• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

CPU guide for newbies

Associate
Joined
5 May 2016
Posts
67
Hi people, I just wanted to open this thread to see if any kind souls round here could compile their vast knowledge, to let newbies like myself understand all the terms you guys are using regarding cpus, basically a starting point for some of the terms and how everything relates together to meet each users needs. I surely have felt overwhelmed, probably because I dont keep up to date. Still, i believe it would be good to have some sort of starting point in order to make it more approachable.

Cheers:p
 
To simplify matters I have only referred to Intel trade names.. AMD have their own set of marketing terms for CPU features as do other manufactures

CPU = Central Processing Unit - the part of a computer that typically does most of the computational work of a computer. One of the most important parts of a computer.

Cores = For most of the history of CPU's up until the mid 2000's most CPU's consisted of one 'core'. However due to the physics of the materials used to make CPU's it was found that it was not possible to make one core CPU's run faster indefinitely. Also by adding more cores to a CPU different tasks could be run in parallel therefore improving multitasking (where a computer is doing more than one task at a time) and by makes some games and applications run faster if they were suitably coded to take advantage of more cores.

I.e. a first person shooting game where one core was used to run the A.I (artificial intelligence) of the computer players where other cores were used to run the game itself.

The downsides of adding more cores are

1) It makes the die's bigger for the same core design and fabrication method (see nm fabrication process below) and therefore more expensive.

2) A CPU with more of the same type of cores compared to one with less of the same cores will generally not be capable of running as fast.

3) There is a limit to how many cores games and applications can utilise even if they are well coded. Only so much work game be completed in parallel

Threads/ Hyper threading(HT - Intel Only) = Some CPU's are capable of using the under utilised parts of their cores to run more than one set of instructions at a time. Each set of work being done in this fashion is called a 'thread' and Intel calls this technology 'Hyper threading' It allows CPU's to run suitable software more quickly (its similar to adding mores cores to a CPU but an additional core will be superior to an additional thread running on a core).

The combination of cores and threads a CPU has is often abbreviated as Xc/Yt (where c = the number of cores and t = the total number of threads those CPU's can run) i.e. a 6700K is an 4c8t CPU

Die = The ''functional' part of a CPU that contains the circuitry that makes the CPU work. Often used when talking about 'die size' i.e how big (or small) the functional part of a chip is

Wafer = a piece of electronic grade silicon (very high purity) onto which lots of identical Die's are copies using a processes like photolithograph. Wafer's come in industry standard sizes and so the size of the die's on a wafer affects the total amount of die's that can be produced per wafer

nm = Nano meter - the current size unit used to measure the size of the circuitry on a die. Often referred to as a 'X' nm fabrication process.

i.e Skylake is a range of CPU's made on a 14nm process.

The smaller the size of the process used the smaller a CPU die will be for the same design (Layout).

Yield = Not all the die's on a wafer are the same quality... Some will be faulty and will have to be discarded and some die's may only be party functional or not be capable of running as fast as other Die's using the same voltage. The 'Yield' is used to discuss the amount of functional die's produced from a wafer and the speeds at which the die's can be used at un der a given voltage range.

i.e. Skylake wafers had a poor yield initially

Clock Speed/frequency = the speed at which a CPU runs - normally measured in GHz where 1 GHZ = 1000MHz

1 MHz represents 1 million cycles per second!

i.e - a Skylake 6700K CPU has a base frequency of 4 GHz and a turbo frequency of 4.2GHz

Base/Turbo speed/Frequency (Intel only) = Intel's more recent CPU's don't run constantly at the same speed. They can 'de clock' (lower their frequency) when not doing much. This is normally done to save power used. Conversely they can run faster when requires and this is sometimes referred to as a 'turbo' speed. On CPU's with multiple cores sometimes the turbo speed is determined by how many cores are being used with less cores equating to a higher turbo speed. This allows for software that does not utilise many cores to be run faster whilst still allowing the CPU to run other software that does benefit from more cores (albeit with the individual cores running at a lower speed)

Stock and Overclocked Speed/Frequency = The stock speed is the speed a CPU is rated to run at by the manufacturer by changing the settings on the computer (normally in the BIOS) and by sometimes increasing the voltages used by the CPU a CPU can be run faster than 'stock' speeds and is then said to have been 'overclocked'. This allows the CPU to run instructions more quickly but uses more power, generates more heat and to a degree can shorten the life of the CPU due to gradual electromigration over time or more sudden failure due to a failure caused by over volting/ overheating of the CPU.

