EC2 Family Comparisons

26 Curated Guides

Engineered multi-way architectural evaluations and generational upgrade guides. Compare benchmark efficiency (CoreMark per dollar), DDR5 memory bandwidth, Nitro networking, and hourly rates without getting lost in single-instance matrices.

Open Custom Interactive Comparator →
High Impact

Flagship Architecture Comparisons

Silicon Trianglec7i · c7a · c7g · c7i-flex

C7i vs C7a vs C7g vs C7i-flex

4th Gen Intel Xeon Sapphire Rapids vs AMD EPYC Genoa vs AWS Graviton3 vs Flex

C7g (Graviton3) delivers the highest CoreMark performance-per-dollar and lowest raw hourly price (~19% cheaper than C7i). C7a (AMD Genoa) leads on raw x86 single-thread throughput. C7i-flex offers a 5% discount over C7i for unconstrained burstable web services.

Explore comparison →
Silicon Trianglem7i · m7a · m7g · m7i-flex

M7i vs M7a vs M7g vs M7i-flex

Comparing modern 4:1 memory-to-vCPU ratio General Purpose workhorses

For modern containerized web fleets, M7g is the runaway value leader with ~20% lower cost than M7i. For enterprise Java or memory-bound x86 monoliths, M7a delivers blistering cache and memory performance.

Explore comparison →
Silicon Triangler7i · r7a · r7g

R7i vs R7a vs R7g

8:1 memory-to-vCPU ratio for in-memory caches, Redis, and transactional databases

R7g offers the lowest cost per GiB of RAM on AWS for standard relational and in-memory databases. R7a offers up to 1.5 TiB of memory on top sizes with AMD Genoa's exceptional memory throughput.

Explore comparison →
Silicon Trianglet4g · t3 · t3a

T4g vs T3 vs T3a

Low-cost baseline compute with CPU credit burst mechanics

T4g is roughly 20% cheaper than T3 and outperforms both T3 and T3a on raw CPU benchmarks. It is the definitive default choice for dev/test and burstable production workloads.

Explore comparison →
Processor Rivals

Silicon Triangle (Intel vs AMD vs Graviton)

Direct head-to-head architectural comparisons between Intel Xeon, AMD EPYC, and AWS Graviton across identical compute tiers.

c7ic7ac7gc7i-flex
→

AWS C7 Generation: C7i vs C7a vs C7g vs C7i-flex

4th Gen Intel Xeon Sapphire Rapids vs AMD EPYC Genoa vs AWS Graviton3 vs Flex

Key Takeaway: C7g (Graviton3) delivers the highest CoreMark performance-per-dollar and lowest raw hourly price (~19% cheaper than C7i). C7a (AMD Genoa) leads on raw x86 single-thread throughput. C7i-flex offers a 5% discount over C7i for unconstrained burstable web services.
m7im7am7gm7i-flex
→

AWS M7 Generation: M7i vs M7a vs M7g vs M7i-flex

Comparing modern 4:1 memory-to-vCPU ratio General Purpose workhorses

Key Takeaway: For modern containerized web fleets, M7g is the runaway value leader with ~20% lower cost than M7i. For enterprise Java or memory-bound x86 monoliths, M7a delivers blistering cache and memory performance.
r7ir7ar7g
→

AWS R7 Generation: R7i vs R7a vs R7g

8:1 memory-to-vCPU ratio for in-memory caches, Redis, and transactional databases

Key Takeaway: R7g offers the lowest cost per GiB of RAM on AWS for standard relational and in-memory databases. R7a offers up to 1.5 TiB of memory on top sizes with AMD Genoa's exceptional memory throughput.
c6ic6ac6g
→

AWS C6 Generation: C6i vs C6a vs C6g

Intel Ice Lake vs AMD Milan vs Graviton2 compute comparison

Key Takeaway: C6g established Graviton as the gold standard for cloud cost optimization, offering up to 40% better price-performance over previous generations. C6a remains the budget x86 alternative.
m6im6am6g
→

AWS M6 Generation: M6i vs M6a vs M6g

General Purpose workhorses across Intel Ice Lake, AMD Milan, and Graviton2

Key Takeaway: M6g remains one of the most widely deployed Graviton instances in production. M6a provides an easy 10% discount for fleets that cannot immediately recompile for ARM64.
r6ir6ar6g
→

