Celeron
From Wikipedia, the free encyclopedia
Celeron is a brand name given by Intel to a large number of different x86 microprocessor models that they produced and marketed as a budget/value CPU line. The Celeron family complements Intel's higher-performance (and more expensive) Pentium CPU family. Introduced in April 1998, the first Celeron was based on the Pentium II core. Later versions were based on the Pentium III, Pentium 4, and Pentium M. These processors are suitable for most applications, but their performance is somewhat limited when it comes to running intense applications, such as cutting edge games or graphical modeling programs, as compared to that of their high-end counterparts.
Contents |
[edit] Background
As a product concept, the Celeron was introduced in response to Intel's loss of the low-end market, in particular to Cyrix's 6x86 and AMD's K6, but also to other competitors such as the [IDT] Winchip. Intel's existing low-end product, the Pentium MMX, was no longer performance competitive. Although a faster Pentium MMX would have been a lower-risk strategy, the industry standard Socket 7 platform hosted a market of competitor CPUs which could be drop-in replacements for the Pentium MMX. Instead, Intel pursued a budget part that was pin-compatible with their high-end Pentium II product, using the Pentium-II's (Slot 1) interface. The Celeron was used in many low end machines and, in some ways, became the standard for non-gaming computers.
[edit] The first Celeron (Covington)
The first Celeron (codenamed Covington) was essentially a 266 MHz Pentium II manufactured without any secondary cache at all. Although clocked at 266 or 300 MHz (substantially higher than the old Pentium MMX), the cacheless Celerons were a good deal slower than the parts they were designed to replace. Substantial numbers were sold on first release, largely on the strength of the Intel name, but the Celeron quickly achieved a poor reputation both in the trade press and among computer professionals. Many people referred to them as Slugeron, Celery, Deceleron, or Smelleron.[citation needed] The initial market interest faded rapidly in the face of its poor performance and with sales at a very low level, Intel felt obliged to develop a substantially faster replacement as soon as possible. Nevertheless the first Celerons were quite popular among some overclockers, for their flexible overclockability and reasonable price.
[edit] Mendocino
Intel was well aware of the poor reputation of the original Celeron and determined not to make the same mistake twice, with the result that the new Mendocino core Celeron was a good performer from the outset. Indeed, most industry analysts regarded the first Mendocino-based Celerons as too successful—performance was sufficiently high to not only compete strongly with rival parts, but also to attract buyers away from Intel's high-profit flagship, the Pentium II.
The key to the new Celeron's performance was cache. Where the old model had no secondary cache at all, the new part included 128 KiB of L2 cache as part of the chip itself. Otherwise, it was identical. With a total of 19.2 million transistors (including cache) on a single chip, the Mendocino Celeron was difficult and expensive to manufacture, but Intel managed a flawless execution of an ambitious project.
The first Mendocino-core Celeron was clocked at a then-modest 300 MHz but was almost twice as fast as the old cacheless Celeron at the same clockspeed. To distinguish it from the old model, Intel called it the 300A. Although the other Mendocino Celerons (the 333 MHz part, for example) did not have an A appended, some people call all Mendocino processors Celeron-A regardless of speed.
The Mendocino Celeron was the first mass-market CPU to utilise on-chip L2 cache. On-chip cache is difficult to manufacture; especially L2 as more of it is needed to attain an adequate level of performance. A benefit of on-die cache is that it can be made to run much faster than individual off-chip cache chips. Contrast this with the other common cache arrangements at that time. Most CPUs used mainboard mounted or slot mounted secondary L2 cache, which was very easy to manufacture, cheap, and simple to enlarge to any desired size. Typical cache sizes were 512 KiB to 1 MiB, typical speeds 66 to 100 MHz. The Pentium II had a pair of moderately high-speed L2 cache chips mounted on a special-purpose board alongside the processor itself. This was expensive and imposed practical cache-size limits, but allowed it to be clocked faster. Typical size was 512 KiB, always running at 1/2 of the processor speed. The new Mendocino Celeron had only 128 KiB of cache, but ran it at full clock speed (typically 300 MHz).
