g3s.xlarge vs p2.xlarge

Accelerators
Instance Type	g3s.xlarge	p2.xlarge
Instance Family	GPU instance	GPU instance
Details	Machine Learning (ML), 3D visualizations, graphics-heavy remote workstations, 3D rendering, application streaming, video encoding, and server-side graphics tasks. g – GPU (Graphics Processing Unit) 3 – Generation xlarge – Size	GPU intensive tasks like machine/deep learning, high-performance computing p – Performance / GPU (Graphics Processing Unit) accelerated 2 – Generation xlarge – Size
vCPUs	4	4
Memory	30.5 GiB	61 GiB
CPU Architecture	n/a	n/a
Processor	Intel Xeon E5-2686 v4 (Broadwell)	Intel Xeon E5-2686 v4 (Broadwell)
Threads Per Core	0	0
GPU	no	no
GPUs	1	1
Inference Accelerator	no	no
Accelerator Memory	8 GiB
FPGA	no	no
Instances
Sizes	g3s.xlarge	p2.16xlarge, p2.8xlarge, p2.xlarge
Free Tier	no	no
Burstable	no	no
Hibernation	no	no
EC2 Compute Units	13	16
Instance Storage	EBS only	EBS only
EBS (Elastic Block Store)
Dedicated Throughput	100 MBps	750 MBps
Networking
Networking Performance	10 Gigabit	High
Enhanced Networking	yes	yes
IPv6 Support	no	no
EFA Support	no	no
Encryption in Transit Support	no	no
ENA SRD Support	no	no
Avg Price Diff	1.00x	1.43x

Regional Prices

US Dollar ($)

Per Hour

Linux

Shared Tenancy

On Demand

Geography	Region	g3s.xlarge	p2.xlarge
Hidden	Hidden (hidden-1)	0.7960	0.9720
Hidden	Hidden (hidden-2)	0.8680	1.0800
Hidden	Hidden (hidden-3)	0.9340	1.3260
Hidden	Hidden (hidden-4)	0.9380	1.4650
Hidden	Hidden (hidden-5)	0.9400	1.5420
Hidden	Hidden (hidden-6)	1.0400	1.5420
Hidden	Hidden (hidden-7)	1.1540	1.7180
Hidden	Hidden (hidden-8)	n/a	1.7180
US	US East (Ohio) (us-east-2)	0.7500	0.9000
US	US East (Virginia) (us-east-1)	0.7500	0.9000
US	US West (Oregon) (us-west-2)	0.7500	0.9000

Other instances worth weighing against these.