g5g.16xlarge by Amazon Web Services

g5g.16xlarge is a Graphics intensive [AWS Graviton processors] Gen5 16xlarge server offered by Amazon Web Services with 64 vCPUs, 128 GiB of memory and 0 GB of storage. The pricing starts at 0.3062 USD per hour.
64 vCPU
128 GiB Memory
2 GPU
Spare SCore
97,057
(All-cores)
1,519
(Single-core)

Specifications

Server Metadata

Vendor ID
aws
Name
g5g.16xlarge
Description
Graphics intensive [AWS Graviton processors] Gen5 16xlarge
Family
g5g
Hw Virt
Average Time To Start
11.33
Status
active
Observed At
2026-10-25T20:43:55.045442

Availability

REGION / IDSPOTONDEMAND
Oregon (US) / us-west-2
0.801233 USD/h 2.744 USD/h
Northern Virgina (US) / us-east-1
0.7158 USD/h 2.744 USD/h
Aragón (ES) / eu-south-2
0.3062 USD/h 3.0622 USD/h
Seoul (KR) / ap-northeast-2
1.5188 USD/h 3.3752 USD/h
Frankfurt (DE) / eu-central-1
1.6149 USD/h 3.4326 USD/h
Tokyo (JP) / ap-northeast-1
1.66725 USD/h 3.7039 USD/h
Singapore (SG) / ap-southeast-1
0.7497 USD/h 3.8395 USD/h

Processor

vCPUs
64
Hypervisor
nitro
CPU Allocation
Dedicated
CPU Cores
64
CPU Speed
2.5 GHz
CPU Architecture
arm64
CPU Manufacturer
AWS
CPU Family
ARMv8
CPU Model
Neoverse-N1
CPU L1D Cache
64 KiB
CPU L1D Cache Total
4 MiB
CPU L1I Cache
64 KiB
CPU L1I Cache Total
4 MiB
CPU L2 Cache
1 MiB
CPU L2 Cache Total
64 MiB
CPU L3 Cache
32 MiB
CPU L3 Cache Total
32 MiB
CPU Flags
fp, asimd, evtstrm, aes, pmull, sha1, sha2, crc32, atomics, fphp, asimdhp, cpuid, asimdrdm, lrcpc, dcpop, asimddp, ssbs
Ecpus
63.9
Scalability
99.84

System Resources and Accelerators

MEMORY
Memory Amount
128 GiB
Memory Amount Actual
126 GiB
ACCELERATORS
Accelerator Count
2
Accelerator Memory Min
16 GiB
Accelerator Memory Total
32 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Turing
Accelerator Model
T4G
Accelerators
  • NVIDIA Turing NVIDIA T4G (Memory amount: 15360, Firmware version: 590.48.01, BIOS version: 90.04.AF.00.02, Clock rate: 1590)
  • NVIDIA Turing NVIDIA T4G (Memory amount: 15360, Firmware version: 590.48.01, BIOS version: 90.04.AF.00.02, Clock rate: 1590)
STORAGE
Storage Size
0 GB
Storages
    NETWORK
    Network Speed Baseline
    25 Gbps
    Network Speed Max
    25 Gbps
    Network Storage Speed Baseline
    19 Gbps
    Network Storage Speed Max
    19 Gbps
    Inbound Traffic
    0 GB/month
    Outbound Traffic
    0 GB/month
    IPv4
    0

    CPU and System Topology

    Server Description

    An ARM64 architecture server combining sixty-four dedicated cores with dual graphics accelerators for multi-threaded and GPU-accelerated workloads.

    GPU AcceleratedCompute Optimized

    Amazon Web Services g5g.16xlarge is an ARM64 server instance featuring 64 dedicated AWS Graviton Neoverse-N1 cores running at 2.5 GHz, paired with 128.0 GB of system memory and two NVIDIA Turing T4G GPUs providing 32 GB of VRAM. Operating on the Nitro hypervisor, it delivers a baseline network speed of 25 Gbps without local storage. Benchmark results show strong multi-core CPU performance and high memory bandwidth for small cached blocks, alongside strong Redis SET operations and static web serving throughput. However, single-core processing and single-threaded compression are weak. LLM inference performance is average for prompt processing on small to medium models but poor to weak for text generation and large models. This instance is suited for multi-threaded ARM applications, graphics-intensive tasks, and web serving workloads.

    Economics

    Average Price per Region

    Prices per Zone

    Lowest Prices

    Performance

    Workload Profiles

    2.3Score
    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) 7,215,330 ops/sec 1,919,289 ops/sec 3.8 50.00% 50.00% +95.00%
    Redis RPS (pipeline=16, SET) 42,322,181 ops/sec 11,728,058 ops/sec 3.5 20.00% 20.00% +28.00%
    PassMark Memory Mark (composite) 1,842 2,416 0.75 10.00% 10.00% -2.80%
    Memory bandwidth (read, 16 MB ~ L3) 171,450 MB/sec 108,466 MB/sec 1.5 10.00% 10.00% +4.10%
    PassMark single-thread CPU 1,111 Mops/s 2,365 Mops/s 0.46 10.00% 10.00% -7.50%

    GPU Bandwidth and Latency

    Node (all GPUs)
    Node baseline directional bandwidth
    		11.3 GB/s
    Node GPU-to-host bandwidth
    		13.2 GB/s
    Node host-to-GPU bandwidth
    		13.2 GB/s
    Single GPU slot
    Per-GPU GPU-to-host bandwidth
    		6.6 GB/s
    Per-GPU host-to-GPU bandwidth
    		6.6 GB/s
    Per-GPU host-device latency
    		753 ns
    Benchmark Environment
    Driver Version
    		 535.183.01
    CUDA Runtime Version
    		 12.6
    nvbandwidth Version
    		 v0.5
    GPU Count
    		 2
    P2P Supported
    No
    Efficiency
    SM/CE copy efficiency
    		1.00

