g5.16xlarge by Amazon Web Services

g5.16xlarge is a Graphics intensive Gen5 16xlarge server offered by Amazon Web Services with 64 vCPUs, 256 GiB of memory and 1.9 TB of storage. The pricing starts at 0.4801 USD per hour.
64 vCPU
256 GiB Memory
1.9 TB Storage
1 GPU
Spare SCore
24,581
(All-cores)
767
(Single-core)

Specifications

Server Metadata

Vendor ID
aws
Name
g5.16xlarge
Description
Graphics intensive Gen5 16xlarge
Family
g5
Hw Virt
Average Time To Start
11
Status
active
Observed At
2026-10-24T08:16:31.033830

Availability

REGION / IDSPOTONDEMAND
Oregon (US) / us-west-2
1.236633 USD/h 4.096 USD/h
Northern Virgina (US) / us-east-1
1.686 USD/h 4.096 USD/h
Ohio (US) / us-east-2
1.922133 USD/h 4.096 USD/h
Aragón (ES) / eu-south-2
4.27085 USD/h 4.3159 USD/h
Stockholm (SE) / eu-north-1
0.5926 USD/h 4.3444 USD/h
Quebec (CA) / ca-central-1
3.1812 USD/h 4.5479 USD/h
Dublin (IE) / eu-west-1
1.500067 USD/h 4.5724 USD/h
Tel Aviv (IL) / il-central-1
1.060833 USD/h 4.801 USD/h
Mumbai (IN) / ap-south-1
1.54895 USD/h 4.9185 USD/h
United Arab Emirates / me-central-1
2.0568 USD/h 5.0243 USD/h
Seoul (KR) / ap-northeast-2
2.271367 USD/h 5.0365 USD/h
Frankfurt (DE) / eu-central-1
3.222667 USD/h 5.122 USD/h
Paris (FR) / eu-west-3
2.9119 USD/h 5.1994 USD/h
London (GB) / eu-west-2
3.0619 USD/h 5.1994 USD/h
Sydney (AU) / ap-southeast-2
3.445 USD/h 5.3256 USD/h
Jakarta (ID) / ap-southeast-3
1.5926 USD/h 5.7328 USD/h
Tokyo (JP) / ap-northeast-1
3.36785 USD/h 5.9404 USD/h
Hong Kong (HK) / ap-east-1
3.1544 USD/h 6.306 USD/h
Sao Paulo (BR) / sa-east-1
5.099 USD/h 6.9624 USD/h

Processor

vCPUs
64
Hypervisor
nitro
CPU Allocation
Dedicated
CPU Cores
32
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
1 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
1 MiB
CPU L2 Cache
512 KiB
CPU L2 Cache Total
16 MiB
CPU L3 Cache
16 MiB
CPU L3 Cache Total
128 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
32.1
Scalability
50.16

System Resources and Accelerators

MEMORY
Memory Amount
256 GiB
Memory Amount Actual
256 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
1900 GB
Storage Type
nvme ssd
Storages
  • 1900 GB nvme ssd
NETWORK
Network Speed Baseline
25 Gbps
Network Speed Max
25 Gbps
Network Storage Speed Baseline
16 Gbps
Network Storage Speed Max
16 Gbps
Inbound Traffic
0 GB/month
Outbound Traffic
0 GB/month
IPv4
0

CPU and System Topology

Server Description

An accelerated virtual server combining dedicated AMD EPYC processors, local NVMe storage, and NVIDIA Ampere graphics for demanding machine learning workloads.

GPU AcceleratedGeneral Purpose

Amazon Web Services g5.16xlarge is a graphics-intensive instance featuring 64 dedicated vCPUs from an AMD EPYC 7R32 processor running at 3.3 GHz, 256.0 GB of DDR4 memory, and a baseline network bandwidth of 25 Gbps. It includes a bundled NVIDIA Ampere A10G GPU with 22 GB of VRAM and 1900 GB of local NVMe SSD storage. Performance benchmarks show a contrast between poor single-core CPU metrics and top-tier LLM inference speeds for small and medium models. While PassMark CPU and memory scores are strong, text generation for large 70B models is weak. This configuration offers qualitative cost efficiency by combining dedicated compute, high-speed local storage, and GPU acceleration on the Nitro hypervisor. It is best suited for machine learning inference, ray tracing, and workloads requiring high-density GPU and storage resources.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

2.1Score
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) 5,141,109 ops/sec 1,919,289 ops/sec 2.7 50.00% 50.00% +64.00%
Redis RPS (pipeline=16, SET) 34,564,529 ops/sec 11,728,058 ops/sec 2.9 20.00% 20.00% +24.00%
PassMark Memory Mark (composite) 2,917 2,416 1.2 10.00% 10.00% +1.80%
Memory bandwidth (read, 16 MB ~ L3) 119,992 MB/sec 108,466 MB/sec 1.1 10.00% 10.00% +0.96%
PassMark single-thread CPU 2,135 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
917 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: 1,332

