both Indian and other manufacturers have been adopting the processor in their UMPC and entry level notebooks. These are competitively priced around 20k or even less depending on the configuration provided. Devices running on these processors are good enough to keep you connected while on the move and run your basic productivity apps and capabilities to run entertainment features.
Processors for notebooks
For notebooks, people are often found confused between platform and processor naming trends, because laptops are based and advertised on the platform instead of the processor alone, a little background check on them is necessary. The platform of a laptop comprises of the main board chipset, the processor and the wireless network interface tweaked to deliver optimum performance, broader wireless interoperability and consume minimum power at the same time. This provides sufficient battery backup while the processor can perform optimally.
There are a number of platforms from Intel including the Napa platform (2006) that comes with the Centrino logo; the Santa Rosa platform, introduced first in 2007 comes with a Centrino Pro logo. Centrino has been Intel's brand name for their notebook platform, and in August the company raised curtains over their fifth generation Centrino2 (Montevina). It claims lesser power consumption than previously available platforms. These new age platforms boast an Intel Core 2 Duo (code named Penryn) 45 nm processor with clock speeds ranging from 2.26 GHz to 3.06 GHz, the mobile chipset offers RAM support for DDR2-667, DDR2-800, DDR3-800, DDR3-1066, DDR3-1333 SODIMM and NAND flash memory caching branded as Intel turbo memory along with Gigabit Ethernet LAN controllers 82567LM and 82567LF. On the other hand AMD has failed to gain popularity due to less advertisements and marketing. The latest in platforms from AMD this year is Puma which features the dual core Turion Ultra or the mobile Sempron processor and graphics from ATI.

AMD's take on Atom
AMD is expected to launch a very simplified x86 processor codenamed Bobcat aiming at very low power processing with TDP between 1 and 10W. Bobcat is expected to make its debut in the UMPC, handheld devices and other small form factor consumer devices.

Page(s)   1