Intel Processor Terminology

Intel has four "major processor families":

All Intel processors support the original x86 instruction set externally, but have very different micro-architectures internally. Some of the micro-architectures have evolved over time and are not the same as the original versions, but the family names have stuck.

Typical Product Numbering Scheme.  For example, lets look at an Intel Core i7-4700HQ chip.

The above example is for the Core i7 product family, the top performance group of the Core products. "Core" chips come in three main versions i3, i5, and i7. This is the fourth generation of "Core" chips. The generation is the node (size of the transistor technology width) and micro-architecture designation. In the above case, it is the 22 nanometer (nm) production node with a fourth generation micro-architecture code named Haswell. Each generation of micro-architecture and node is given a code name. The exact model is the 700. The higher the number, the faster the processor. The HQ suffix designates the processor as a quad-core, meaning the chip contains 4 micro-processors. A "core" is one micro-processor. This exact numbering scheme is not used for other product families, but the Core family is the most popular one for business and residential customers. Other product family numbering schemes are similar but not exactly the same.

Note: all "road maps" shown below are Intel generated.  Top

The "Xeon Phi" Processors

Intel Client Roadmap

In 2012, Intel announced that Xeon Phi would be the brand name used for products based on their Many Integrated Core (MIC) architecture. Xeon Phi chips are a family of chips delivering extremely high performance for workloads that are executed in parallel fashion. The first Xeon Phi product, code named Knights Corner, was announced in 2011, manufactured on a 22 nm (nanometer) process node (die shrink), delivered in 2012, and was in operational products in 2013. In Knights Corner systems, a Xeon processor (Ivy Bridge) was used as the main engine and the Xeon Phi chip, Knights Corner, was used as a co-processor. In the subsequent architecture, Knights Landing, the Xeon Phi chip can be used either as the main processor or as a co-processor. Knights Corner chips have in excess of 60 cores that can all be running at the same time. The Xeon Phi 7120 chip is capable of producing over 1.0 teraflops (TFLOPS - trillion floating point operations per second) of double precision floating point with 352 gigabytes per second (GB/sec) memory bandwidth.

Knights Landing Die

By 2013 Xeon processors were ubiquitous in supercomputers - more than 80% of the top 500 machines in 2013 used them. For the very fastest machines, much of the performance comes from computer accelerators, i.e. the Xeon Phi. The first computers using Xeon Phi processors appeared in June 2012 and by June 2013 it was used in the fastest computer in the world.

A second generation Xeon Phi product, code named Knights Landing, on a 14 nm process was announced in June 2013 and shipments were made in late 2015 to early customers. See the Knights Landing chip package to the left. Note that the die size is extremely large. The "master die" is 3.5 inches wide by 2.25 inches high - much bigger than any previous full sized die. Note the "near" memory, called MCDRAM (multi-channel DRAM), composed of 8 small memory die of 2 gigabytes (GB) each located around the processor, making 16 GB total on the master die.

Intel says Knights Landing comes in two variants. The first will be a version mounted on an add-in board. In this case, Knights Landing plugs into a traditional server and acts as a co-processor. However, the other variant will come as a stand-alone processor. Intel is really playing up the stand-alone version and claims that more than 60 system providers will be selling units with Knights Landing as a stand-alone processor by the end of 2016. HPC (high performance computing) made up a significant portion of the company's data-center revenues in 2015 and it is expected to be a very fast-growing segment through 2018. See the Knights Landing Overview chart below.

Knights Landing Overview

The cores on the Knights Landing chip are based on a heavily modified “Silvermont” Atom core that has four threads on it. (More on Atom chips below.) These are tiled in pairs, with each core having two AVX512 512-bit vector processing units (VPUs) and 1 MB of L2 memory shared across the tile. See the tile layout in the upper right hand corner of the chart. The tiles are linked to each other using a 2D mesh interconnect, which also hooks into the two DDR4 memory controllers that feed into the DDR4 "far memory". The far memory scales up to 384 GB capacity and delivers about 90 GB/sec of bandwidth on the STREAM Triad memory benchmark test. The 2D mesh also hooks the cores into eight chunks of high bandwidth stacked MCDRAM on-package memory, which scales to 16 GB of capacity. This is known as the “near memory” and offers more than 400 GB/sec of bandwidth to keep the cores well fed.

