g5.4xlarge by Amazon Web Services

g5.4xlarge is a Graphics intensive Gen5 4xlarge server offered by Amazon Web Services with 16 vCPUs, 64 GiB of memory and 600 GB of storage. The pricing starts at 0.3079 USD per hour.
16 vCPU
64 GiB Memory
600 GB Storage
1 GPU
Spare SCore
6,147
(All-cores)
767
(Single-core)

Specifications

Server Metadata

Vendor ID
aws
Name
g5.4xlarge
Description
Graphics intensive Gen5 4xlarge
Family
g5
Hw Virt
Status
active
Observed At
2026-10-25T20:43:55.027788

Availability

REGION / IDSPOTONDEMAND
Oregon (US) / us-west-2
0.8945 USD/h 1.624 USD/h
Ohio (US) / us-east-2
0.8553 USD/h 1.624 USD/h
Northern Virgina (US) / us-east-1
1.10962 USD/h 1.624 USD/h
Aragón (ES) / eu-south-2
1.6807 USD/h 1.7112 USD/h
Stockholm (SE) / eu-north-1
0.49805 USD/h 1.7225 USD/h
Quebec (CA) / ca-central-1
1.8032 USD/h 1.8032 USD/h
Dublin (IE) / eu-west-1
1.1238 USD/h 1.8129 USD/h
Tel Aviv (IL) / il-central-1
1.833733 USD/h 1.9035 USD/h
Mumbai (IN) / ap-south-1
0.8685 USD/h 1.9501 USD/h
United Arab Emirates / me-central-1
0.8155 USD/h 1.9921 USD/h
Seoul (KR) / ap-northeast-2
0.9122 USD/h 1.9969 USD/h
Frankfurt (DE) / eu-central-1
0.882933 USD/h 2.0308 USD/h
London (GB) / eu-west-2
0.75745 USD/h 2.0615 USD/h
Paris (FR) / eu-west-3
1.501067 USD/h 2.0615 USD/h
Sydney (AU) / ap-southeast-2
1.83935 USD/h 2.1115 USD/h
Jakarta (ID) / ap-southeast-3
0.71835 USD/h 2.2729 USD/h
Tokyo (JP) / ap-northeast-1
1.18925 USD/h 2.3553 USD/h
Hong Kong (HK) / ap-east-1
0.62625 USD/h 2.5002 USD/h
Sao Paulo (BR) / sa-east-1
1.4171 USD/h 2.7605 USD/h

Processor

vCPUs
16
Hypervisor
nitro
CPU Allocation
Dedicated
CPU Cores
8
CPU Speed
3.3 GHz
CPU Architecture
x86_64
CPU Manufacturer
AMD
CPU Family
EPYC
CPU Model
7R32
CPU L1D Cache
32 KiB
CPU L1D Cache Total
256 KiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
256 KiB
CPU L2 Cache
512 KiB
CPU L2 Cache Total
4 MiB
CPU L3 Cache
16 MiB
CPU L3 Cache Total
32 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, ht, syscall, nx, mmxext, fxsr_opt, pdpe1gb, rdtscp, lm, constant_tsc, rep_good, nopl, nonstop_tsc, cpuid, extd_apicid, aperfmperf, tsc_known_freq, pni, pclmulqdq, ssse3, fma, cx16, sse4_1, sse4_2, movbe, popcnt, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, cmp_legacy, cr8_legacy, abm, sse4a, misalignsse, 3dnowprefetch, topoext, ssbd, ibrs, ibpb, stibp, vmmcall, fsgsbase, bmi1, avx2, smep, bmi2, rdseed, adx, smap, clflushopt, clwb, sha_ni, xsaveopt, xsavec, xgetbv1, clzero, xsaveerptr, rdpru, wbnoinvd, arat, npt, nrip_save, rdpid
Ecpus
8
Scalability
50

