g2-standard-96 by Google Cloud Platform

g2-standard-96 is a Graphics Optimized: 8 NVIDIA L4 GPUs, 96 vCPUs, 384GB RAM server offered by Google Cloud Platform with 96 vCPUs, 384 GiB of memory and 0 GB of storage. The pricing starts at 3.4745 USD per hour.
96 vCPU
384 GiB Memory
8 GPU
Spare SCore
75,746
(All-cores)
1,201
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1600096
Name
g2-standard-96
Description
Graphics Optimized: 8 NVIDIA L4 GPUs, 96 vCPUs, 384GB RAM
Family
g2
Hw Virt
Average Time To Start
20
Status
active
Observed At
2026-10-25T09:08:14.057458

Availability

REGION / IDSPOTONDEMAND
Ashburn (US) / us-east4
4.2561 USD/h 7.9771 USD/h
Council Bluffs (US) / us-central1
4.7892 USD/h 8.0033 USD/h
The Dalles (US) / us-west1
4.7892 USD/h 8.0033 USD/h
Moncks Corner (US) / us-east1
4.7892 USD/h 8.0033 USD/h
Mumbai (IN) / asia-south1
4.2432 USD/h 8.3311 USD/h
Eemshaven (NL) / europe-west4
5.0466 USD/h 8.4119 USD/h
Montréal (CA) / northamerica-northeast1
4.9975 USD/h 8.8125 USD/h
Toronto (CA) / northamerica-northeast2
5.0093 USD/h 8.8125 USD/h
St. Ghislain (BE) / europe-west1
5.2876 USD/h 8.8125 USD/h
Las Vegas (US) / us-west4
5.3951 USD/h 9.0142 USD/h
London (GB) / europe-west2
5.4675 USD/h 9.1125 USD/h
Changhua County (TW) / asia-east1
5.5522 USD/h 9.2671 USD/h
Frankfurt (DE) / europe-west3
5.6313 USD/h 9.4322 USD/h
Jurong West (SG) / asia-southeast1
5.9249 USD/h 9.8737 USD/h
Seoul (KR) / asia-northeast3
3.4745 USD/h 10.2779 USD/h
Tokyo (JP) / asia-northeast1
5.2737 USD/h 10.2779 USD/h
Zurich (CH) / europe-west6
6.5131 USD/h 10.8829 USD/h
Dammam (SA) / me-central2
7.2251 USD/h 12.8053 USD/h

Processor

vCPUs
96
CPU Allocation
Dedicated
CPU Cores
48
CPU Speed
2.2 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU L1D Cache
32 KiB
CPU L1D Cache Total
2 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
2 MiB
CPU L2 Cache
1 MiB
CPU L2 Cache Total
48 MiB
CPU L3 Cache
39 MiB
CPU L3 Cache Total
77 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, tsc_known_freq, pni, pclmulqdq, ssse3, fma, cx16, pcid, sse4_1, sse4_2, x2apic, movbe, popcnt, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, abm, 3dnowprefetch, ssbd, ibrs, ibpb, stibp, ibrs_enhanced, fsgsbase, tsc_adjust, bmi1, hle, avx2, smep, bmi2, erms, invpcid, rtm, mpx, avx512f, avx512dq, rdseed, adx, smap, clflushopt, clwb, avx512cd, avx512bw, avx512vl, xsaveopt, xsavec, xgetbv1, xsaves, arat, avx512_vnni, md_clear, arch_capabilities
Ecpus
63
Scalability
65.62

System Resources and Accelerators

MEMORY
Memory Amount
384 GiB
Memory Amount Actual
384 GiB
ACCELERATORS
Accelerator Count
8
Accelerator Memory Min
24 GiB
Accelerator Memory Total
192 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ada Lovelace
Accelerator Model
L4
Accelerators
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
  • NVIDIA Ada Lovelace NVIDIA L4 (Memory amount: 23034, Firmware version: 590.48.01, BIOS version: 95.04.29.00.07, Clock rate: 2040)
STORAGE
Storage Size
0 GB
Storages
    NETWORK
    Inbound Traffic
    0 GB/month
    Outbound Traffic
    0 GB/month
    IPv4
    0

    CPU and System Topology

    Server Description

    A high-density GPU server featuring eight NVIDIA L4 accelerators and ninety-six vCPUs optimized for parallel computing and large language model inference.