Intel has a number of different modes of memory addressing in the Knights Landing processor, including using the combined memory as a single address space or using the MCDRAM as an L3 cache for the DRAM memory. Add it all up, and Intel says that a Knights Landing processor delivers more than 6 teraflops of single precision and more than 3 teraflops of double precision floating point performance. Intel expects to ship more than 100,000 Xeon Phi units into the HPC market mainly in the second half of 2016.

See the four different versions of the Knights Landing chips below that Intel formally announced in June, 2016.

Xeon Phi Table

In the fourth quarter of 2016 Intel will ship the variants of these four Knights Landing parts with an on-package Omni-Path 100 series port. The Omni-Path Architecture (OPA) is a new high speed 100 Gb/sec (gigabits/sec. not gigabytes/sec.) Intel interconnect system to connect tens of thousands of nodes in an HPC environment. Adding the Omni-Path links to the chip package boosts the price by $278 and raises the thermal design point by another 15 watts. Two deals involving Omni-Path networking that have been publicly announced are at the Pittsburgh Supercomputing Center in the United States and Fudan University in China.  Top

The "Xeon" Processors

Xeon Haswell Processors - Version 3

Haswelll Zeon Processor

The Xeon family of processors are mainly aimed at the single socket server market. (However the newer E5 and E7 versions do support multiple sockets.) The overall main benefit of the Haswell micro-architecture is much lower power efficiency. The variants have 4 to 8 cores and support hyper threading.

A processor "thread" is a sequence of active instructions in one core. A single-core computer processing unit (cpu) can have only one thread active at a time. A 2-core cpu has two threads active, one per core. Hyper threading is two virtual processors each sharing one physical core. Hyper threading can be utilized only with an operating system specifically optimized for it.

The version 3 (Haswell) Xeon has a DRAM limit of 32 gigabytes which might appear rather small at first glance. However, Xeon processors are designed to a very specific thermal power limit and are energy efficient above everything else.

Performance per watt is the most important design factor in the server world. Cooling hundreds (maybe thousands) of servers is paramount in the "cloud" market.

There are three types of Xeon chips on the market, classed as the E3, E5, and E7 series:

Xeon Broadwell Processors - Version 4

Broadwell Zeon E5 v4

The newest Broadwell line (version 4) of very high performance processors are aimed towards the enterprise and workstation markets that are configured in single and two socket platforms. At the Intel Developer Forum in May 2015 (IDF15), an Intel spokes person told the group that the Broadwell lineup of Xeon E5-2600 v4 series processors (2 socket platform) was going to launch in the fourth quarter of 2015. This launch was subsequently delayed one quarter into Q1 of 2016. The Xeon E5-4600 v4 (4 socket platform) was announced in March, 2016. The Broadwell-EX/EN based Xeon E7-4800 v4 (4 socket platform) and Xeon E7-8800 (8 socket platform) were launched in June, 2016.

Broadwell-EP Xeon E5-2600 v4 processors feature up to 22 cores while the EX series will feature up to 24 cores. The 22 core version will feature a total of 44 threads and 48 for the EX. Intel has announced up to 55 MB of Level 3 cache on these processors. Maximum power ranges are from 145 watts to 165 watts.

The Broadwell Xeon E5-4600 v4 family is aimed at workloads that do not require the large memory footprint of the Xeon E7, which allows for up to 24 memory slots per socket compared to the 12 memory slots in the Xeon E5s. E5s are aimed at server virtualization and database acceleration, but not with the extreme memory capacity that is only available with the Xeon E7s. The Xeon family of processors is illustrated in the table below.