System Resources and Accelerators

MEMORY
Memory Amount
64 GiB
Memory Amount Actual
64 GiB
Memory Generation
DDR4
Memory Speed
3200 Mhz
ACCELERATORS
Accelerator Count
1
Accelerator Memory Min
22 GiB
Accelerator Memory Total
22 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ampere
Accelerator Model
A10G
Accelerators
  • NVIDIA Ampere NVIDIA A10G (Memory amount: 23028, Firmware version: 590.48.01, BIOS version: 94.02.75.00.01, Clock rate: 1710)
STORAGE
Storage Size
600 GB
Storage Type
nvme ssd
Storages
  • 600 GB nvme ssd
NETWORK
Network Speed Baseline
10 Gbps
Network Speed Max
25 Gbps
Network Storage Speed Baseline
4.75 Gbps
Network Storage Speed Max
4.75 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 NVMe storage and an NVIDIA Ampere GPU, optimized for machine learning inference and graphics-intensive workloads.

GPU AcceleratedStorage & DatabaseGeneral Purpose

Amazon Web Services g5.4xlarge is a GPU-accelerated server powered by the Nitro hypervisor. It features an AMD EPYC 7R32 processor with 16 vCPUs, 64 GB of DDR4 memory, 600 GB of local NVMe SSD storage, and a single NVIDIA Ampere A10G GPU with 22 GB of VRAM. While synthetic CPU and memory bandwidth benchmarks place the instance in the average to weak categories, its machine learning capabilities are highly competitive. Specifically, the server achieves top-tier performance in LLM inference for small and medium models, and strong performance for large model prompt processing. This hardware profile makes the g5.4xlarge highly cost-efficient for GPU-driven workloads, including low-to-medium parameter LLM deployment, ray tracing, and graphics rendering, whereas it presents tradeoffs for pure CPU-bound or high-memory-bandwidth applications.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

0.69Score
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) 1,311,632 ops/sec 1,919,289 ops/sec 0.68 50.00% 50.00% -18.00%
Redis RPS (pipeline=16, SET) 8,635,272 ops/sec 11,728,058 ops/sec 0.72 20.00% 20.00% -6.40%
PassMark Memory Mark (composite) 2,668 2,416 1.1 10.00% 10.00% +0.96%
Memory bandwidth (read, 16 MB ~ L3) 37,793 MB/sec 108,466 MB/sec 0.35 10.00% 10.00% -10.00%
PassMark single-thread CPU 2,133 Mops/s 2,365 Mops/s 0.88 10.00% 10.00% -1.30%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
11.0 GB/s
Node GPU-to-host bandwidth
13.2 GB/s
Node host-to-GPU bandwidth
13.5 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
13.2 GB/s
Per-GPU host-to-GPU bandwidth
13.5 GB/s
Per-GPU host-device latency
923 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Runtime Version
 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: 1,340

Geekbench Multi-Core

Score: 8,024

Passmark CPU Scores

BENCHMARKSCORE
Mark
 19,545
Compression
 237,423
Encryption
 17,037
Extended Instructions
 14,785
Floating Point Maths
 32,192
Integer Maths
 53,922
Physics
 1,750
Prime Numbers
 84
Single Threaded
 2,133
String Sorting
 33,351

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,668
Database Operations
 5,612
Memory Latency
 51
Memory Read Cached
 24,688
Memory Read Uncached
 12,822
Memory Write
 13,221