Although the Mendocino Celeron cache was rather small, its high clock speed more than overcame that handicap, and the Mendocino Celeron was a success, particularly with the enthusiast market. Overclockers soon discovered that, given a high-end motherboard, the Celeron-A/300 could run reliably at 450 MHz. This was achieved by simply increasing the Front Side Bus (FSB) speed from the stock 66 MT/s to the 100 MT/s spec of the Pentium II. At this speed, the Mendocino Celeron rivalled the fastest x86 processors available.
However, this only came about due to Intel converting one of their manufacturing plants that was at the time producing Pentium 2 processors at 450 mhz to produce Celerons that ran at 300 mhz, in response to a new budget aimed processor from AMD, which despite returning a loss allowed Intel to regain their reputation in the lower cost market, making the Celeron 300A a crippled Pentium 2 rather than a proper Celeron.
Over time, newer Mendocino processors were released at 333, 366, 400, 433, 466, 500, and 533 MHz. They all ran on the 66 MT/s FSB, which was not a particular issue with the 300, but became a major limitation with the faster-clocked parts, and Mendocino Celerons from about 433 MHz upwards were adequate rather than good.
The Mendocino Celerons also introduced new packaging. When the Mendocinos debuted they came in both a Slot 1 and Socket 370 PPGA package. The Slot 1 form had been designed to accommodate the off-chip cache of the Pentium II and had mounting problems with motherboards. Because all Celerons are a single-chip design, however, there was no reason to retain the slot packaging for L2 cache storage, and Intel discontinued the Slot 1 variant: beginning with the 466 MHz part, only the PPGA Socket 370 form was offered. (Third-party manufacturers made motherboard slot-to-socket adaptors (nicknamed Slotkets) available for a few dollars, which allowed, for example, a Celeron 500 to be fitted to a Slot 1 motherboard.) One interesting note about the PPGA Socket 370 Mendocinos is that SMP (symmetric multiprocessing) mode was available, and there was at least one motherboard released (the ABIT BP6) which took advantage of this fact.
The Mendocino also came in a mobile variant, with speeds from 266, 300, 333, 366, 400, 433, and 466 MHz.
In Intel's "Family/Model/Stepping" scheme, Mendocino CPUs are family 6, model 6.
[edit] Coppermine-128
The next generation Celeron was the Coppermine-128 (sometimes known as the Celeron II). These were a derivative of Intel's latest high-end part, the "Coppermine" Pentium III and were released in March 2000. Like the Mendocino, the Celeron-128 used 128 KiB of on-chip L2 cache and was restricted to a 66 MT/s bus speed. It was identical to the Pentium III except for the smaller secondary cache and the much slower bus.
Although in theory the Celeron-128 had an updated core, the benefit of this was not noticeable. The Celeron was, by this time, the only mainstream CPU still using a 66 MT/s bus and running 66 MHz RAM, and was significantly slower than any major competitor. Sales gradually dwindled as the industry moved on and, although making a 100 MT/s bus version would have been a trivially easy task, Intel chose not to—probably because they had major production problems at that time and preferred to concentrate available resources on trying to ship sufficient volumes of the high-margin Pentium III.
All Coppermine-128s were produced in the same FCPGA Socket 370 format that most Coppermine Pentium IIIs used. These Celerons started at 533 MHz and continued through 566, 600, 633, 666, 700, 733, and 766 MHz. Because of the limitations of the 66 MT/s bus, there was very little performance difference between the higher-clocked models; so long as the major competition was the elderly AMD K6-2, this was acceptable. In July 2000, however, AMD released their Athlon-derived Duron—a budget CPU with a faster bus and larger caches. The Celeron-128 became almost as uncompetitive as the original 266 MHz part had been.
However some benefit on performance could be gained through overclocking, by operating the Coppermine-128 CPUs with a 100 MT/s FSB. Overclocking a 566 MHz Celeron at 850 MHz or a 600 MHz at 900 MHz gave the best results, since these systems usually proved stable without the requirement of additional cooling or overvolting. In particular a number of 600 MHz Celerons had been marketed which operated with a default core voltage of 1.7 V: when running on a 100 MHz FSB these processors proved to be even more stable if compared to previous batches of 600 MHz Celerons with core voltage of 1.5 V.
On January 3, 2001, Intel finally switched to a 100 MT/s bus, and the performance improvement was startling. Although the Celeron 800 (the first of the 100 MT/s bus parts) was still less powerful than the Duron, it was within a few percent and a perfectly viable option. All Celeron-128s from the 800 up used the 100 MT/s Front Side Bus. Various models were made at 800, 850, 900, 950, 1000, and 1100 MHz.