Often abbreviated to 'OC' i.e my CPU is oc'd to 4.5GHz or my CPU can run a 5GHz OC at 1.35volts

Cache (memory) = Modern CPU's have memory built into their die's for short quick storage of information to speed up processes. Intel refers to this memory as L1, L2, L3 and for some CPU's L4 cache in order of decreasing cache speed and increasing cache size for a given set of CPU's. The L1 and L2 cache's are very small - for example the Skylake set of CPU's has: -

L1 cache - 64 KiB per core , L2 cache 256 KiB per core, L3 cache - 8192 KiB (8MB) shared by all cores

(where one KiB (Kibibyte not to be confused with Kilobyte) = 1024 bytes with 1 Bute being made up of eight 'bits' and with each 'bit' representing either a 1 or a 0 in binary)

Intel has added larger 'L4' caches to some recent (and upcoming) CPU's which it has called 'Iris Pro' in sales speak. This memory is primarily intended to speed up the operation of onboard graphics (iGPU) but can assist the CPU in running non graphics related tasks as well - most recently seen on some 'Broadwell' '14nm' CPU's which had 128 MB of eDRAM L4 cache

CPU Socket = The mechanical attachment point on a motherboard where a CPU is placed. Normally CPU sockets are made so that they mechanically wont fit an incompatible CPU but this is not always the case so you need to check your motherboard manual. Intel sockets are named for the amount of pins in the CPU socket (I.e. socket 1366, 1156 etc) - The pins in the socket connect to points on the underside of a CPU linking the motherboard electrically to the motherboard. A CPU socket will include a physical retention system for holding the CPU in place.

Integrated Heat Spreader (IHS) = placed onto of a CPU die to facilitate heat transfer from a CPU to a cooling system and to protect the die from damage. Intel CPU's can either be soldered or use TIM (see below) to join the die to the spreader (Soldered is generally superior)

TIM - Thermal interface material = A paste use to connect different object to assist in the transfer of heat. Commonly used in relation to CPU's between the IHS and the die and between the IHS and CPU cooler

Delidding = The process of removing an IHS with a view to changing the TIM for a better one. **Can cause damage to the die if not performed properly**

IPC - Instructions per cycle/clock =
A measure of how much work a CPU can be done either per cycle (so one of those many millions of cycles per second!) or more commonly used to compare different CPU's with the same clock/frequency. i.e. a 6700K@4GHz is a 4c/8t with higher IPC than a 4c/8t 3770K@4GHz (the Skylake CPU will be quicker despite having the same clock and core/thread count)
 
Last edited:
Intel always provide a detailed data sheet for their CPUs, it tells you all the technical stuff like voltages and speeds etc and provides pictures to give you a better idea of whats going on in their cpus.

If ya google Skylake Data Sheet youll find this which is referring to the latest generation of intel cpu.
 
Carakus this is amazing thank you very much, this is a clear and concise explanation. Obviously from what ive read, there is more than one technology per socket? So lets say for the skylake, there would be different versions for i3 i5 and i7, but also for sockets with different pins? Or am I terribly mistaken?
 
Carakus this is amazing thank you very much, this is a clear and concise explanation. Obviously from what ive read, there is more than one technology per socket? So lets say for the skylake, there would be different versions for i3 i5 and i7, but also for sockets with different pins? Or am I terribly mistaken?

Usually there are two different sockets per "generation" - a consumer/mainstream aimed socket (so for skylake this is 1151, for Haswell it is 1150, for Sandybridge it was 1155) this relates to the number of pins in the socket. It can also be described by the number given to it's motherboard chipset so Skylake is also referred to as Z170, Haswell Z97 and Sandy Z77.

There's also a socket aimed at the higher end of CPU performance, so for Haswell-e this is X99 chipset or socket 2011. Same goes for the high performance versions of other sockets/generations.

