✅ Multicore

• CPU is multicore if the neumann machine is replicated several times
• However, even if replicated, not all blocks are replicated

✔️ In CU

clock, sequencer is NOT replicated

• clock: NOT replicated❌ we need only one boss
• sequencer: NOT replicated❌ only one clock, only one sequencer to set only one rythm
• Program counter: Replicated⭕️ each core has one PC, every core can run own instuctions, every PC pointing to their own address and has each has own instructions, this is multi-tasking
• Instruction Register: Replicated⭕️ as each core has each PC
• Decoder: Replicated⭕️ to help IR

✔️ In ALU

ALL replicated

• InRA and InRA: Replicated⭕️
• Operational Circuit: Replicated⭕️
• ACCUM: Replicated⭕️

• State Register: Replicated⭕️

  • 👉🏻 Thus, all ALU blocks are replicated
  • bc all the instructions use different numbers

✔️ In MU

RAM is unique
Memory Controller

• The RAM is unique, outside the micro-processor
• thus the RAM is NOT replicated❌
• However, as 4 instructions are running at the same time, MAR, MDR and MS will be more stressed!
• 4 instructions will come at the same time
• 😫 So RAM is unique, but the internal blocks(MAR, MDR, MS) will be more stressed
• 💊 Memory Controller: So we have a super block
• So instead of having MAR, MDR, MS, we have ONE BIG BLOCK for controlling everything
• This huge block works simillar to replicating the RAM blocks(MAR, MDR, MS), and reduces the stree of the RAM
• sometimes, next to the Memory Controller, there is a System Agent block
• System Agent block: to help coordinate several RAMs

• When you have several RAMs, the System Agent block helps coordinating the divided RAM

• In the picture, there are 4 cores, so there are 4 neumann machines
• So if we have 4 cores, we will have
  • one clock, one sequencer
  • 4 PCs, IRs and Decoder, all blocks of ALU blocks
• if we have 4 cores we can run 4 instructions at the same time simultaneously
• And we have one RAM, but we have Memory Controller

✔️ In Peripherals

Transport Bus

• 😫 In multicore, the peripherals are also collapsed, very stressed running several instructions at the same time
• 💊 Transport Bus(also called Input/Output memory controller (I/O))
• used for controlling the peripherals at a very high speed
• ⭐️ In the exam, this block will be called Transport Bus⭕️, and NOT Input/Output memory controller❌

☑️ Graphics Processing Card

• If we are multicore, there can be a portion named GPU
• however, not mandatory
• helps the other standard graphics card with the graphics(images, videos)
• 🛠️ For games, music, movies

✅ Back Side Bus BSB

road inside core

• Each core inside has an internal sturcture for communication
• like internal streets/roads
• ⭐️ one BSB per core
• one core will have their own BSB
• 🐢 BSB is relatively slower to FSB

❓ If in core 3, the data from InRA is 3,
but cannot reach the OC,
what is the problem?

A: The BSB in core 3 is broken, having problems.

✅ Front Side Bus FSB

road to communicate among cores

• Apart from BSB, there is a circular highway to communicate AMONG cores
• circular road to connect all the cores
• ⭐️ there is only ONE FSB in one computer

❓ In 4 core computer, how many BSB and FSB?

A: There will be 4 BSBs and 1 FSB

• Sometimes, we have to send information from one core to another
• 🛠️ Then we need to use FSB

❓ If the result of operation 1 in core 1 cannot reach InRB of core 4, what is the problem?

A: The FSB is not working properly.

✔️ Speed of FSB

• 🐇 FSB is the fast speed bus
• If my clock has 2GHz, my FSB tends to be 2GHz too

• so the FSB has more or less same speed as the clock

• If the clock is much faster than the FSB, it means there is a bottleneck in the FSB
• example: If my clock is 2.5GB, and my FSB is 1GHz

✅ How many cores?

• In computing, every number should be a power of 2

• 1 core: mono-core

• However, sometimes core is not a power of 2, for example: 6 core computer
• then, it means 4 core and 2 cores are for mirroring
• 🛠️ for protection, backups

❓ In a 8 core computer, what is the result of this?
FSB + BSB + MC + TB
= 1 + 8 + 1 + 1
= 11

✅ Multi-thread

Dividing one process into several execution threads

• each of the threads(parts of process) will be processed by different blocks
• If you want your job to be done without interruption/disconnection
• use an independent execution thread

✔️ Resource sharing in multi-thread

• being multi-threading means more or less multi-tasking

• you can only do multi-threading if you have a resource sharing technique

• Mono thread: takes the whole micro processor to itself
• yellow runs, then blue
• Multi thread: the threads have to share the micro processor

• The microprocessor would do blue and yellow at the same time

• If I am multi-core, it is easy to be multi-thread
• If you have many cores, it does NOT necessarily mean your computer is faster ❌
  • It means your computer can run more things at the same time
  • It means your computer micro processor is more stressed, more hotter, need to ventilate more, need to refresh your computer more

✔️ Optimization

• If you are mono-core you can be multi-thread ⭕️
• But you have to share your resources very well
• 💊 Optimization tools: help share resources among threads in mono-core

✅ Logic Core

• Physical core: number of cores we have, 8 core = 8 physical core
• Threads per core: how many thread in a core
• Logic cores: how many threads I have in total

❓ If we have 4 cores, and each core executes 2 threads, how many threads in total can we execute?
A: 8 threads in total

Physical core: 4
Threads per core: 2
Logic cores: 8

I have a computer of physical core of 4, thread per core is 2
then I have 8 logic cores
⚠️ I have 4 physical cores, but have 8 logic cores, bc I feel like I have 8 workers!

✅

✅