    GPU AcceleratedCompute OptimizedGeneral Purpose

    Google Cloud Platform g2-standard-96 is a graphics-optimized server featuring eight NVIDIA L4 GPUs with 192 GB of total VRAM, 96 dedicated Intel Xeon vCPUs, and 384.0 GB of system memory. Built on the x86_64 architecture, this server does not include local storage. While single-core CPU performance and single-threaded tasks are relatively weak, the instance delivers strong multi-core capabilities and top-tier LLM inference speeds, processing up to 11350.23 tokens per second for small models. It also exhibits high memory bandwidth and strong Redis performance. Qualitatively, the high density of GPUs and dedicated vCPUs makes it a cost-efficient choice for parallelized, GPU-bound workloads rather than single-threaded applications. The g2-standard-96 is highly suited for LLM inference, multi-threaded data processing, and GPU-accelerated compute tasks.

    Economics

    Average Price per Region

    Prices per Zone

    Lowest Prices

    Performance

    Workload Profiles

    2.5Score
    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) 6,655,418 ops/sec 1,919,289 ops/sec 3.5 50.00% 50.00% +87.00%
    Redis RPS (pipeline=16, SET) 39,460,725 ops/sec 11,728,058 ops/sec 3.2 20.00% 20.00% +26.00%
    PassMark Memory Mark (composite) 2,675 2,416 1.1 10.00% 10.00% +0.96%
    Memory bandwidth (read, 16 MB ~ L3) 221,008 MB/sec 108,466 MB/sec 2 10.00% 10.00% +7.20%
    PassMark single-thread CPU 1,694 Mops/s 2,365 Mops/s 0.71 10.00% 10.00% -3.40%

    GPU Bandwidth and Latency

    Node (all GPUs)
    Node baseline directional bandwidth
    		25.8 GB/s
    Node GPU-to-host bandwidth
    		52.7 GB/s
    Node host-to-GPU bandwidth
    		49.5 GB/s
    Peer-to-peer
    GPU-to-GPU P2P duplex bandwidth
    		8.2 GB/s
    GPU-to-GPU P2P gather bandwidth
    		9.4 GB/s
    GPU-to-GPU P2P latency
    		946 ns
    GPU-to-GPU P2P bandwidth
    		10.1 GB/s
    Single GPU slot
    Per-GPU GPU-to-host bandwidth
    		13.2 GB/s
    Per-GPU host-to-GPU bandwidth
    		12.4 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
    		 8
    P2P Supported
    Yes
    Efficiency
    SM/CE copy efficiency
    		1.00

    Memory Bandwidth

    Compression

    OpenSSL

    Geekbench Single-Core

    Score: 1,072

    Geekbench Multi-Core

    Score: 11,560

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 49,105
    Compression
    		 1,032,296
    Encryption
    		 31,079
    Extended Instructions
    		 65,697
    Floating Point Maths
    		 136,123
    Integer Maths
    		 253,019
    Physics
    		 4,904
    Prime Numbers
    		 318
    Single Threaded
    		 1,694
    String Sorting
    		 115,764

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 2,675
    Database Operations
    		 20,576
    Memory Latency
    		 50
    Memory Read Cached
    		 20,747
    Memory Read Uncached
    		 10,667
    Memory Write
    		 10,022