AWS R6 Generation: R6i vs R6a vs R6g

6th Gen Memory-Optimized Battleground (8:1 RAM-to-vCPU ratio)

Key Takeaway: R6g is the default high-efficiency choice for open-source caching and databases. R6a offers large memory footprints at lower cost than R6i.
t4gt3t3a
→

AWS Burstable Comparison: T4g vs T3 vs T3a

Low-cost baseline compute with CPU credit burst mechanics

Key Takeaway: T4g is roughly 20% cheaper than T3 and outperforms both T3 and T3a on raw CPU benchmarks. It is the definitive default choice for dev/test and burstable production workloads.
c8gm8gr8g
→

AWS Graviton4 Flagship Comparison: C8g vs M8g vs R8g

Next-gen Graviton4 ARM Neoverse V2 architecture across compute, memory, and general tiers

Key Takeaway: Graviton4 raises the bar across all dimensions. C8g is optimal for CPU-heavy microservices, M8g balances 4:1 RAM, and R8g brings 8:1 RAM with branch protection and enhanced security.
Upgrade Paths

Generational Evolution & Upgrades

Historical benchmarks, pricing drops, and Nitro architecture leaps across successive AWS instance generations.

c4c5c6ic7ic7g
→

AWS Compute Series Evolution: C4 vs C5 vs C6i vs C7i vs C7g

Tracking a decade of compute optimization from Xen hypervisors to Graviton3 Nitro

Key Takeaway: Upgrading from C5 to C7g delivers over 40% higher benchmark performance while dropping hourly compute costs by ~15-20%. Legacy C4 instances should be retired immediately.
m4m5m6im7im7g
→

AWS General Purpose Evolution: M4 vs M5 vs M6i vs M7i vs M7g

From Haswell to Sapphire Rapids and Graviton3 in the 4:1 RAM workhorse tier

Key Takeaway: M7g offers nearly triple the CoreMark performance-per-dollar compared to aging M4 instances, with 4x higher baseline network throughput.
r4r5r6ir7ir7g
→

AWS Memory Optimized Evolution: R4 vs R5 vs R6i vs R7i vs R7g

Memory tier advancement: DDR3 to DDR5, EBS throughput, and per-GB pricing

Key Takeaway: Upgrading database hosts from R5 to R7g lowers RAM cost by 20% while unlocking DDR5 memory channels that significantly decrease query latency for memory-bound workloads.
t2t3t4g
→

AWS Burstable Evolution: T2 vs T3 vs T4g

How AWS burstable compute changed from Xen T2 to Graviton2 T4g

Key Takeaway: T4g is superior to T2 in every measurable metric: 20% cheaper, significantly faster, and equipped with modern Nitro networking instead of legacy Xen queues.
a1c6gc7gc8g
→

AWS Graviton Evolution: A1 vs C6g vs C7g vs C8g

The complete journey of AWS custom ARM silicon from experimental A1 to Graviton4

Key Takeaway: Graviton transformed from an experimental alternative in A1 to AWS's premier high-performance compute silicon in Graviton4. Performance per core has more than quadrupled since 2018.
c5c6ic7i
→

AWS Pure Intel Compute Upgrade Path: C5 vs C6i vs C7i

Skylake/Cascade Lake vs Ice Lake vs Sapphire Rapids Intel compute

Key Takeaway: C7i provides a noticeable generational bump with DDR5 memory and Intel AMX, but costs slightly more per hour than older generations. Upgrading to C7i-flex mitigates the price increase.
m5m6im7i
→

AWS Pure Intel General Purpose Upgrade: M5 vs M6i vs M7i

Modernizing Intel Xeon fleets without recompiling for ARM

Key Takeaway: Moving from M5 to M6i or M7i provides substantial network and EBS throughput improvements, even if raw compute price-per-hour remains relatively flat.
r5r6ir7i
→

AWS Pure Intel Memory Upgrade: R5 vs R6i vs R7i

Intel Xeon memory tier evolution across three enterprise generations

Key Takeaway: R7i's DDR5 memory architecture provides crucial bandwidth advantages for memory-saturating databases that cannot run on Graviton.
Right-Sizing