Intel Xeon Families (June 2016)


E3-1200 v5

E3-1500 v5
E3-1500M v5

E5-1600 v4
E5-2600 v4

E7-4800 v4

E7-8800 v4

 Core Family






 Core Count

2 to 4

2 to 4

4 to 22

8 to 16

4 to 24

 DRAM Channels







64 GB

64 GB

1536 GB

3072 GB


 Multi-Socket Support



2600: 1S or 2S

Up to 4S

Up to 8S

 PCIe Lanes






 Target Market

Entry Workstation

Memory Compute

High-End Workstation

Many-Core Server

World Domination


The "Core" Family Of Processors

Intel's "Core" brand of products are aimed at individual users ("Clients") and this product family makes up the bulk of computer processors on the market today. The current Core products come in five smaller groups: the i3,i5, i7, i7vPro, and the M series. Intel also offers Celeron and Pentium versions for entry level users at very good prices. Celeron versions are aimed at very budget oriented customers and have a very reduced version of the Core architecture. Pentiums are also based on a reduced Core architecture typically by lowering the clock frequency and disabling some features, such as hyper-threading, virtualization and some cache memory. (Current Pentium processors have only the name in common with the early Pentiums.) In general, the Core i3 is the everyday reasonably priced processor for the average user. It has good performance for the majority of people. The i5 is the medium processor for advanced scientific type users. The i7 is the top of the line processor for extremely heavy workloads (high-end gamers). The i7 vPro version has some features unique to advanced business users (heavy security and manageability). The M versions are the "mobile" equivalents of the desktop units and come in M3, M5, M7 and M7 vPro versions.

The chart below is an Intel originated chart showing the processor progression for mobile products from generation 6 (Skylake) to generation 7 (Kaby Lake) to generation 8 (Cannon Lake) to generation 9 (Coffee Lake) in early 2018. It is really hard for lay people to keep track of all the different processor versions in different market segments in shifting time frames. If all this sounds complicated, it is.

Intel Roadmap

Lets look at the abbreviations at the left hand side of the above illustration.

What makes the generations different?

Gordon Moore, a co-founder of Intel, discovered that every two years the number of transistors in processors could be doubled. The cost of the wafers increased some, but the number of transistors double making the cost per transistor decrease dramatically. This became know as Moore's Law and has been good for 50 years. See the Moore's Law Section. Moore's Law is now universally used in the semiconductor industry to set future targets in research and development.

Intel Technology Leadership

Intel and other chip suppliers now plan a new node (die shrink) every two years. See the technology chart to the left. A new node means that Intel can double the transistors on the same size die which then uses less power and generates less heat with an increase in performance. Between node advances every other year, a new more advanced micro-architecture is introduced. Hence there is a tick-tock rhythm to the semiconductor process. Every tick is a new node, every tock is a new advanced micro-architecture.

Ivy Bridge was a 22nm node ("tick") down from 32nm Sandy Bridge, while Haswell was a more advanced micro-architecture ("tock"). Broadwell was a new 14 nm node ("tick"), while Skylake is an advanced micro-architecture ("tock"). However, with Intel's announcement that Cannon Lake will be delayed about a year, Kaby Lake will be a second new micro-architecture from the same node. Intel now has a tick-tock-tock system in effect. The transfer from 14nm to 10nm is proving very difficult and we may be seeing 2.5 to 3.0 years between node upgrades in the future.  Top

Core i3

Even though the Core i3 is at the base of the Core family, it is still a very good processor that has received very good reviews for its cost and performance by the majority of experts and customers alike. The technology behind Core i3 processors includes dual cores, hyper threading support, and virtualization. Core i3 processors also support 64-bit versions of Windows.