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 293.26 350.95 1.20×
bitops
 120,712 1,219,232 10.10×
branch
 153,368 2,128,476 13.88×
funccall
 501.52 5,419 10.80×
funcret
 14,067 170,736 12.14×
hash
 131,669 1,642,939 12.48×
intmath
 19,638 170,865 8.70×
nop
 236,914 1,912,144 8.07×
prime
 347,158 3,663,014 10.55×
rotate
 59,895 498,104 8.32×
str
 613,814 6,485,693 10.57×
wcs
 475,127 4,301,806 9.05×
CryptographySingleMultiScaling
crypt
 2,814 24,531 8.72×
CompressionSingleMultiScaling
jpeg
 668.47 6,998 10.47×
zlib
 51.02 682.2 13.37×
HypervisorSingleMultiScaling
clock
 19,611 165,777 8.45×
clone
 177.75 1,355 7.62×
context
 3,382 46,019 13.61×
cpu-sched
 65.34 1,021 15.62×
eventfd
 21,979 820,683 37.34×
fault
 8,352 85,866 10.28×
fork
 3,073 28,218 9.18×
getrandom
 36,388 390,868 10.74×
hrtimers
 340,619 1,367,310 4.01×
mprotect
 685,613 2,780,797 4.06×
mtx
 18,493,484 641,425,034 34.68×
munmap
 2,179 21,259 9.75×
null
 417,659 5,377,698 12.88×
pipe
 1,714,162 5,520,168 3.22×
priv-instr
 425,988 5,513,192 12.94×
resched
 3,190,174 948,288 0.30×
switch
 29,796 2,145,691 72.01×
timermix
 139,011 1,824,076 13.12×
tlb-shootdown
 9,407 36,074 3.83×
tsc
 654,695 10,179,699 15.55×
usersyscall
 277,622 3,832,506 13.80×
vdso
 78,982,701 897,736,224 11.37×
vfork
 16,341 123,088 7.53×
x86syscall
 5,284,640 68,712,549 13.00×
yield
 8,672,870 7,967,720 0.92×
zero
 2,101,826 26,684,170 12.70×
CPU float/SIMDSingleMultiScaling
eigen
 12,420 117,689 9.48×
expmath
 2,773 35,003 12.62×
fma
 740,293 9,345,841 12.62×
fp
 208.6 2,467 11.83×
fractal
 6.4 80.41 12.56×
hyperbolic
 3,642 43,757 12.01×
logmath
 4,110 47,901 11.65×
matrix
 2,753 39,879 14.48×
matrix-3d
 183.26 1,459 7.96×
monte-carlo
 183.44 2,408 13.13×
mpfr
 1,278 13,259 10.37×
powmath
 1,338 17,309 12.94×
trig
 1,559 19,027 12.20×
veccmp
 17,135 166,011 9.69×
vecfp
 1,385 16,853 12.16×
vecmath
 3,372 43,477 12.89×
vecshuf
 483.58 4,793 9.91×
vecwide
 26,578 253,591 9.54×
AlgorithmsSingleMultiScaling
bitonicsort
 6.8 67.6 9.94×
bsearch
 227.59 2,307 10.14×
bubblesort
 0.8 6.4 8.00×
fibsearch
 259.14 2,677 10.33×
heapsort
 8.6 97.6 11.35×
hsearch
 5,375 51,499 9.58×
insertionsort
 6.6 61.4 9.30×
malloc
 2,652,116 30,211,057 11.39×
mcontend
 127.1 2,639 20.76×
mergesort
 25.8 313.99 12.17×
qsort
 16.6 200.59 12.08×
skiplist
 5,526 70,340 12.73×
tree
 1.8 9.59 5.33×
MemorySingleMultiScaling
cache
 485,409 971,406 2.00×
cachehammer
 295,440 3,951,230 13.37×
cacheline
 50.93 853.68 16.76×
l1cache
 198.59 1,541 7.76×
memcpy
 355.9 2,961 8.32×
memrate
 1.95 6.07 3.11×
memthrash
 98.82 382 3.87×
ptr-chase
 29,312,793 134,833,184 4.60×
spinmem
 237.19 788.34 3.32×
stream
 176.6 5,368 30.40×
vm
 57,743 292,986 5.07×

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
g5.xlarge 416 GiB1
g5.2xlarge 832 GiB1
g5.8xlarge 32128 GiB1
g5.12xlarge 48192 GiB4
g5.16xlarge 64256 GiB1
g5.24xlarge 96384 GiB4
g5.48xlarge 192768 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

g5.4xlarge FAQs