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 198.28 113.04 0.57×
    bitops
    		 92,216 4,998,254 54.20×
    branch
    		 131,369 9,521,207 72.48×
    funccall
    		 463.15 21,639 46.72×
    funcret
    		 9,724 625,222 64.30×
    hash
    		 107,604 7,271,430 67.58×
    intmath
    		 19,000 1,340,476 70.55×
    nop
    		 163,257 6,763,402 41.43×
    prime
    		 271,793 13,768,510 50.66×
    rotate
    		 82,415 5,420,020 65.76×
    str
    		 638,050 28,849,441 45.22×
    wcs
    		 393,577 18,190,542 46.22×
    CryptographySingleMultiScaling
    crypt
    		 2,078 93,942 45.21×
    CompressionSingleMultiScaling
    jpeg
    		 499.13 25,962 52.01×
    zlib
    		 30.15 3,109 103.11×
    HypervisorSingleMultiScaling
    clock
    		 21,992 173,947 7.91×
    clone
    		 300.63 7,405 24.63×
    context
    		 2,244 114,855 51.19×
    cpu-sched
    		 72.77 6,076 83.50×
    eventfd
    		 46,443 1,797,487 38.70×
    fault
    		 9,462 143,098 15.12×
    fork
    		 3,410 46,415 13.61×
    getrandom
    		 38,207 2,137,655 55.95×
    hrtimers
    		 1,083,128 8,925,685 8.24×
    mprotect
    		 604,785 11,404,392 18.86×
    mtx
    		 8,711,250 1,810,095,797 207.79×
    munmap
    		 2,271 53,812 23.70×
    null
    		 253,234 11,132,400 43.96×
    pipe
    		 813,735 19,942,208 24.51×
    priv-instr
    		 482,697 9,952,553 20.62×
    resched
    		 4,187,829 6,011,707 1.44×
    switch
    		 76,826 12,744,132 165.88×
    timermix
    		 164,302 10,948,520 66.64×
    tlb-shootdown
    		 5,999 72,767 12.13×
    tsc
    		 850,505 56,651,986 66.61×
    usersyscall
    		 332,897 8,729,901 26.22×
    vdso
    		 70,510,681 4,345,706,238 61.63×
    vfork
    		 22,018 115,331 5.24×
    x86syscall
    		 2,898,981 125,559,662 43.31×
    yield
    		 43,254,758 48,902,795 1.13×
    zero
    		 1,706,301 86,314,750 50.59×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 13,486 762,252 56.52×
    expmath
    		 2,279 157,431 69.08×
    fma
    		 509,894 38,421,318 75.35×
    fp
    		 104.76 6,086 58.09×
    fractal
    		 4.2 279.32 66.50×
    hyperbolic
    		 2,381 171,107 71.86×
    logmath
    		 3,862 259,440 67.17×
    matrix
    		 2,597 188,684 72.66×
    matrix-3d
    		 618.51 6,403 10.35×
    monte-carlo
    		 118.64 6,125 51.63×
    mpfr
    		 1,033 49,291 47.71×
    powmath
    		 1,350 92,493 68.51×
    trig
    		 1,456 83,280 57.22×
    veccmp
    		 13,278 605,307 45.59×
    vecfp
    		 1,153 67,616 58.66×
    vecmath
    		 1,864 132,698 71.18×
    vecshuf
    		 481.57 24,302 50.46×
    vecwide
    		 21,010 1,069,327 50.90×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 6.2 308.99 49.84×
    bsearch
    		 205.14 10,723 52.27×
    bubblesort
    		 0.6 19.2 32.00×
    fibsearch
    		 208.13 11,453 55.03×
    heapsort
    		 7.4 393.99 53.24×
    hsearch
    		 4,189 236,314 56.41×
    insertionsort
    		 7 312.2 44.60×
    malloc
    		 2,037,084 55,294,577 27.14×
    mcontend
    		 67.06 17,412 259.65×
    mergesort
    		 19.8 1,128 56.96×
    qsort
    		 12.6 642.58 51.00×
    skiplist
    		 4,160 268,369 64.52×
    tree
    		 1.2 38.38 31.98×
    MemorySingleMultiScaling
    cache
    		 276,337 1,706,375 6.17×
    cachehammer
    		 356,543 13,878,225 38.92×
    cacheline
    		 16.16 1,397 86.45×
    l1cache
    		 193.31 9,402 48.64×
    memcpy
    		 296.34 14,042 47.39×
    memrate
    		 1.17 14.87 12.71×
    memthrash
    		 285.4 1,288 4.51×
    ptr-chase
    		 30,842,278 843,298,951 27.34×
    spinmem
    		 197.99 3,495 17.65×
    stream
    		 13.34 23,902 1791.76×
    vm
    		 53,168 534,321 10.05×

    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
    g2-standard-4 416 GiB1
    g2-standard-8 832 GiB1
    g2-standard-12 1248 GiB1
    g2-standard-16 1664 GiB1
    g2-standard-24 2496 GiB2
    g2-standard-32 32128 GiB1
    g2-standard-48 48192 GiB4

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    g2-standard-96 FAQs