EC2 Instance Types — Family Codes & Meanings

The 5 EC2 categories (mental models)

Type	Meaning	Use for	Families
General	Balanced CPU/mem/net	Web servers, dev, small DB	t3, m5, m7g
Compute Optimized	CPU-intensive	Transcoding, batch, gaming, ML inference	c5, c7g
Memory Optimized	RAM-intensive	In-memory caches (Redis), in-memory DB, SAP HANA	r5, x2gd, u-*, z1d
Accelerated	GPU/FPGA/AI chips	ML training, rendering, simulation	p4, g5, inf2, trn1
Storage Optimized	I/O-bound, FAST disk (not just big)	NoSQL (Cassandra/Mongo), data warehouse, ElasticSearch	i3, d2, h1

Storage Optimized ≠ “more storage” — it’s low-latency high-I/O (NVMe SSD), for random read/write heavy workloads.

Family code letters

[family][generation][g?].[size] e.g. c5.large, m7g.xlarge, t3.micro

Letter	Meaning
t	Burstable (CPU credits) — cheap, low avg load
m	General purpose
c	Compute
r	Memory (RAM)
x	Extra-large memory
z	High-frequency (high clock speed)
u	Ultra-high memory (up to 24TB)
p	GPU (Tesla)
g	GPU (graphics)
inf	Inferentia (ML inference chip)
trn	Trainium (ML training chip)
i	Storage / I/O optimized
d	Dense storage (HDD, throughput)
h	High disk throughput
f	FPGA
a	ARM (first-gen Graviton)
mac	Mac instances

Generation + Graviton

Number = generation (5 = 5th gen)
g suffix = Graviton (AWS ARM chip): m5 → m6g → m7g. ~20% better price/performance
No g = x86 (Intel/AMD)

Sizes

nano → micro → small → medium → large → xlarge → 2xlarge → ... → 16xlarge → 24xlarge → metal

Each step ≈ doubles vCPU/RAM
metal = bare metal (no virtualization)
Free tier = t3.micro (1 vCPU, 1GB RAM, 750h/mo first 12 months)

Example decode

t3.micro      → t (burstable) gen3, micro size
c7g.2xlarge   → c (compute) gen7 + g (Graviton ARM), 2xlarge
r5.large      → r (memory) gen5, large
p4d.24xlarge  → p (GPU) gen4, 24xlarge

Exam relevance

t = burstable cheap, c = compute, r = memory, m = general
g suffix = Graviton/ARM (cheaper)
Scenario → match type (compute-intensive → c, RAM-heavy → r, GPU → p/g)

[[aws/index]]
[[aws/compute-map/index]]