The important suffixes for desktop Core processors are K, R, and T. The K models have an unlocked core multiplier that makes them easier for enthusiasts to overclock the master clock (make it run faster). The R suffix is new for 4th generation parts and is used to designate a processor with Iris Pro integrated graphics. The T models are very low power versions built to draw as little power as possible. Standard laptop processors use the M suffix. New low-voltage parts meant for Ultrabooks and Tablets replace the M with a U or Y.  Top

Core i5 and i7

Core i5 & i7 Families
Intel i7 Chip

Core i5 and i7 processors are the “mid-range” and "high-end" processors produced by Intel. The above chart compares 4th generation (Haswell) i5 and i7 processors. Unlocked means that the base clock speed can be adjusted by the customer.

Note the base clock speed of 4.0 for the i7 (top line) is close to the maximum clock speed (maximum in a production computer environment is about 4.4 GHz as shown in the Turbo Frequency column). Note however, that the base clock speed is only a gross approximation of a processor's true speed.

The "micro-processor" clock speed can be as much as 10 times the base clock rate. Also complex instruction sets take longer to execute than do simpler instruction sets. Sometimes they are slower than simpler instructions, but sometimes faster depending upon the nature of the work involved. Benchmark runoffs are the ultimate test of a computer's speed.

Core i5’s offer enough performance to do video editing and gaming, and more than enough performance to do basic things like word processing, internet surfing, and email. A Core i5 processor is a great mid-range priced processor for people who use their computers frequently and multi task.

Core i7’s are the top of the line of Core series processors sold into highly technical environments and extreme gamers. They are also the most expensive. i7's have 4 more threads that can be running simultaneously and two more megabytes of cache memory than i5's. For most computer users, an i7 processor is not necessary. But... if you want the latest and the fastest, that’s what the i7 gives you.  Top

Core M Processors

M SeriesPackage

As mentioned above, the M processor series are mobile versions of the i3. i5, i7 and the i7 vPro. When talking about Core mobile processors we are discussing laptops and 2 in 1 type tablets, but not very small tablets, smart phones, or watches. When processors run they generate heat - lots of heat. Therefore the heat generated by the processor must be tightly controlled because of the space and cooling limitations in a thin portable computer.

The 5th generation (Broadwell) Intel Core M processor package eats up just 495 square millimeters of space, about half the size of the 960 square-millimeter 4th generation package. See the illustration to the left. Note that the graphics processor and memory controller are integrated on the processor die.

In the i5 and i7 unlocked chart above for desktop processors, the heat limit is specified by the column TDP which stands for "Thermal Design Power", i.e. the amount of heat that the cooling system of the computer must dissipate. The i5's and i7's dissipate 84 to 88 watts of energy just from the processor chip. This is a fair amount of heat and heat sinks and fans are required to cool the processors. A similar TDP number for the Skylake (6th generation) mobile processors is 4.5 watts - a huge difference for very thin packages that operate without fans.

By in large, micro-architectures of the desktop and mobile processors of the same generation are similar. Programs written for a desktop will run unchanged on a laptop and vice versa. Portables do have a few different instructions that are unique to portables - such as powering down when the lid is closed. The main difference is in clock speeds. Laptop clock speeds are limited to 3 GHz or less. Cache sizes are limited to 4 megabytes versus up to 8 Mbs in desktops. Also, all Core M processors have 2 cores and 4 threads as opposed to 4 cores and up to 8 threads in desktops.

However, laptops provide very good overall performance, and in ordinary day to day workloads, perform as well as a desktop. Only special high performance workloads need to be done on a desktop these days.  Top

Intel Announces Kaby Lake (7th Core Generation)

&th Generation Kaby Lake

On August 30, 2016, Intel announced Kaby Lake - its 7th generation of Core processors. (Kaby rhymes with baby and is named after a lake in Canada.) As mentioned above, Intel has gone from a tick-tock series of architectures to a tick-tock-tock mode. Kaby Lake is the third set of 14 nm processors and the node (size of the transistor technology width) time has gone from 2 years to about 3 years between new node releases.