Architecture & Sizing Tradeoffs

RAM ratios (2:1 vs 4:1 vs 8:1), local NVMe SSD vs EBS-only, 200 Gbps networking, and Intel Flex tiers.

c7gm7gr7g
→

RAM-to-vCPU Sizing Guide: C7g vs M7g vs R7g

Choosing between 2:1, 4:1, and 8:1 RAM-to-vCPU ratios on Graviton3

Key Takeaway: Never over-provision memory. C7g gives you the most CPU for your dollar (2:1 RAM). M7g is the safe general baseline (4:1 RAM). R7g is strictly for memory-saturated data stores (8:1 RAM).
c7ic7i-flex
→

AWS Intel Flex Comparison: C7i vs C7i-flex

Is the 5% discount on C7i-flex worth the CPU baseline constraints?

Key Takeaway: C7i-flex is a fantastic drop-in discount for web servers, dev environments, and microservices that do not peg CPU at 100% continuously. For scientific modeling or heavy compilation, stick to standard C7i.
m7im7i-flex
→

AWS General Purpose Flex Comparison: M7i vs M7i-flex

5% savings on standard 4:1 RAM general purpose workloads

Key Takeaway: M7i-flex provides effortless savings for general web workloads on Intel, making it the preferred x86 default over standard M7i for non-continuous compute.
c7gc7gd
→

Local NVMe SSD vs EBS-Only: C7g vs C7gd

When ephemeral local NVMe storage beats EBS gp3 and io2

Key Takeaway: C7gd costs slightly more (~10%) but includes blazing-fast local NVMe SSDs. Ideal for scratch disks, temporary build caches, and distributed caches where data persistence across instance stops is not required.
m7gm7gd
→

General Purpose NVMe Comparison: M7g vs M7gd

Local high-speed NVMe storage for general purpose application tiers

Key Takeaway: M7gd is the premier choice for distributed search and buffer tiers that need fast, unmetered local disk I/O alongside balanced memory and compute.
c6inc7gnc6gnc5n
→

AWS High-Networking Compute: C6in vs C7gn vs C6gn vs C5n

Comparing 100 Gbps and 200 Gbps network-optimized compute instances

Key Takeaway: C7gn brings Graviton3 efficiency to 200 Gbps networking with Elastic Fabric Adapter (EFA) support, making it the most cost-efficient high-bandwidth compute node available.
r7izz1dm5zn
→

AWS High-Frequency Compute: R7iz vs Z1d vs M5zn

Ultra-high single-core clock speeds up to 4.5 GHz for EDA, financial modeling, and per-core licensing

Key Takeaway: R7iz represents the state of the art in high-frequency compute, pairing Intel Sapphire Rapids cores boosting up to 3.9 GHz with up to 1 TiB of DDR5 memory.
Workload Specific

Specialized & Accelerated Compute

Ultra-dense NVMe storage engines, AI inference accelerators (NVIDIA vs Inferentia), and tightly-coupled HPC clusters.

i3i3eni4ii4g
→

Storage-Optimized IOPS Comparison: I3 vs I3en vs I4i vs I4g

Massive NVMe density for Cassandra, ScyllaDB, MongoDB, and Elasticsearch

Key Takeaway: I4i (AWS Nitro SSD) dramatically reduces latency jitter compared to legacy I3 instances. For ARM-ready databases, I4g delivers superior price-performance with AWS Nitro NVMe controllers.
g4dng5g6inf2
→

AWS AI Inference Acceleration: G4dn vs G5 vs G6 vs Inf2

NVIDIA T4 vs A10G vs L4 vs AWS Inferentia2 for LLM and ML model serving

Key Takeaway: G6 (NVIDIA L4) is the modern successor to G4dn/G5 for general PyTorch and TensorRT serving. Inf2 provides unmatched price-per-token for compatible transformer models compiled with AWS Neuron.
hpc6ahpc7ghpc7a
→

AWS High-Performance Computing: HPC6a vs HPC7g vs HPC7a

Tightly coupled clusters for computational fluid dynamics (CFD) and weather forecasting

Key Takeaway: HPC7a delivers the highest raw computing power for MPI simulations using AMD Genoa cores. HPC7g offers outstanding energy and cost efficiency for ARM-optimized scientific packages.