a2-megagpu-16g by Google Cloud Platform

a2-megagpu-16g is a Accelerator Optimized: 16 NVIDIA Tesla A100 GPUs, 96 vCPUs, 1360GB RAM server offered by Google Cloud Platform with 96 vCPUs, 1360 GiB of memory and 0 GB of storage. The pricing starts at 26.2846 USD per hour.
96 vCPU
1360 GiB Memory
16 GPU
Spare SCore
75,766
(All-cores)
1,201
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1001096
Name
a2-megagpu-16g
Description
Accelerator Optimized: 16 NVIDIA Tesla A100 GPUs, 96 vCPUs, 1360GB RAM
Family
a2
Hw Virt
Average Time To Start
88
Status
active
Observed At
2026-10-16T14:05:04.549105

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
30.1826 USD/h 55.7395 USD/h
Eemshaven (NL) / europe-west4
26.2846 USD/h 56.6266 USD/h
Jurong West (SG) / asia-southeast1
36.1955 USD/h 60.4762 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
63.1
Scalability
65.73

System Resources and Accelerators

MEMORY
Memory Amount
1 TB
Memory Amount Actual
1 TB
ACCELERATORS
Accelerator Count
16
Accelerator Memory Min
40 GiB
Accelerator Memory Total
640 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ampere
Accelerator Model
A100
Accelerators
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.45.00.03, Clock rate: 1410)
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

    An accelerator-optimized platform combining sixteen Ampere GPUs and substantial memory capacity for highly parallelized, multi-threaded computing workloads.

    GPU AcceleratedMemory OptimizedCompute Optimized

    Google Cloud Platform a2-megagpu-16g is an accelerator-optimized server featuring 16 NVIDIA Ampere A100 GPUs with 640 GB of total VRAM, 96 vCPUs on an Intel Xeon architecture, and 1360.0 GB of system memory. It contains no local storage. Benchmarks show a significant performance divergence: single-core CPU tasks and single-threaded compression rank in the bottom 25%, whereas multi-core CPU performance and cached/uncached memory read bandwidth rank in the top 25%. Other metrics, such as database operations and static web serving, perform in the average range. Qualitatively, this server offers high resource density for GPU-heavy tasks, though its single-core performance and lack of local storage present clear architectural tradeoffs. It is designed for highly parallelized workloads, including deep learning, large-scale simulations, and GPU-accelerated data science.

    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) 6,462,670 ops/sec 1,936,298 ops/sec 3.4 50.00% 50.00% +84.00%
    Redis RPS (pipeline=16, SET) 38,061,426 ops/sec 11,797,257 ops/sec 3.2 20.00% 20.00% +26.00%
    PassMark Memory Mark (composite) 2,214 2,407 0.92 10.00% 10.00% -0.83%
    Memory bandwidth (read, 16 MB ~ L3) 167,479 MB/sec 108,490 MB/sec 1.5 10.00% 10.00% +4.10%
    PassMark single-thread CPU 1,685 Mops/s 2,365 Mops/s 0.71 10.00% 10.00% -3.40%

    Memory Bandwidth

    Compression

    OpenSSL

    Geekbench Single-Core

    Score: 1,046

    Geekbench Multi-Core

    Score: 11,593

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 48,596
    Compression
    		 1,023,301
    Encryption
    		 30,720
    Extended Instructions
    		 64,981
    Floating Point Maths
    		 136,109
    Integer Maths
    		 253,182
    Physics
    		 4,744
    Prime Numbers
    		 288
    Single Threaded
    		 1,685
    String Sorting
    		 115,178

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 2,214
    Database Operations
    		 19,695
    Memory Latency
    		 61
    Memory Read Cached
    		 20,871
    Memory Read Uncached
    		 7,770
    Memory Write
    		 7,671