The move from Intel’s 14nm process to 10nm has been a long, slow one taking much longer than any previous new node introduction. The 14 nm process (Broadwell) was first introduced in Q3 of 2014. The 10 nm processors (Cannon Lake) are expected to be announced sometime in the second half of 2017.

Intel no longer refers to the processor rhythm as tick-tock. It is now referred to as Process-Architecture-Optimization, or PAO for short. The first sector (Process) is a new manufacturing node, the second sector is a revised computer architecture, and the third sector is an optimized version of the architecture, called "14nm Plus" (or 14nm+) by Intel.

As can be seen from the Intel chart to the left, the various processor versions will come out in waves. The first wave will consist of three Kaby Lake parts of about 4.5 watts of power aimed at high-end tablets and 2-in-1 computers. Also, three more Kaby Lake parts at 15 watts will be dedicated to powerful notebooks (6 Kaby Lake models in all).

Intel Core i7 Kaby Lake

The fundamental micro-architecture between Skylake and the new Kaby Lake parts is practically unchanged. However, the "improved fin profile" (in the finFET transistor technology) is expected to produce a 12% to 19% CPU performance increase by running at higher clock frequencies with no increase in power. In addition, Kaby Lake has some nice upgrades of Generation 9 graphics aimed at the upcoming era of 4K ultrahigh definition (UHD) TV screen support (current TVs are 1K in number of screen pixels).

On January 3, at CES 2017, Intel announced the remaining line of Kaby Lake processors. Intel made a comprehensive update to its entire product line. See the Kaby Lake Core i7 processor pictured to the left. The initial Kaby Lake "mobile" refresh was limited to a handful of processors. With this launch Intel is bringing out a complete line of Kaby Lake processors intended for every price point. Kaby Lake is priced nearly identical to Skylake in almost every case.

Apart from the 4K ultrahigh definition (UHD) TV screen support, clock speed increases, and the Core i3-7350K (the first unlocked Core i3), the Kaby Lake family refresh is a pretty standard update to Intel’s road map. A little more clock speed, a little more performance, but not many new features to get excited about. However, to Intel's credit, squeezing an extra 8 to 10% performance out of Skylake with equivalent power is no small accomplishment given how hard it has been to move the ball on x86 performance.  Top

Intel Will Release 8th Gen Coffee Lake Chips in Late 2017

8th Gen Coffee Lake

Intel's eighth-generation Core CPUs, code-named Coffee Lake, will launch in the second half of 2017 - far earlier than the 2018 launch previously predicted. Intel confirmed at its Investor Day on February 9, 2017 that its 8th-gen chips will also be based on its 14 nanometer (nm) process. Coffee Lake will pack up to six processor cores, rather than the maximum of four cores that the company has previously offered with such processors.

The early launch of Coffee Lake is something of an anomaly. In early 2016, Intel introduced a three phase rhythm of Process, Architecture, and Optimization (PAO). But now with Coffee Lake, it seems like Intel might be modifying that model as Coffee Lake is the fourth phase on 14nm technology.

Intel is promising a 15 percent (maybe more) jump in performance versus Kaby Lake. See the Intel chart to the above left. Intel hasn't explained whether the chart refers to desktop or mobile chips or both. Ashraf Eassa of The Fool believes that the reason that Intel is keeping its high performance notebook and desktop processors on its 14nm technology is cost. By late 2017/early 2018 (when Coffee Lake is expected to launch), Intel's 14nm technology will be quite mature while the company's 10nm technology will have just begun production and will be at the top of the learning curve.

Another reason for moving the Coffee Lake chips forward is because Intel will want to go head-to-head with AMD's Ryzen CPUs which launched in March, 2017. (Ryzen features eight cores and 16 threads in its top-end part, with six and quad core versions making up the rest of the line.) The Ryzen chips are expected to give Intel a run for the money in both performance and cost.

