Processors

Every CPU contains:

1. Arithmetic and Logic Unit (ALU).

• Performs arithmetic operations (+, −, ×, ÷).
• Performs logical operations (AND, OR, NOT, comparisons).
• The 'doer' of computation.

2. Control Unit (CU).

• Fetches instructions from RAM.
• Decodes what each instruction means.
• Directs the flow of data between ALU, registers and memory.
• Coordinates the Fetch-Decode-Execute cycle — the basic operation of every CPU.

3. Registers.

• Tiny, ultra-fast storage locations inside the CPU.
• Hold the current instruction, operands, results and addresses.
• The fastest memory in the entire computer (sub-nanosecond access).

Plus the cache hierarchy:

• L1 cache — smallest, fastest (per core).
• L2 cache — larger, slightly slower (per core).
• L3 cache — largest, slower (shared between cores).

Clock speed is measured in Hertz (Hz):

• 1 Hz = 1 cycle/second.
• 1 GHz = 1,000,000,000 cycles/second.

Each clock cycle can execute part of an instruction. Higher clock = more cycles per second = more work done.

Cores are independent processing units within a CPU.

• Single-core: 1 core (legacy).
• Dual-core: 2 cores.
• Quad-core: 4 cores (common entry-level).
• Octa-core: 8 cores (modern standard for laptops).
• Modern professional CPUs: 16, 24, 32, 64+ cores.

Hyper-Threading / SMT (Simultaneous Multi-Threading): allows each core to handle 2 threads, presenting twice as many 'logical cores' to the OS. Boosts multi-threaded performance by ~30%.

Theoretical ops/sec = Clock speed × Cores.

BUT: real-world performance depends on:

• Whether software is multi-threaded.
• Memory latency (cache misses).
• Branch prediction accuracy.
• Power/thermal throttling.

Modern reality: a 2.5 GHz octa-core often outperforms a 4.0 GHz dual-core in multi-threaded work (rendering, compiling).

Bit width refers to the size of data the CPU can process in one operation and the size of memory addresses it can handle.

32-bit CPU:

• Can process 32-bit numbers natively.
• Can address 2³² = 4,294,967,296 bytes = 4 GB of RAM.
• Found in older systems and some embedded devices.

64-bit CPU:

• Can process 64-bit numbers natively.
• Can address 2⁶⁴ ≈ 1.8 × 10¹⁹ bytes ≈ 16 EB (Exabytes).
• Standard in modern desktops, laptops, phones.

Practical implications:

• A 32-bit OS cannot use more than 4 GB RAM, even if more is installed.
• 64-bit software can handle larger files and datasets in memory.
• All major modern OSes (Windows 11, macOS, Linux, Android, iOS) are 64-bit.

Why 32-bit still exists:

• Embedded systems with limited memory.
• Legacy industrial controllers.
• Some IoT devices.

x86 (and x86-64):

• Manufacturer: Intel, AMD.
• Architecture: CISC (Complex Instruction Set Computing).
• Strength: high peak performance, mature software ecosystem.
• Power: relatively high (laptops typically 15–45 W).
• Used in: desktops, gaming PCs, servers, traditional laptops.

ARM:

• Manufacturer: many licensees — Apple (M-series), Qualcomm (Snapdragon), Samsung (Exynos), MediaTek, AWS (Graviton).
• Architecture: RISC (Reduced Instruction Set Computing).
• Strength: excellent power efficiency, scales from microwatts to high-end laptops.
• Power: very low (smartphones 1–5 W, M-series laptops 10–30 W with much longer battery).
• Used in: smartphones, tablets, embedded systems, modern Apple Macs, Snapdragon X Windows laptops.

Trend: ARM is encroaching on desktop computing. Apple's switch to M-series (2020) showed ARM can match or beat x86 while using FAR less power. Microsoft now ships Windows on ARM. Servers at AWS, Google and Microsoft are moving to ARM (Graviton, etc.) for power savings.

Why both still exist:

• x86 has decades of legacy Windows software.
• ARM has the best power efficiency.
• Software ecosystems are still bridging.

When evaluating a CPU spec sheet, look at:

1. Clock speed — base AND boost (modern CPUs run faster when fewer cores are active).
2. Cores and threads — physical cores vs logical (SMT) threads.
3. Cache — total L3 cache especially matters for multi-app workloads.
4. Architecture / generation — newer is generally better per-clock (Intel 13th-gen > 8th-gen at same GHz).
5. TDP (Thermal Design Power) — heat output ≈ continuous power draw. Higher TDP = more sustained performance but more cooling needed.
6. Integrated graphics — important if no dedicated GPU.
7. Memory support — what RAM speed and capacity the CPU supports.

Watch out for:

• 'Boost' clocks that only apply briefly to 1 or 2 cores.
• Marketing claiming 'X+ cores' without distinguishing physical vs logical.
• Comparing across generations using only clock speed (misleading).

Even the fastest CPU can be bottlenecked by:

• Slow RAM — CPU waits for data; cache misses are expensive.
• Slow storage — if frequently loading from disk.
• Single-threaded software — can't use multiple cores.
• Cooling — CPUs throttle (reduce clock speed) when too hot.
• Power delivery — laptops on battery may reduce clock to save power.
• GPU bottleneck — if the system is GPU-intensive, CPU may be idle.

Understanding the bottleneck is key to upgrading effectively. Adding a faster CPU doesn't help if the limit is slow disk or insufficient RAM.