The Coppermine-128 was used well into 2002. Early on it was popular in entry level desktops such as the eMachines eTower series. Often the processor would be permanently soldered onto the motherboard rather than being installed in a ZIF socket. Although always near the bottom of the performance table, it sometimes found a particular niche in low-power applications.
In Intel's "Family/Model/Stepping" scheme, Coppermine CPUs are family 6, model 8.
[edit] Tualatin Celerons
The next series of Celeron processors was based on Pentium III Tualatin core, and made with a 0.13 micrometer process. They were nicknamed Tualeron — a portmanteau of the words Tualatin and Celeron. Some software and users refer to the chips as "Celeron-S", referring to the chip's lineage with the Pentium III-S, but this is not an official designation. The series began with 1000 and 1100 MHz parts (which were given the extension "A" to their name to differentiate them from the Coppermine-128 of the same speed they replaced) and the line continued with 1200, 1300, and 1400 MHz chips.
Tualerons were nearly identical to their Pentium III sibling. The most significant difference was that the Celeron used a 100 MT/s bus rather than the 133 MT/s bus of the Tualatin-core Pentium III processors. The Tualeron also had a slightly slower L2 cache than the Pentium III, but not enough to seriously affect performance. They were excellent overclockers, since they had higher multipliers and users could put them on a 133 MT/s bus easily. The Celerons came with 256 KiB cache, while the Tualatin-core Pentium III processors came with either 256 KiB or 512 KiB cache.
The Tualatin-based Celeron was the last of the P6-core Celerons. For some time they were manufactured and sold in parallel with their replacement, the Pentium 4-based Celerons. Most users assumed that the newer parts, with their much higher clockspeeds, were significantly faster: they were not, and canny buyers snapped up the last of the P6 Tualatin Celerons, especially in notebook systems where the much lower power consumption translated into longer battery life.
In Intel's "Family/Model/Stepping" scheme, Tualatin CPUs are family 6, model 11.
[edit] Pentium M Celerons
[edit] Banias-512
This Celeron (sold under the Celeron M brand) is based on the Pentium M, and differs from its parent in that it has half the L2 cache, and does not support the clock-varying SpeedStep technology. It performs reasonably well compared to the Pentium M, but battery life is noticeably shorter on a Celeron M–based notebook than it is on a comparable Pentium M notebook.
A system based on the Celeron M processor may not use the Centrino brand name, regardless of what chipset and Wi-Fi components are used.
[edit] Dothan-1024
A 90 nm Celeron M with half of the L2 cache of the 90 nm (Dothan) Pentium Ms (twice the L2 cache of the 130nm Celeron Ms, though), and, like its predecessor, lacking SpeedStep.
[edit] Shelton (aka Banias-0)
The Shelton core is a Banias core without any L2 cache, and without SpeedStep. It is used in Intel's small form factor D845GVSH motherboard, intended for Asian and South American markets. The processor identifies itself as a "Intel Celeron 1.0B GHz", to differentiate it from the previous Coppermine-128 and "Tualeron" 1.0 GHz processors
[edit] Celeron Yonah-1024
The Celeron Yonah is a 65 nm Celeron M based on the Core Solo version (single-core) of the Yonah Core. Like its predecessors in the Celeron M series, this Celeron M has half of the L2 cache (1 MiB) of Core Solo and lacks SpeedStep. This core also brings new features to Celeron M including a faster front side bus (533 MT/s), SSE3 instructions, and support for the XD bit. This, like the Celeron D Cedar Mill-512, is probably the last Intel budget processor to use the "Celeron" name.
[edit] NetBurst-class Celerons
[edit] Willamette-128
The next series of Celerons was based on the Pentium 4 "Willamette" core and were, in consequence, a completely different design. These are often known as the Celeron 4. They have 128 KiB rather than 256 KiB or 512 KiB of L2 cache, but are otherwise very similar. Although the P4-based Celerons suffer considerably from their smaller caches, some speed grades have been favored in the enthusiast market, because like the old 300A, they can run well above their rated speeds. Some overclockers say that the Celeron is the best chip for overclocking because of its flexibility and lower heat production due to its smaller cache size.
