g4dn.8xlarge by Amazon Web Services

g4dn.8xlarge is a Graphics intensive [Instance store volumes] [Network and EBS optimized] Gen4 8xlarge server offered by Amazon Web Services with 32 vCPUs, 128 GiB of memory and 900 GB of storage. The pricing starts at 0.2308 USD per hour.
32 vCPU
128 GiB Memory
900 GB Storage
1 GPU
Spare SCore
27,104
(All-cores)
1,286
(Single-core)

Specifications

Server Metadata

Vendor ID
aws
Name
g4dn.8xlarge
Description
Graphics intensive [Instance store volumes] [Network and EBS optimized] Gen4 8xlarge
Family
g4dn
Hw Virt
Status
active
Observed At
2026-11-01T13:05:52.367835

Availability

REGION / IDSPOTONDEMAND
Northern Virgina (US) / us-east-1
0.78572 USD/h 2.176 USD/h
Oregon (US) / us-west-2
0.9128 USD/h 2.176 USD/h
Ohio (US) / us-east-2
1.0735 USD/h 2.176 USD/h
Stockholm (SE) / eu-north-1
0.302267 USD/h 2.308 USD/h
Mumbai (IN) / ap-south-1
0.770467 USD/h 2.395 USD/h
Quebec (CA) / ca-central-1
0.53475 USD/h 2.416 USD/h
Dublin (IE) / eu-west-1
1.001567 USD/h 2.426 USD/h
Paris (FR) / eu-west-3
0.7819 USD/h 2.544 USD/h
London (GB) / eu-west-2
0.539 USD/h 2.546 USD/h
Milan (IT) / eu-south-1
0.3973 USD/h 2.547 USD/h
California (US) / us-west-1
1.0251 USD/h 2.611 USD/h
Seoul (KR) / ap-northeast-2
1.2103 USD/h 2.677 USD/h
Frankfurt (DE) / eu-central-1
1.045367 USD/h 2.72 USD/h
Sydney (AU) / ap-southeast-2
1.4575 USD/h 2.83 USD/h
Cape Town (ZA) / af-south-1
0.62975 USD/h 2.887 USD/h
Osaka (JP) / ap-northeast-3
0.70745 USD/h 2.938 USD/h
Tokyo (JP) / ap-northeast-1
1.447733 USD/h 2.938 USD/h
Singapore (SG) / ap-southeast-1
0.997267 USD/h 3.045 USD/h
Hong Kong (HK) / ap-east-1
0.514367 USD/h 3.351 USD/h
Sao Paulo (BR) / sa-east-1
0.382967 USD/h 3.698 USD/h

Processor

vCPUs
32
Hypervisor
nitro
CPU Allocation
Dedicated
CPU Cores
16
CPU Speed
2.5 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU Model
8259CL
CPU L1D Cache
32 KiB
CPU L1D Cache Total
512 KiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
512 KiB
CPU L2 Cache
1 MiB
CPU L2 Cache Total
16 MiB
CPU L3 Cache
36 MiB
CPU L3 Cache Total
36 MiB
CPU Flags
fpu, vme, de, pse, tsc, msr, pae, mce, cx8, apic, sep, mtrr, pge, mca, cmov, pat, pse36, clflush, mmx, fxsr, sse, sse2, ss, ht, syscall, nx, pdpe1gb, rdtscp, lm, constant_tsc, rep_good, nopl, xtopology, nonstop_tsc, cpuid, aperfmperf, tsc_known_freq, pni, pclmulqdq, ssse3, fma, cx16, pcid, sse4_1, sse4_2, x2apic, movbe, popcnt, tsc_deadline_timer, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, abm, 3dnowprefetch, pti, fsgsbase, tsc_adjust, bmi1, avx2, smep, bmi2, erms, invpcid, mpx, avx512f, avx512dq, rdseed, adx, smap, clflushopt, clwb, avx512cd, avx512bw, avx512vl, xsaveopt, xsavec, xgetbv1, xsaves, ida, arat, pku, ospke, avx512_vnni
Ecpus
21.1
Scalability
65.94

System Resources and Accelerators

MEMORY
Memory Amount
128 GiB
Memory Amount Actual
128 GiB
Memory Generation
DDR4
Memory Speed
2933 Mhz
ACCELERATORS
Accelerator Count
1
Accelerator Memory Min
16 GiB
Accelerator Memory Total
16 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Turing
Accelerator Model
T4
Accelerators
  • NVIDIA Turing Tesla T4 (Memory amount: 15360, Firmware version: 590.48.01, BIOS version: 90.04.96.00.02, Clock rate: 1590)
STORAGE
Storage Size
900 GB
Storage Type
nvme ssd
Storages
  • 900 GB nvme ssd
NETWORK
Network Speed Baseline
50 Gbps
Network Speed Max
50 Gbps
Network Storage Speed Baseline
9.5 Gbps
Network Storage Speed Max
9.5 Gbps
Inbound Traffic
0 GB/month
Outbound Traffic
0 GB/month
IPv4
0

CPU and System Topology

Server Description

A GPU-accelerated instance featuring dedicated NVIDIA T4 graphics, local NVMe storage, and high-speed networking for machine learning inference.

