g3.8xlargevsg5.8xlarge

g3.8xlarge is 7% cheaper than g5.8xlarge on demand in US East (N. Virginia), saving $122.64 a month.

At a glanceg3.8xlargeg5.8xlarge
On-demand$2.28/hr7% less$2.448/hr
Lowest rate$0.9581/hrSpot, can be interrupted$0.8379/hr13% lessSpot, can be interrupted
Size32 vCPUs, 244 GiB32 vCPUs, 128 GiB
ProcessorIntel Xeon E5-2686 v4 (Broadwell)AMD EPYC 7R32
Architecturex86_64x86_64

Prices in US East (N. Virginia)

“Off” is the discount against each type’s own on-demand price. Linux, USD per hour.

Optiong3.8xlargeg5.8xlargeg5.8xlarge vs g3.8xlarge
On-demandPay by the hour$2.28$2.448▲ 7%
SpotCan be interrupted58% off$0.958166% off$0.8379▼ 13%
1-yr reservedNo upfront—37% off$1.542—
3-yr reservedNo upfront—57% off$1.058—
1-yr savings planHourly commitment—21% off$1.928—
3-yr savings planHourly commitment—46% off$1.322—

Key differences

The biggest lead each one has.

Where g3.8xlarge leads
On-demand price$2.28 vs $2.448 per hour
7% cheaper
GPUs2 vs 1
2× higher
Memory (GiB)244 vs 128
91% higher
Where g5.8xlarge leads
GPU memory24 GB vs 16 GB
50% higher
Clock speed2.8 GHz vs 2.3 GHz
22% higher

Specs

41 of 45 specs differ. Those are in bold.

Compute9 of 11 differg3.8xlargeg5.8xlarge
vCPUs3232
Physical cores–16
Threads per core–2
Architecturex86_64x86_64
ProcessorIntel Xeon E5-2686 v4 (Broadwell)AMD EPYC 7R32
CPU maker–AMD
Clock speed2.3 GHz2.8 GHz
Sustained clock–3.3 GHz
Processor featuresIntel AVX; Intel AVX2; Intel TurboAMD Turbo; AVX; AVX2
Burstable–No
ECU97–
Memory1 of 1 differg3.8xlargeg5.8xlarge
Memory (GiB)244128
Accelerators3 of 3 differg3.8xlargeg5.8xlarge
GPUs21
GPU2 × NVIDIA Tesla M601 × NVIDIA A10G
GPU memory16 GB24 GB
Networking8 of 8 differg3.8xlargeg5.8xlarge
Network10 Gigabit25 Gigabit
Network baseline–25 Gbps
Network peak–25 Gbps
Enhanced networking–Yes (ENA)
EFA–Yes
IPv6–Yes
Max network interfaces–8
IPv4 addresses per interface–30
Storage12 of 12 differg3.8xlargeg5.8xlarge
StorageEBS only1 x 900 GB NVMe SSD
Local disks–1 × 900 GB SSD
EBS optimized–Yes
EBS over NVMe–Required
EBS throughput7000 Mbps16000 Mbps
EBS baseline bandwidth–16,000 Mbps
EBS max bandwidth–16,000 Mbps
EBS baseline throughput–2,000 MB/s
EBS max throughput–2,000 MB/s
EBS baseline IOPS–65,000
EBS max IOPS–65,000
Max EBS volumes–25
Instance details8 of 10 differg3.8xlargeg5.8xlarge
Familyg3g5
CategoryGPU instanceGPU instance
Current generationYesYes
Normalization factor642.433399602
Bare metal–No
Hypervisor–Nitro
Placement groups–Cluster, Partition, Spread
Hibernation–No
Nitro Enclaves–Yes
Boot modes–Legacy BIOS, UEFI

On-demand prices by region

Cheapest for g3.8xlarge first. Linux, USD per hour.

Regiong3.8xlargeg5.8xlargeDifference
US East (N. Virginia)Used in summaryus-east-1$2.28$2.448▲ 7%
US East (Ohio)us-east-2$2.28$2.448▲ 7%
US West (Oregon)us-west-2$2.28$2.448▲ 7%
EU (Ireland)eu-west-1$2.42$2.733▲ 13%
Canada (Central)ca-central-1$2.832$2.718▼ 4%
Asia Pacific (Seoul)ap-northeast-2$2.84$3.01▲ 6%
EU (Frankfurt)eu-central-1$2.85$3.061▲ 7%
EU (London)eu-west-2$2.858$3.107▲ 9%
Asia Pacific (Tokyo)ap-northeast-1$3.16$3.55▲ 12%
Asia Pacific (Sydney)ap-southeast-2$3.508$3.183▼ 9%

Keep exploring

Each type’s own page, and how it compares with others.

g3.8xlargeGPU instance, g3 family
g3.8xlarge prices and specsEvery size in the g3 family

Compare g3.8xlarge with

g2.8xlarge
g5.8xlargeGPU instance, g5 family
g5.8xlarge prices and specsEvery size in the g5 family

Compare g5.8xlarge with

g5g.8xlargeg6.8xlarge

Pay the committed rate, with commitment protection

The best committed rate on these two is 0–57% below on-demand. Connect your cloud accounts to unlock greater savings with commitment protection.

Connect AWS account

AWS EC2 Linux prices in USD. Monthly figures use 730 hours. Prices last updated 12 Nov 2026.