In Intel's "Family/Model/Stepping" scheme, Willamette CPUs are family 15, model 1.
[edit] Northwood-128
These Celerons are based on the "Northwood" core, and also have 128 KiB of L2 cache. The only difference between the Northwood-128 and the Willamette-128 Celeron is the fact that it was built on the new 0.13 micrometre process which shrunk the die size, increased the transistor count, and lowered the core voltage from 1.7 V on the Willamette-128 to 1.52 V for the Northwood-128. Despite these differences, they are functionally the same as the Willamette-128 Celeron, and perform largely the same clock-for-clock.
In Intel's "Family/Model/Stepping" scheme, Northwood CPUs are family 15, model 2.
[edit] Northwood-256
These are the Mobile Celeron range used in Notebook Computers. Also based on the "Northwood" core, they feature a 256 KiB L2 Cache.
[edit] Celeron D (Prescott-256)
The Celeron D processor is based on the Prescott core and has a larger (than the previous NetBurst Celerons) 256 KiB cache. It also features a 533 MT/s bus and SSE3, and a 3xx model number (compared to 5xx for Pentium 4s and 7xx for Pentium Ms); specifically, they have been released thus far bearing model numbers of 355 (3.33 GHz), 350 (3.2 GHz), 345 (3.06 GHz) 340 (2.93 GHz), 335 (2.80 GHz), 330 (2.66 GHz), 325 (2.53 GHz), 320 (2.40 GHz), 315 (2.26 GHz), and 310 (2.13 GHz). They also have hardware-level support of Intel's EM64T technology by virtue of it also being built into the Prescott core, although the feature is disabled in all 3x0/3x5 models (with the exception of the Celeron D 355). It has been activated in all 3x1/3x6 models. The Intel Celeron D processor works with the Intel 845 and 865 chipset families. It should be noted that the "D" suffix actually stands for nothing. This name was given out by intel in order to differentiate this generation of Celeron from its predecessors.[1] It should also be stated that despite its name, the Celeron D is not a dual core processor.
In mid-2005, Intel refreshed the Celeron D with EM64T and XD Bit (eXecute Disable) enabled. Model numbers increase by 1 over the previous generation (e.g. 330 became 331), and were only manufactured for LGA 775.
[edit] Celeron D (Cedar Mill-512)
Based on the Pentium 4's Cedar Mill core, this version of the Celeron continues the 3xx naming scheme with the Celeron D 347 (3.06 GHz), 352 (3.2 GHz), 356 (3.33 GHz), and 360 (3.46 GHz) and retains the Prescott-V's feature set, except with double the L2 cache (512KB) and based on a 65nm manufacturing process.
[edit] Intel Core microarchitecture Celerons
These processors are not yet released, but are on Intel's roadmap. There is little current information about them, except that they will use the Conroe-L core,[2] and will use a 400 model number sequence. They will be single-core processors for the value segment of the market, much like the Core 2-based Pentium.
[edit] See also
[edit] References
- ^ Aubrey, John. Celeron D: the Little Processor that Could. Dev Hardware. Retrieved on 2006-09-23.
- ^ Huynh, Anh. Intel "Conroe-L" Details Unveiled. Daily Tech. Retrieved on 2006-09-23.
[edit] External links
- Budget CPU Shootout - Popular hardware review website Anandtech compares Celerons to other similarly priced CPUs
- Intel Pentium II Celeron Covington and Mendocino specifications
- Intel Pentium III Celeron Coppermine and Tualatin specifications
- Intel Pentium4 Celeron Willamette, Northwood and Prescott specifications
- Intel's Celeron M homepage
- Intel Celeron M Banias, Dothan, and Yonah specifications
- Intel Pentium II and Pentium III based Celerons at cpu-collection.de
|
Intel processors 4004 | 4040 | 8008 | 8080 | 8085 | 8086 | 8088 | iAPX 432 | 80186 | 80188 | 80286 | 80386 | 80486 | i860 | i960 | Pentium | Pentium Pro | Pentium II | Celeron | Pentium III | XScale | Pentium 4 | Pentium M | Pentium D | Pentium Extreme Edition | Xeon | Core | Core 2 | Itanium | Itanium 2 (italics indicate non-x86 processors) |