GPU AcceleratedStorage & Database

Amazon Web Services g4dn.8xlarge is a GPU-accelerated instance optimized for graphics-intensive and machine learning workloads. Built on the Nitro hypervisor, it features an Intel Xeon 8259CL processor with 32 dedicated vCPUs, 128.0 GB of DDR4 memory, and a baseline network bandwidth of 50 Gbps. Hardware acceleration is provided by a single NVIDIA Turing T4 GPU with 16 GB of VRAM, complemented by 900 GB of local NVMe SSD storage. While single-core CPU benchmarks are relatively weak, the instance delivers average multi-core performance and top-tier results in LLM inference, particularly for small and medium models. This hardware profile makes the g4dn.8xlarge highly cost-efficient for machine learning inference, graphics rendering, and video transcoding workloads that benefit from local high-speed storage and dedicated GPU acceleration.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

1.2Score
Precomputed compound score for Cache Intensive workloads. A weighted average (geometric mean) of benchmark scores compared to their medians: score = ∏ (x_i / m_i)^(w_i / Σw). The score of 1.0 represents a synthetic baseline server with the median performance of each component benchmark; 0.5 means roughly half the performance; and 2.0 means twice the performance of that reference profile. Component weights: 50% Redis RPS (pipeline=1, SET), 20% Redis RPS (pipeline=16, SET), 10% PassMark Memory Mark (composite), 10% Memory bandwidth (read, 16 MB ~ L3), 10% PassMark single-thread CPU. Rationale for component selection: In-memory key-value store workload, mixing direct Redis performance metrics with memory speed and latency benchmarks, and single-core CPU performance profiles.
Component Score Weight Impact
Raw Reference Normalized Target Actual
Redis RPS (pipeline=1, SET) 2,693,331 ops/sec 1,918,343 ops/sec 1.4 50.00% 50.00% +18.00%
Redis RPS (pipeline=16, SET) 15,974,891 ops/sec 11,704,734 ops/sec 1.3 20.00% 20.00% +5.40%
PassMark Memory Mark (composite) 2,531 2,416 1 10.00% 10.00% 0.00%
Memory bandwidth (read, 16 MB ~ L3) 115,570 MB/sec 108,466 MB/sec 1.1 10.00% 10.00% +0.96%
PassMark single-thread CPU 1,816 Mops/s 2,365 Mops/s 0.75 10.00% 10.00% -2.80%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
5.7 GB/s
Node GPU-to-host bandwidth
6.6 GB/s
Node host-to-GPU bandwidth
6.3 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
6.6 GB/s
Per-GPU host-to-GPU bandwidth
6.3 GB/s
Per-GPU host-device latency
655 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Version (driver)
 12.6
nvbandwidth Version
 v0.5
GPU Count
 1
P2P Supported
No
Efficiency
SM/CE copy efficiency
1.00

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 1,131

Geekbench Multi-Core

Score: 9,943

Passmark CPU Scores

BENCHMARKSCORE
Mark
 25,379
Compression
 395,777
Encryption
 12,445
Extended Instructions
 24,908
Floating Point Maths
 48,638
Integer Maths
 90,764
Physics
 2,297
Prime Numbers
 126
Single Threaded
 1,816
String Sorting
 46,256

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,531
Database Operations
 10,400
Memory Latency
 54
Memory Read Cached
 22,170
Memory Read Uncached
 10,239
Memory Write
 9,799

Stress-ng Raw Scores

Stress-ng Relative Multicore Performance

LLM Inference Speed for Prompt Processing

LLM Inference Speed for Text Generation

Static Web Server

Redis

Alternatives

Servers of the Same Family

INSTANCEvCPUsMEMORYAccelerators
g4dn.xlarge 416 GiB1
g4dn.2xlarge 832 GiB1
g4dn.4xlarge 1664 GiB1
g4dn.12xlarge 48192 GiB4
g4dn.16xlarge 64256 GiB1
g4dn.metal 96384 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

g4dn.8xlarge FAQs