CPU Performance Metrics
Clock speed (GHz): number of cycles per second the CPU can execute instructions. Higher GHz = faster — but only when comparing the same CPU generation and architecture. A 3.5 GHz modern CPU outperforms a 5 GHz decade-old CPU due to IPC (Instructions Per Cycle) improvements. Core count: number of independent processing units. Multi-core CPUs execute multiple threads simultaneously. Cores vs threads: physical cores execute instructions. Hyper-threading (Intel) / SMT (AMD): presents each physical core as two logical processors to the OS — improves throughput for multi-threaded workloads but does not double performance. An 8-core CPU with HT appears as 16 logical processors in Task Manager. L1/L2/L3 cache: on-chip memory for frequently accessed data. L1: fastest, smallest (32-64 KB per core). L2: larger, slightly slower (256 KB-1 MB per core). L3: shared among all cores, largest (8-64 MB). More cache = fewer slow RAM accesses.
CPU Thermal Management
CPUs generate significant heat during operation. Without adequate cooling, thermal throttling occurs: the CPU reduces clock speed automatically to prevent overheating. Thermal Design Power (TDP): rated maximum heat output in watts — guides cooler selection. A CPU with 65W TDP needs a cooler rated for at least 65W. CPU socket contact: IHS (Integrated Heat Spreader) on the CPU makes contact with the cooler. Thermal paste (thermal interface material — TIM): fills microscopic gaps between IHS and cooler to improve heat transfer. Apply properly: pea-sized dot in center, let mounting pressure spread it, or thin uniform layer. Too much thermal paste: can spill onto socket contacts. Too little: poor thermal contact. Old paste dries out — reapply when removing a cooler. Cooler types: stock cooler (included with CPU — adequate for stock speeds), aftermarket air cooler (larger heatsink + fan — better thermal performance), AIO liquid cooler (all-in-one liquid cooling — best for high TDP / overclocking), custom liquid loop (best performance, complex).
Intel vs AMD Platform Differences
Intel CPUs: Socket LGA (Land Grid Array) — pins on the motherboard, flat contacts on CPU. Common modern sockets: LGA1700 (12th/13th/14th gen Intel Core), LGA1200 (10th/11th gen). Intel Core nomenclature: Core i3 (entry), Core i5 (mainstream), Core i7 (performance), Core i9 (enthusiast). Suffix: K = unlocked multiplier (overclockable), F = no integrated graphics, H = high performance mobile. AMD CPUs: Socket AM (Array Module) or TR (Threadripper) — pins on the CPU, ZIF (zero insertion force) socket on motherboard. Common modern sockets: AM5 (Ryzen 7000+), AM4 (Ryzen 5000 and earlier). AM4 motherboards are reused across multiple CPU generations (good longevity). AMD naming: Ryzen 3 (entry), Ryzen 5 (mainstream), Ryzen 7 (performance), Ryzen 9 (enthusiast), Threadripper (workstation/HEDT). Intel and AMD CPUs are NOT cross-compatible — socket specific.
Overclocking
Overclocking: running the CPU (or RAM, GPU) above its rated specifications to achieve higher performance. Risks: increased heat output (requires better cooling), instability (system crashes, data corruption), reduced component lifespan, voids processor warranty (typically). Requirements: unlocked CPU (Intel K-series, AMD Ryzen X/non-X/all Ryzen on X570+), compatible motherboard (Z-series for Intel, X-series or B-series for AMD), adequate cooling (stock coolers insufficient for significant OC). How it works: increase the CPU multiplier in BIOS/UEFI. Also increase voltage (Vcore) to maintain stability — increases heat proportionally. Stability testing: Prime95, OCCT, Cinebench. Temperature monitoring: HWiNFO64, Core Temp. Safe temperature limits: generally keep under 90°C for sustained loads (80°C preferred). Underclocking/undervolting: reducing clock speed and voltage for lower power consumption and heat — useful for small form factor builds or quiet systems.
Integrated vs Discrete Graphics
Integrated GPU (iGPU): built into the CPU. Intel: Intel HD/UHD/Iris Xe Graphics (suffix F = no iGPU). AMD: Radeon Graphics (Ryzen CPUs with 'G' suffix have iGPU — e.g., Ryzen 5 5600G). Advantages: no separate GPU card required (lower cost, lower power). Shares system RAM (uses a portion of RAM as VRAM). Adequate for: office work, video playback, light web browsing. Not suitable for: gaming, 3D rendering, machine learning, video editing with GPU acceleration. Discrete GPU: separate graphics card with dedicated VRAM. Connects via PCIe x16 slot. Superior performance for graphics-intensive workloads. Dual graphics: if both iGPU and discrete GPU are present, the discrete GPU is typically used by default. iGPU can be used for quick-sync (video encoding acceleration) alongside discrete GPU in some configurations.