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 252.61 116.85 0.46×
    bitops
    		 104,432 5,062,199 48.47×
    branch
    		 148,659 9,532,760 64.12×
    funccall
    		 524.92 21,861 41.65×
    funcret
    		 10,943 632,898 57.84×
    hash
    		 121,635 7,392,185 60.77×
    intmath
    		 21,526 1,356,205 63.00×
    nop
    		 184,973 6,758,274 36.54×
    prime
    		 304,979 13,942,414 45.72×
    rotate
    		 93,284 5,415,288 58.05×
    str
    		 712,598 29,951,861 42.03×
    wcs
    		 467,772 18,969,066 40.55×
    CryptographySingleMultiScaling
    crypt
    		 2,359 95,497 40.49×
    CompressionSingleMultiScaling
    jpeg
    		 561.88 26,305 46.82×
    zlib
    		 55.93 3,088 55.21×
    HypervisorSingleMultiScaling
    clock
    		 23,829 167,254 7.02×
    clone
    		 303.59 6,440 21.21×
    context
    		 2,515 115,449 45.90×
    cpu-sched
    		 54.24 5,874 108.29×
    eventfd
    		 51,637 1,820,170 35.25×
    fault
    		 10,644 137,786 12.94×
    fork
    		 3,210 43,147 13.44×
    getrandom
    		 43,228 2,169,499 50.19×
    hrtimers
    		 333,734 7,587,594 22.74×
    mprotect
    		 605,091 11,585,213 19.15×
    mtx
    		 10,363,059 1,732,175,542 167.15×
    munmap
    		 2,210 51,697 23.39×
    null
    		 274,574 11,195,289 40.77×
    pipe
    		 836,791 21,780,143 26.03×
    priv-instr
    		 543,001 9,577,004 17.64×
    resched
    		 4,087,869 7,029,458 1.72×
    switch
    		 85,481 13,220,283 154.66×
    timermix
    		 184,600 10,884,832 58.96×
    tlb-shootdown
    		 8,864 74,803 8.44×
    tsc
    		 963,558 56,707,008 58.85×
    usersyscall
    		 371,870 8,531,235 22.94×
    vdso
    		 80,104,310 4,466,583,107 55.76×
    vfork
    		 20,819 97,741 4.69×
    x86syscall
    		 3,239,580 126,120,792 38.93×
    yield
    		 44,992,588 47,302,403 1.05×
    zero
    		 1,953,746 87,079,079 44.57×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 15,606 772,184 49.48×
    expmath
    		 2,574 159,866 62.11×
    fma
    		 584,278 39,534,805 67.66×
    fp
    		 121.62 6,102 50.17×
    fractal
    		 4.8 283.21 59.00×
    hyperbolic
    		 2,695 173,431 64.35×
    logmath
    		 4,375 261,182 59.69×
    matrix
    		 2,947 191,034 64.82×
    matrix-3d
    		 639.37 5,904 9.23×
    monte-carlo
    		 139.07 6,133 44.10×
    mpfr
    		 1,171 50,025 42.73×
    powmath
    		 1,526 94,500 61.94×
    trig
    		 1,651 84,531 51.19×
    veccmp
    		 15,019 616,531 41.05×
    vecfp
    		 1,306 68,922 52.77×
    vecmath
    		 2,117 134,280 63.43×
    vecshuf
    		 544.45 24,652 45.28×
    vecwide
    		 24,427 1,084,148 44.38×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 7 310.39 44.34×
    bsearch
    		 233.4 11,132 47.69×
    bubblesort
    		 0.6 19.2 32.00×
    fibsearch
    		 236.36 11,485 48.59×
    heapsort
    		 8.4 394.79 47.00×
    hsearch
    		 4,738 240,618 50.79×
    insertionsort
    		 8 307.4 38.42×
    malloc
    		 2,091,890 60,479,117 28.91×
    mcontend
    		 69.05 17,004 246.25×
    mergesort
    		 22.4 1,128 50.36×
    qsort
    		 14.4 643.58 44.69×
    skiplist
    		 4,715 269,452 57.15×
    tree
    		 1.4 38.37 27.41×
    MemorySingleMultiScaling
    cache
    		 191,799 1,679,128 8.75×
    cachehammer
    		 329,508 12,980,449 39.39×
    cacheline
    		 15.37 1,387 90.22×
    l1cache
    		 218.96 9,415 43.00×
    memcpy
    		 336.52 14,263 42.39×
    memrate
    		 1.17 14.84 12.68×
    memthrash
    		 286.62 995.52 3.47×
    ptr-chase
    		 32,501,273 759,375,984 23.36×
    spinmem
    		 208.99 3,464 16.58×
    stream
    		 13.11 20,930 1596.50×
    vm
    		 56,501 360,374 6.38×

    Stress-ng Raw Scores

    Stress-ng Relative Multicore Performance

    Static Web Server

    Redis

    Alternatives

    Servers of the Same Family

    INSTANCEvCPUsMEMORYAccelerators
    a2-highgpu-1g 1285 GiB1
    a2-ultragpu-1g 12170 GiB1
    a2-highgpu-2g 24170 GiB2
    a2-ultragpu-2g 24340 GiB2
    a2-highgpu-4g 48340 GiB4
    a2-ultragpu-4g 48680 GiB4
    a2-highgpu-8g 96680 GiB8

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    a2-megagpu-16g FAQs