Month: June 2021

CPU Common Terms

Processing Cores

“Cores” are a basic part of any CPU, the Cores are what handle all the workload. Some of that workload is handed off to their little brothers called “Threads”, which we will talk about next. The main job of your cores is to provide both stability and speed to the processing of data on your PC. They come in many different configurations from both Intel and AMD, depending on processor series and model numbers they differ from the likes of a dual-core for light workloads, all the way up to the monsters for special use cases that are 32 and sometimes 64 cores. They all do the same job the same way at different speeds and use different configurations of cores and hardware to interface with the motherboard, Think of a processing core as a piece of silicon that reads 1’s and 0’s at an alarmingly fast rate, and relays that data out to your GPU, hard disk/SSD, RAM, Etc.

Threads

Threads were introduced in the early 2000s when Intel released the “Pentium 4 with Hyperthreading”, The premise of threads is very basic, it divides the load handed to the core in half and hands it off to 2 worker threads inside each core, therefore processing data faster. Threads are that second thing you’ll read on every CPU box you pick up at your local parts store, and they are normally double the count of cores on a CPU. Say you pick up a CPU that is 6 cores, more often than not it will also list as having 12 threads on the box as each core has 2 threads.

IMC

The IMC or “Integrated Memory Controller” controls the communication between the CPU and RAM mostly, but also talks to other parts of your PC, like your hard drives or SSDs to fetch data for the CPU to process, and to your GPU to offload that data from RAM to be rendered on your screen, it plays a key part in how your system operates, That’s why you hear so many people on enthusiast forums talking about avoiding high voltages to reduce chances of damaging the IMC, if it becomes damaged you lose function of some RAM slots or “channels” on your motherboard thus reducing your max memory capacity, permanently.

Chiplets

The chiplet design has become a more recent way to pack many cores together in one space, for example, AMD’s chiplets are up to 8 cores and 16 threads each, and multiple of them can be assembled onto one CPU “package” or “substrate”, providing larger core counts for ever-increasing workloads in a smaller amount of space. The density is what makes this so easy to scale up for manufacturers. Put simply, Chiplet designs allow you to pack in more cores onto a CPU.

Substrate

This is the thing you hold with your fingers when placing the CPU in its socket, and is also what all the connections from the CPU to the motherboard are part of, this magic piece of engineering allows all the cores and threads to talk to each other and the rest of the computer via the motherboard to make everything operate.

ALU

The ALU or “Arithmetic and Logic Unit” is what modern CPUs use to process mathematics efficiently, simply put any math you do on your PC runs through this unit, everything down to simply using the windows calculator function, this removes stress from your cores to do other operational procedures for your PC and also spits out results in a very timely manner.

Cache

All modern CPUs come equipped with a cache, commonly referred to as  L1, L2, or L3. L1 cache is a set amount linked to each core and cannot be accessed by other cores for use, this is great for multithreaded processes as each core will get assigned a job in order as it finishes its previous. L2 and L3 cache differ in two ways, L2 cache is usually shared between a “cluster” of cores, 4 cores at maximum normally and they can communicate with each other via this L2 cache. The L2 cache is much larger than the L1 cache but is usually half that of the L3 cache. L3 cache is shared by the entire CPU for incoming and outgoing data, it is the data flood gate if you will, constantly feeding each other level of cache all the work to do while your PC is on, keeping your cores and threads hard at work.

There are different cache types, L1, L2 & L3 and they differ in cache size and operational speed. L1 is the fastest of the lot but has the lowest cache so far as size in the region of 1MB. Next, you have L2 which has a larger cache in the region of 8MB which is more than L1 but it is slower. Lastly, you have an L3 which will have the largest amount of cache in the region of 64MB. While L3 Cache is far larger in size and is able to store far more data, it is also the slowest of the bunch.

FSB

The FSB or “front-side bus” controls the speed at which your RAM sends your CPU information, normally measured in MHZ most commonly. This controls the rate at which data is processed. The faster it gets loaded into your CPUs cache, the faster it gets processed.

Lanes

Lanes are used in CPU jargon all the time, they’re all over spec sheets and motherboard sheets, but what are they really? Put simply. A x1 PCI-E slot transfers data via 1 data path to the CPU, which is very slow and rarely used in modern times. x4 PCI-E slots have 4 data channels connected to the CPU providing 4 times the data bandwidth or transfer speed if you will. The same terms carry to x8 PCI-E slots and x16 PCI-E slots. On some cheaper motherboards, some of these lanes will be controlled by a “southbridge” which is a chip that communicates with a CPU at a slower rate than those of direct connections. It is not very noticeable until you get into situations that need those lanes to be direct, like GPUs or you sacrifice performance. Each CPU is only given so many lanes for these data interactions, so eventually, you will run out if you have too many devices installed. It is worth noting that any device that utilizes the PCI-e interface will utilize CPU lanes, this includes storage devices such as NVMe drives.

Clock Speed

Clock Speed determines how many instructions per clock cycle, clock cycles are measured by how much data a CPU can process in 1 second. say your CPU is 400MHZ, it would complete 400 million instructions per second. If your CPU is 3.2GHZ it would complete an amazing 3.2 BILLION instructions per second. This speed has climbed over time, as the world has become ever hungrier for faster and more powerful CPUs and the companies manufacturing them are forever pushing the speed limits higher.

