CPU

Intel or AMD

Only 2 brands to consider if you want to build or buy a PC. Previously Intel was the clear leader, though AMD's Ryzen series make it comparable with Intel in terms of performance.

AMD has successfully moved to 7nm since Zen 2 architecture, while Intel is still at 10 or 14 nm as of 2020.

AMD tends to have more cores and threads than Intel.

The product lines are well aligned:

• entry level: Intel CORE i3, AMD Ryzen 3
• mid level: Intel CORE i5, AMD Ryzen 5
• high end: Intel CORE i7, AMD Ryzen 7
• top level: Intel CORE i9, AMD Ryzen 9

Intel

Naming scheme:

• categry: the first part of the name, Core i7-8700 is Core i7
• generation: the first digit of the second part, e.g. Core i7-8700 is 8th Gen, Core i9-10900K is 10th Gen
• codes:
  • H: high performance
  • P: consumer
  • U: ultra-low power
  • K: unlocked for overclocking (AMD Ryzen processors are almost all unlocked by default)
  • F: no integrated GPU
  • E: embedded
  • T: low power

Integrated graphics:

• Intel Iris Xe: introduced in 11th gen (Tiger Lake)
• Intel UHD Graphics: the "old" integrated graphics

Intel Desktop CPU Generations Comparison:

Gen	CPU Family	Process	Cores	TDPs	Chipset	Platform	Memory	PCIe	Launch
2nd Gen	Sandy Bridge	32nm	4/8	35-95W	6-Series	LGA 1155	DDR3	PCIe 2.0	2011
3rd Gen	Ivy Bridge	22nm	4/8	35-77W	7-Series	LGA 1155	DDR3	PCIe 3.0	2012
4th Gen	Haswell	22nm	4/8	35-84W	8-Series	LGA 1150	DDR3	PCIe 3.0	2013
5th Gen	Broadwell	14nm	4/8	65-65W	9-Series	LGA 1150	DDR3	PCIe 3.0	2015
6th Gen	Skylake	14nm	4/8	35-91W	100-Series	LGA 1151	DDR4	PCIe 3.0	2015
7th Gen	Kaby Lake	14nm	4/8	35-91W	200-Series	LGA 1151	DDR4	PCIe 3.0	2017
8th Gen	Coffee Lake	14nm	6/12	35-95W	300-Series	LGA 1151	DDR4	PCIe 3.0	2017
9th Gen	Coffee Lake	14nm	8/16	35-95W	300-Series	LGA 1151	DDR4	PCIe 3.0	2018
10th Gen	Comet Lake	14nm	10/20	35-125W	400-Series	LGA 1200	DDR4	PCIe 3.0	2020
11th Gen	Rocket Lake	14nm	8/16	TBA	500-Series	LGA 1200	DDR4	PCIe 4.0	2021
12th Gen	Alder Lake	10nm	16/24?	TBA	600 Series?	LGA 1700	DDR5	PCIe 5.0?	2021
13th Gen	Meteor Lake	7nm?	TBA	TBA	700 Series?	LGA 1700	DDR5	PCIe 5.0?	2022?
14th Gen	Lunar Lake	TBA	TBA	TBA	800 Series?	TBA	DDR5	PCIe 5.0?	2023?

12th Gen Alder Lake processors: Based on the 7nm (10nm ESF) node, these chips leverage two core architectures: Golden Cove for performance, and Gracemont for the efficiency cores. 8 P-Cores/4 E-Cores.

Intel's Process Roadmap

New naming: Intel 10, Intel 7, Intel 4 (2022), Intel 3 (2023), Intel 20A (2024) (angstrom) and Intel 18A (2025)

Intel 18A node, equivalent to 1.8nm, will enter production in the second half of 2024.

The A standing for Ångström, or 10A is equal to 1nm.

AMD

Naming scheme:

• category: the first part of the name, e.g. Ryzen 5
• generation: same as Intel, the first digit of the second part, e.g. Ryzen 5 2600 is 2nd Gen
• suffix:
  • no suffix: standard desktop CPU.
  • X: high performance.
  • G: with a Radeon graphics card.
  • X3D: with AMD 3D V-Cache Technology.
  • S: low power desktop with GFX.
  • H: higher 45W TDP SKUs that offer higher performance.
  • HS: more efficient, with 35W TDP ratings.
  • HX: higher performance than regular H.
  • U: low TDP versions for better battery life and reduced cooling needs.

Zen Architecture:

• Zen 4: 5nm
• Zen 3: 7nm, since 2020; using the same 7nm process as Zen 2. The last microarchitecture before switches to DDR5 and new sockets.
• Zen 2: 7nm. Ppowering the 3-rd gen of Ryzen processors (Ryzen 3000 and Threadripper 3000).
• Zen+: 12nm, released in 2018.
• Zen: 14nm, released in 2017.

Threadrippers: high-performance multi-core processors.

APU

Accelerated Processing Unit (APU) is AMD’s marketing term for CPUs that have both a conventional CPU and a GPU on a single die. This is just a CPU with integrated graphics.

Specs

Clock Speed

measured in GHz, the higher the faster. Most of mainstream modern CPUs are below 4 GHz, and this speed is NOT increase year over year. Instead new generations of CPUs focus on adding more cores and being more efficient.

"Overclocking" is to increase this speed beyond the spec. Intel chips with "K" and almost all AMD chips can overclock.

Instructions Per Cycle (IPC)

The average number of instructions executed for each clock cycle. It depends on CPU architecture.

Cores and Threads

Cores is the number of processors. Each core may have one or multiple threads ("multithreading"). E.g. some CPU spec says "6 Cores / 12 Threads".

Cache

Cache on chip. If the data needed by CPU can be found in Cache, it saves a trip to RAM.

3 layers: L1, L2, and L3. L1 is the fastest and smallest, L3 is largest while slower than L1 and L2, but still faster than RAM.

Thermal design power (TDP)

Measured in watts (W), assuming not overclocking. The mainstream is 65W / 95W, and higher for high end CPUs. Your choice of PSU (Power Supply Unit) and CPU cooler should depend on this.

CPU base clock and CPU ratio

Base clock vs boost clock:

• base clock: the default clock speed at which a central processing unit (CPU) operates under normal conditions.
• boost clock: a higher clock speed that the CPU can reach for short bursts when more processing power is needed.

CPU Frequency = CPU base clock x CPU ratio
3.5 GHz = 100 MHz x 35

Max Boost

Max boost for AMD processors is the maximum frequency achievable by a single core on the processor running a bursty single-threaded workload.

Matching the motherboard

Make sure your motherboard is compatible

• PCIe Version
• Socket