So it's possible to have two "current" i7 CPUs and they're both completely different and non-interchangeable, so the Haswell mainstream i7 4790k is a 4 core 8 thread CPU for socket 1150, where as the other 6 core+ performance Haswell i7 chips are socket 2011. This is often a big area of misconception as people assume an i7 is an i7.
 
In terms of performance, what is main difference between the mainstream and the specialist or higher end? I suppose it relates to computer power and the threads (maybe?) but if I was looking to spend more, is there any real impact on my gaming experience lets say. Between these two chip tiers
 
In terms of performance, what is main difference between the mainstream and the specialist or higher end? I suppose it relates to computer power and the threads (maybe?) but if I was looking to spend more, is there any real impact on my gaming experience lets say. Between these two chip tiers

I can't say there's a lot of impact, depends on games somewhat. For example Call of Duty is pretty cpu-light game, while ArmA is quite cpu-heavy, however general consensus is, enthusiast CPU prices are too high for gaming only. You're better off with an i5 (mainstream short term) or an i7 (mainstream longer term) rather than enthusiast grade i7. There are exceptions of course, but generally speaking, enthusiast cpus are enthusiast cpus for a reason, it being 'high compute power'.

A good example is 6700K SKylake (4c8t) versus 5820k Haswell-E (6c12t), they cost nearly the same, excluding motherboard and ram, 6700k is mainstream high-end, 5820k would be enthusiast low-end (comparing with other X99 cpus). In games they perform pretty similar because there aren't many of them that utilise the extra cores X99 has over Z170, but throw a compute task at these two, the 5820k will wipe the floor with 6700k and then some.
 
In terms of performance, what is main difference between the mainstream and the specialist or higher end? I suppose it relates to computer power and the threads (maybe?) but if I was looking to spend more, is there any real impact on my gaming experience lets say. Between these two chip tiers

It would depend on the resolution that you are gaming at. Most people usually go for a quadcore without hyperthreading such as an i5 6600k, and put more money towards the graphics card because its the component that provides the most gaming performance for your money
 
I can't say there's a lot of impact, depends on games somewhat. For example Call of Duty is pretty cpu-light game, while ArmA is quite cpu-heavy, however general consensus is, enthusiast CPU prices are too high for gaming only. You're better off with an i5 (mainstream short term) or an i7 (mainstream longer term) rather than enthusiast grade i7. There are exceptions of course, but generally speaking, enthusiast cpus are enthusiast cpus for a reason, it being 'high compute power'.

A good example is 6700K SKylake (4c8t) versus 5820k Haswell-E (6c12t), they cost nearly the same, excluding motherboard and ram, 6700k is mainstream high-end, 5820k would be enthusiast low-end (comparing with other X99 cpus). In games they perform pretty similar because there aren't many of them that utilise the extra cores X99 has over Z170, but throw a compute task at these two, the 5820k will wipe the floor with 6700k and then some.


What is the meaning of the X99? Is it related to the X86 files you see in windows c:/ file? And also by task do you mean running a background program such as a temp monitor or is it running a full program such as I dont know a recording program?
 
And also by task do you mean running a background program such as a temp monitor or is it running a full program such as I dont know a recording program?

Anything compute related and programmed to use multithreading. Includes WinRAR I believe, video making programs, image processing, server related workload. More cores you have, more properly programmed tasks you can have at once.

Let's say, compiling an hour worth of video can be done in background while gaming on a 5820k, but not so good on a 6700k, just as an example.
 
I havebeen reading about the alphanuremic designation of processors, taking into account what I have read from another blog, I assume that the 6700K Skylake, is from a later generation than the 5820K Hashwell-e chip and, however this does not neccessarily mean that the latest is the best option, as explained by Sixense. I assume newer generations have other built in utilities that do not directly correlate with computing power? or is there another real reason for wanting an older generation of CPU over a newer one?
 
