HIGH-PERFORMANCE ARCHITECTURE
SLIDE 01
Cache Memory in
Gaming Computers
Working • Internals • Pros and Cons • Applications
L3
L2
L1
CPU
Hardware Architecture & Game Optimization
1
What Is Cache Memory?
⚡
Tiny, ultra-fast memory beside CPU
⚡
Built from SRAM, not DRAM
⚡
Stores frequently used data
⚡
Hides slow RAM latency
⚡
Managed automatically by hardware
⚡
Sized in KB to MB
Memory hierarchy: fast and small → slow and large
Registers
L1 • ~1 ns
L2 • ~4 ns
L3 • ~10 ns
RAM • ~70 ns
SSD • ~100 µs
2
How Cache Works
CPU needs data
Look in L1 cache
Hit?
Yes
HIT: data returned (~1 ns)
No
Search L2 and L3
Fetch from RAM
Fill cache line, return
Miss path: slower, then data cached
3
Inside a Cache
Memory address
Tag
Identifies the memory block
Index
Selects the cache row
Offset
Picks byte in line
| Valid | Tag | Data block (64 B) |
|---|---|---|
| 1 | 0x1F | 64 bytes |
| 1 | 0x3A | 64 bytes |
| 1 | 0x9C | 64 bytes |
| 1 | 0x07 | 64 bytes |
= ?
HIT: send word
MISS: next level
Index finds row; tag confirms hit
4
Advantages and Disadvantages
⚡
Advantages
- ▸ Far lower memory latency
- ▸ Higher, smoother frame rates
- ▸ Fewer slow RAM accesses
- ▸ Better CPU and GPU utilisation
- ▸ Lower power per access
⚠️
Disadvantages
- ▸ Costly per megabyte
- ▸ Uses large chip area
- ▸ Limited, fixed capacity
- ▸ Adds heat and power draw
- ▸ Misses still stall processing
5
Applications in Gaming
⚡
CPU L3
AMD X3D chips boost game FPS
⚡
GPU caches
Textures, shaders, geometry stay on-chip
⚡
Console APUs
Shared caches for CPU and GPU
⚡
SSD cache
DRAM cache speeds game loading
⚡
Game engines
Cache-friendly data layouts, fewer misses
⚡
Asset streaming
Hot assets cached, fewer stutters
⚡
Takeaway: more cache, steadier FPS
6