    Memory Bandwidth

    Compression

    OpenSSL

    Geekbench Single-Core

    Score: 973

    Geekbench Multi-Core

    Score: 13,890

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 28,013
    Compression
    		 304,927
    Encryption
    		 8,699
    Extended Instructions
    		 39,781
    Floating Point Maths
    		 191,572
    Integer Maths
    		 252,597
    Physics
    		 3,904
    Prime Numbers
    		 243
    Single Threaded
    		 1,111
    String Sorting
    		 175,019

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 1,842
    Database Operations
    		 20,423
    Memory Latency
    		 70
    Memory Read Cached
    		 9,487
    Memory Read Uncached
    		 9,201
    Memory Write
    		 12,573

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 122.68 40.26 0.33×
    bitops
    		 90,261 5,764,022 63.86×
    branch
    		 183,232 11,711,739 63.92×
    funccall
    		 1,112 70,956 63.80×
    funcret
    		 17,215 1,105,928 64.24×
    hash
    		 106,457 6,781,497 63.70×
    intmath
    		 26,424 1,665,662 63.04×
    nop
    		 145,080 9,176,411 63.25×
    prime
    		 283,773 18,093,336 63.76×
    rotate
    		 80,320 5,116,734 63.70×
    str
    		 428,642 27,345,673 63.80×
    wcs
    		 99,252 6,330,987 63.79×
    CryptographySingleMultiScaling
    crypt
    		 2,408 153,499 63.74×
    CompressionSingleMultiScaling
    jpeg
    		 438.15 24,807 56.62×
    zlib
    		 23.35 1,839 78.76×
    HypervisorSingleMultiScaling
    clock
    		 26,749 295,453 11.05×
    clone
    		 247.25 9,409 38.05×
    context
    		 2,192 139,590 63.68×
    cpu-sched
    		 60.97 4,080 66.92×
    eventfd
    		 52,526 1,567,439 29.84×
    fault
    		 11,038 307,644 27.87×
    fork
    		 3,537 53,601 15.16×
    getrandom
    		 40,061 2,547,836 63.60×
    hrtimers
    		 200,487 10,002,617 49.89×
    mprotect
    		 981,636 16,801,388 17.12×
    mtx
    		 7,404,827 2,246,978,206 303.45×
    munmap
    		 2,477 64,564 26.07×
    null
    		 383,635 24,462,315 63.76×
    pipe
    		 691,783 24,350,992 35.20×
    priv-instr
    		 607,306 9,902,584 16.31×
    resched
    		 4,552,880 7,055,184 1.55×
    switch
    		 82,756 16,746,480 202.36×
    timermix
    		 277,353 17,769,550 64.07×
    tlb-shootdown
    		 4,367 4,777 1.09×
    vdso
    		 41,281,236 2,625,195,790 63.59×
    vfork
    		 22,873 215,501 9.42×
    yield
    		 45,643,000 52,990,743 1.16×
    zero
    		 1,717,075 109,223,087 63.61×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 1,111 70,805 63.76×
    expmath
    		 199.62 12,699 63.61×
    fma
    		 533,090 33,966,943 63.72×
    fp
    		 86.4 5,493 63.57×
    fractal
    		 5.4 341.77 63.29×
    hyperbolic
    		 203.13 12,956 63.78×
    logmath
    		 244.24 15,571 63.75×
    matrix
    		 4,448 282,912 63.61×
    matrix-3d
    		 563.07 5,963 10.59×
    monte-carlo
    		 161.08 10,266 63.74×
    mpfr
    		 533.44 33,947 63.64×
    powmath
    		 146.78 9,359 63.76×
    trig
    		 274.17 17,464 63.70×
    veccmp
    		 14,296 908,099 63.52×
    vecfp
    		 620.76 39,557 63.72×
    vecmath
    		 2,569 163,763 63.75×
    vecshuf
    		 558.78 35,646 63.79×
    vecwide
    		 15,307 972,488 63.53×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 4.4 281.57 63.99×
    bsearch
    		 214.78 13,717 63.86×
    bubblesort
    		 0.6 38.4 64.00×
    fibsearch
    		 227.53 14,501 63.73×
    heapsort
    		 7 435.16 62.17×
    hsearch
    		 4,036 260,095 64.45×
    insertionsort
    		 3.8 242.2 63.74×
    malloc
    		 2,620,886 106,142,034 40.50×
    mcontend
    		 226.48 118,891 524.95×
    mergesort
    		 23.2 1,434 61.79×
    qsort
    		 16.4 1,035 63.10×
    skiplist
    		 4,445 282,715 63.61×
    tree
    		 1.2 38.3 31.92×
    MemorySingleMultiScaling
    cache
    		 360,836 701,527 1.94×
    cachehammer
    		 98,248 231,720 2.36×
    cacheline
    		 69.77 3,916 56.13×
    l1cache
    		 151.47 9,452 62.40×
    memcpy
    		 368.91 23,504 63.71×
    memrate
    		 3.32 35 10.54×
    memthrash
    		 249.45 537.6 2.16×
    ptr-chase
    		 23,131,080 547,529,109 23.67×
    spinmem
    		 462.98 820.18 1.77×
    stream
    		 81.86 58,097 709.71×
    vm
    		 55,542 2,686,266 48.36×

    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
    g5g.xlarge 48 GiB1
    g5g.2xlarge 816 GiB1
    g5g.4xlarge 1632 GiB1
    g5g.8xlarge 3264 GiB1
    g5g.metal 64128 GiB2

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    g5g.16xlarge FAQs