The alphanumeric model numbers are a reflection of 'currentness'.

mainstream (up to 4 core cpu)
LGA 1155 3000 series is ivybridge
LGA 1150 4000 series is haswell
LGA 1150 5000 series is broadwell
current: LGA 1151 6000 series is skylake

enthusiast (4+ cores)
LGA 2011 4000 series is ivybridge
current: LGA 2011 5000 series is haswell
current - v.near future LGA 2011 6000 series will be broadwell

The enthusiast platform is subject to intel's server development process. The architecture in use lags behind mainstream as it takes a greater amount of time for design/validation/testing (helps to have the uarch tested in the wild through mainstream platform)
 
Last edited:
Older CPUs generally tend to cost less than new, even if they offer similar compute performance, there's quite a difference in the chipsets used, not only the CPUs. Now, I'm not very experienced in all this, just reading up in my free time, but the main jumps I believe are these:

LGA 1156 to LGA 1155. Going from first iteration of iX (like i7-920) to Sandy brought us SATA 3.0.

Z68 to Z77 aka Sandy Bridge to Ivy Bridge. Same socket LGA 1155, means you can use Sandy cpus in Ivy boards and vice versa I believe. However, Ivy brought PCIE 3.0 while Sandy is stuck on PCIE 2.0 motherboards. USB 3.0 was introduced in Ivy Bridge too I think.

Z87 to Z97 aka Haswell to Haswell Refresh. Again same socket, LGA 1150. Z97 introduced M.2 support. DDR4 RAM was introduced for X99 (Haswell-E) platform.

Z170 Skylake brought DDR4 RAM to mainstream.

These are mainly chipset changes, so more motherboard related. Now to answer the question...

An i5-2500k might perform very similar to an i5-6600k in games*, especially ones harder on the gpu, as shown in >>>LINK<<<, however you'll be stuck to PCIE2.0, no M.2 SSDs and your Graphics card's bandwith to the cpu/ram/hard drive will be thinner than 6600k's.

You might see some people upgrading their i7-920s for X5650 of sorts. It gives a boost over LGA1156's isevens, however they don't have the fancy SATA 3.0 transfer speeds.

Newer mainstream processors usually have a better integrated gpu (although weak) so you don't need a dedicated card, not sure if K-series cpu have them, and none of the enthusiast or Xeon cpus have an iGPU.

*in compute tasks a newer cpu will nearly always beat the cpu it's replacing
 
Wow guys this is turning into a really informative thread, thank you all for your input. The next question might open up to a whole new thread, but I have tried to get my head round the next topic, what are threads? I have read and come to understand that threads is what aids the cores in processing commands and calculations, however they are something not fully "physical"? correct me if I am wrong, once again I want to know what is the impact of threads and hyperthreading (Intel) when it comes to choosing a processor and how it helps to run your computer smoothly or how it aids with graphics and day to day computer functions.
 
I think Currently my advice would be to avoid AMD, this might change with the arrival of zen but generally Intel is far superior for the most part at any price point
 
YEah thats the feeling I got from reading the forum, I do currently have a bullzoder CPU, cant remember the specifics and it has run smoothly. However I believe now I am going tos tart doing a bit more with my pc, some video editing and stuff and I really ant to know the ins and out of processors. The insight you all give is great since it also comes from personal experience so I hope you can still provide me with your knowledge XD
 
Wow guys this is turning into a really informative thread, thank you all for your input. The next question might open up to a whole new thread, but I have tried to get my head round the next topic, what are threads? I have read and come to understand that threads is what aids the cores in processing commands and calculations, however they are something not fully "physical"? correct me if I am wrong, once again I want to know what is the impact of threads and hyperthreading (Intel) when it comes to choosing a processor and how it helps to run your computer smoothly or how it aids with graphics and day to day computer functions.

The thread is the stream of instructions that the CPU is processing. Its essentially the work the CPU is doing/has to do which is being fed to it by whatever programs are running. In general, a single core can only process a single thread at a time which is why modern CPUs have multiple cores so that they can run multiple threads in parallel to increase performance.

Hyperthreading is intels marketing term for a technology that they have developed which allows a single cpu core to process two threads simultaneously. This is not as fast as having two dedicated cores but still provides a nice performance boost over a single core without hyperthreading.
 
It's possible to have more threads per core, but that depends if the core is capable enough. Intel has only designed their cores to handle 2 threads each, as the second thread is basically the leftover resources on the core that the 1st thread is not using.

Oracle's Sparc T3 and beyond on the other hand does 8 threads per core :p.
 
Back
Top Bottom