Intel Shows Off Cannon Lake 2-in-1 PC

Cannon Lake CPU

During Intel’s press conference at the CES 2017 convention in Las Vegas on January 3rd, 2017, CEO Brian Krzanich showcased a 2-in-1 PC sporting a processor based on the company’s “Cannon Lake” design. See the CES photo to the left. The chip was created using 10nm process technology, meaning Intel is able to cram even more transistors into a standard 12 inch wafer. Essentially, the lower the process number (10nm), the smaller the transistors are, creating more powerful, more power efficient chips.

For the last several years there has been a lot of talk about the death of Moore’s Law - that transistors will only get so small. However, Krzanich said that Moore’s Law is not dead and that Intel’s Cannon Lake processor design is a perfect example. “Moore’s Law is alive and well and flourishing,” he added.

Processors and products based on Intel’s Cannon Lake design are expected to ship before the end of 2017. Unfortunately, that is all Krzanich provided in regards to the embedded 10nm processor. However, he did indicate that 10nm process technology will help evolve the current virtual reality market thanks to slimmer devices, better power efficiency, and better computing performance.  Top

The "Atom" Family Of Processors

Atom x5 and x7

The "Atom" is the Intel brand name for its line of ultra-low-power processors. Atom processors are mainly used in netbooks, embedded applications ranging from health care to advanced robotics, and mobile internet devices (MIDs).

The Atom family of processors was first introduced in March of 2008. In December of 2012, Intel launched its first 64 bit versions of Atom processors. In March of 2015 Intel announced its x series of Atom processors which come in three versions: x3, x5, and x7 SoCs (systems-on-a-chip). The x5 and x7 (code named Cherry Trail) are 14nm technology. Intel has not formally announced the TDP for the x series, but it is believed to be under 4.0 watts.

The x3 chip (branded SoFIA) was aimed at computing devices priced under $75, while the x5 and x7 are aimed at products priced at $120 or more. The x3 also marks the first time Intel has been able to "integrate a modem" into a SoC, available in both 3G and LTE variants.

Intel Withdraws From Tablets And Phones

On April 26, 2016 Intel unexpectedly announced that it had made a major decision to drop the "mobile" versions of its Atom SoCs (Systems on a Chip), code named SoFIA and Broxton (smartphones and tablets), as part of their new revised strategy. Here is a summary paragraph of the strategic announcement of what is "in" at Intel by President Krzanich:

"Intel is accelerating its transformation from a PC company to one that powers the cloud and billions of smart, connected computing devices. We will intensify our investments to fuel the virtuous cycle of growth in the data center, IoT, memory and FPGA businesses, and to drive more profitable mobile and PC businesses. Intel delivers a broad range of computing and connectivity technologies that are foundational to this strategy and that position us well to lead the end-to-end transition to 5G. Our connectivity strategy includes increased investment in wired and wireless communications technology for connecting all things, devices and people to the cloud, and to power the communications infrastructure behind it. We re-evaluated projects to better align to this strategy."

What was "not included" in the strategic direction were "smartphones and tablets". Having fared poorly in mobile smartphones and tablets, Intel has abandoned its efforts to play catchup. It confirmed to trade publications that it was cancelling its Broxton and SoFIA chips designed for 3G and 4G phone technologies. Intel sent out a second message to trade publications clarifying that the Broxton Platform was being canceled for both smartphones and tablets, as tablets were not mentioned in the original message. Intel had invested billions of dollars subsidizing chip sales to tablet and low cost phone suppliers at huge profit losses to the company. However, Intel hasn't given up completely on smartphones as they have included the forthcoming 5G phone technologies in their strategic direction.

Also, Intel’s newest generation 14nm Atom platform, Goldmont, was announced at IDF Shenzhen in China in April of 2016 as part of the Apollo Lake netbook and low-cost PC platforms. So, it appears that Intel is not completely abandoning the Atom product family, just the current mobile smartphone and tablet "markets" that are not profitable for them. One other item not mentioned in the strategic direction is the future of the Intel contract semiconductor manufacturing business. It is currently not clear whether or not Intel is going to actively pursue its contract fab business.