Geekbench Multi-Core

Score: 14,069

Passmark CPU Scores

BENCHMARKSCORE
Mark
 55,615
Compression
 953,024
Encryption
 68,193
Extended Instructions
 56,952
Floating Point Maths
 128,866
Integer Maths
 216,068
Physics
 5,371
Prime Numbers
 296
Single Threaded
 2,135
String Sorting
 126,195

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,917
Database Operations
 21,347
Memory Latency
 51
Memory Read Cached
 24,632
Memory Read Uncached
 12,768
Memory Write
 13,485

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 414.39 339.06 0.82×
bitops
 120,296 4,867,019 40.46×
branch
 152,541 8,500,430 55.73×
funccall
 501.38 21,391 42.66×
funcret
 14,029 660,934 47.11×
hash
 131,506 6,540,594 49.74×
intmath
 19,043 675,603 35.48×
nop
 235,911 7,640,520 32.39×
prime
 345,155 14,605,719 42.32×
rotate
 59,782 1,983,190 33.17×
str
 625,967 24,917,346 39.81×
wcs
 463,033 17,447,455 37.68×
CryptographySingleMultiScaling
crypt
 2,809 95,673 34.06×
CompressionSingleMultiScaling
jpeg
 663.23 26,599 40.11×
zlib
 50.91 3,018 59.29×
HypervisorSingleMultiScaling
clock
 15,486 279,453 18.05×
clone
 138.5 4,596 33.19×
context
 3,379 182,056 53.87×
cpu-sched
 69.28 4,123 59.52×
eventfd
 23,007 2,922,939 127.05×
fault
 8,220 212,068 25.80×
fork
 2,790 54,522 19.54×
getrandom
 35,743 1,559,296 43.63×
hrtimers
 884,545 5,003,229 5.66×
mprotect
 503,445 10,387,592 20.63×
mtx
 29,270,498 2,191,436,369 74.87×
munmap
 1,984 55,129 27.78×
null
 418,182 21,322,497 50.99×
pipe
 1,733,696 24,225,110 13.97×
priv-instr
 416,641 16,879,757 40.51×
resched
 5,329,705 3,189,264 0.60×
switch
 53,373 8,548,367 160.16×
timermix
 142,290 7,381,447 51.88×
tlb-shootdown
 8,510 102,334 12.02×
tsc
 648,756 40,415,832 62.30×
usersyscall
 288,619 13,846,347 47.97×
vdso
 77,999,879 3,160,543,255 40.52×
vfork
 14,286 148,899 10.42×
x86syscall
 5,233,090 266,568,769 50.94×
yield
 29,754,987 32,161,337 1.08×
zero
 2,117,846 105,703,118 49.91×
CPU float/SIMDSingleMultiScaling
eigen
 12,400 464,858 37.49×
expmath
 2,773 139,521 50.31×
fma
 736,239 37,151,032 50.46×
fp
 207.89 9,753 46.92×
fractal
 6.4 320.04 50.01×
hyperbolic
 3,639 174,575 47.97×
logmath
 4,095 190,725 46.58×
matrix
 2,737 153,273 56.00×
matrix-3d
 145.45 3,728 25.63×
monte-carlo
 183.07 9,598 52.43×
mpfr
 1,273 52,725 41.43×
powmath
 1,336 69,198 51.80×
trig
 1,555 76,125 48.96×
veccmp
 17,103 658,157 38.48×
vecfp
 1,385 65,740 47.47×
vecmath
 3,368 173,563 51.54×
vecshuf
 483.59 19,135 39.57×
vecwide
 26,538 972,191 36.63×
AlgorithmsSingleMultiScaling
bitonicsort
 6.8 274.99 40.44×
bsearch
 227.83 9,256 40.63×
bubblesort
 0.8 25.6 32.00×
fibsearch
 257.77 10,546 40.91×
heapsort
 8.6 388.98 45.23×
hsearch
 5,390 209,108 38.79×
insertionsort
 6.6 242 36.67×
malloc
 2,591,378 76,588,123 29.55×
mcontend
 85.88 9,336 108.71×
mergesort
 25.6 1,240 48.44×
qsort
 16.6 799.57 48.17×
skiplist
 5,509 281,229 51.05×
tree
 1.8 38.31 21.28×
MemorySingleMultiScaling
cache
 464,796 891,518 1.92×
cachehammer
 259,900 12,239,305 47.09×
cacheline
 52.96 3,520 66.46×
l1cache
 197.1 6,105 30.98×
memcpy
 357.55 11,819 33.05×
memrate
 1.95 11.57 5.93×
memthrash
 202.06 1,005 4.97×
ptr-chase
 29,510,623 449,463,728 15.23×
spinmem
 245.39 3,413 13.91×
stream
 178.96 62,384 348.59×
vm
 53,351 300,002 5.62×

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
g5.4xlarge 1664 GiB1
g5.8xlarge 32128 GiB1
g5.12xlarge 48192 GiB4
g5.24xlarge 96384 GiB4
g5.48xlarge 192768 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

g5.16xlarge FAQs