Instruction Sets

Instruction sets are what make a modern processor tick, and do its work in different ways to optimize speed and efficiency, ill give the basics of what each one does here if I went into full detail this would be several pages in length.

MMX

MMX is intel’s first MAJOR instruction set change, arriving with Pentium 2 and still used today. MMX is used for unpacking data in order at a faster rate than normal I/O Speeds once it is in a CPUs cache, It is however limited to 32 bits of unpacking because it was designed as an x86 instruction set.

SSE2/SSE3/SSE4

These three instruction sets share a name and an exact replication of functions, other than bit size increases as newer processor architectures hit the market as x64 instead of x86 instructions, the basic premise of SSE since its birth was floating-point integer calculation, the number value getting a larger theoretical maximum at each iteration along the way, SSE4 topping out at 128 bits of floating-point activity at once, the former SSE2 being only 32 bits at release.

AVX/AVX2

AVX and AVX2 do floating-point vector calculations, the former being limited to 64 bits and the latter boasting a massive 256 bits of data crunching AVX was designed for the long data code lines of modern programs today, it is mostly showcased in video encoding and editing software.

AMD-V

AMD-V is well, AMD’s instruction set for virtualization, it is quite a bit more complex than most think. With it disabled in the BIOS a virtual machine will not run on your system as the hardware acceleration required will not be available.

Intel VT-x

Intel’s VT-x is the same thing as AMD’s counterpart above, again, with it disabled in the BIOS a virtual machine will not run on your system as the hardware acceleration required will not be available.

NOTE: THE LIST OF INSTRUCTION SETS LISTED IS NOT EXHAUSTIVE, ONLY THE MOST COMMON WERE EXPLAINED IN MINOR DETAIL.

CPU Building Blocks

Die Lithography

I know the above name sounds like the name of some cheap book, but let me explain there’s quite a bit to this. You see the NM arms race in the CPU world is partially driven by this, die lithography is the overall size of the CPU dies without the heat spreader also referred to as an IHS. Measured in millimeters Squared, for example, intel’s current die size for their six core is 180MM squared, whereas AMD’s current die size for 8 cores is 80.7mm squared, this is due to different lithography’s being used. Namely, Intel uses what is called “monolithic” die lithography, which means they pack everything they can into one chip all connected to each other usually causing the heat output to be focused on the central area of the heat spreader. AMD, however, uses a “chiplet” type lithography, separating sets of cores and the I/O die from each other into their own separate chips, and spreading them out upon the substrate, this means that generally the heat output to the heat spreader is more evenly distributed.

Die NM (Nanometre) and what it means

When you hear the technology news say “so and so the company is moving to a 7NM process” it’s based on an old measurement that has been used for a long time, the NM figure in chip manufacturing refers to the gap between transistors inside the CPU Die, for example back in the early 2000’s we had 65NM, which compared to today’s 7NM spacing is quite a big leap in closing the gap. The smaller the gap (essentially wasted space), the more space that is then created to house more transistors and ultimately allows CPU manufacturers to create more efficient/powerful CPUs. Competition is fierce between the likes of Intel and AMD as to who can get to the next smallest NM iteration and can be the difference as to who wins the CPU performance crown at any given time.

Main Players In The Industry

Intel

Intel has been around since the late 60s, and 1968 in fact. They originally made processors such as the 4004 for use in complex commercial calculators, the invention of the microprocessor was pioneered by Ted Hoff, one of the founding 8 men of Intel back then. This breakthrough didn’t hit until 1971, it took years of engineering to get that far, but intel soon stormed the world with their invention, it was in such demand they needed 3^rd party manufacturers to make enough to keep up with the demand. Which is ironically were their only competitor originally started off from.

AMD

AMD or Advanced Micro Devices started up in early May of 1969 as a 3^rd party manufacturer of Intel’s 4004 microprocessor. Intel also licensed them all of the schematics and specifications to produce the product, later on in 1971 AMD produced its very first processor of its own, the AM2501 Logic Processor. The rivalry between the two companies began then and is still in effect today.

Strengths And Weaknesses

There is the age-old question, “who is faster than who” when it comes to Intel and AMD, but I will point out what each is good at from a general standpoint. In the end, you, the end consumer has the choice to make depending on your use cases, The list below sums up both sides of the fence, short and brief.

Intel

As of this article’s writing, intel has the following strengths over AMD:

• Higher Boost Speed Clock
• Gaming FPS (higher than AMD in single-threaded games in most cases)
• Faster Single-Threaded IPC (Instructions Per Clock)

And the following Weaknesses, Compared to AMD:

• Increased operating temperatures
• Multithreaded workloads
• Power-hungry

AMD

As of this article’s writing, AMD has the following strengths over Intel:

• Multithreaded workloads
• Video Encoding
• Streaming
• Core Count

And The Following Weaknesses, Compared to Intel:

• Slower Boost Speed Clock
• Gaming FPS (lower than intel in single-threaded games in most cases)
• Slower Single-Threaded IPC by a very small margin (Instructions Per Clock)

Summary

This about sums up the basics of what makes up a CPU in 2022. As with most topics, you could really go in-depth. CPUs are no exception to this rule due to their sheer complexity. However, I hope this article gives you a basic understanding and somewhere to start.

As always, if you enjoy my writing, feel free to let me know.