a2-highgpu-4g by Google Cloud Platform

a2-highgpu-4g is a Accelerator Optimized: 4 NVIDIA Tesla A100 GPUs, 48 vCPUs, 340GB RAM server offered by Google Cloud Platform with 48 vCPUs, 340 GiB of memory and 0 GB of storage. The pricing starts at 7.3253 USD per hour.
48 vCPU
340 GiB Memory
4 GPU
Spare SCore
37,929
(All-cores)
1,201
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1000048
Name
a2-highgpu-4g
Description
Accelerator Optimized: 4 NVIDIA Tesla A100 GPUs, 48 vCPUs, 340GB RAM
Family
a2
Hw Virt
Average Time To Start
43
Status
active
Observed At
2026-10-27T15:26:20.211122

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
8.3753 USD/h 14.6935 USD/h
The Dalles (US) / us-west1
8.3753 USD/h 14.6935 USD/h
Moncks Corner (US) / us-east1
8.3753 USD/h 14.6935 USD/h
Eemshaven (NL) / europe-west4
7.3253 USD/h 14.9919 USD/h
Las Vegas (US) / us-west4
7.6909 USD/h 15.7374 USD/h
Jurong West (SG) / asia-southeast1
9.609 USD/h 16.055 USD/h
Tel Aviv (IL) / me-west1
9.6975 USD/h 16.1629 USD/h
Seoul (KR) / asia-northeast3
8.7711 USD/h 16.1984 USD/h
Tokyo (JP) / asia-northeast1
9.6699 USD/h 16.1984 USD/h
Salt Lake City (US) / us-west3
8.8776 USD/h 17.3003 USD/h

Processor

vCPUs
48
CPU Allocation
Dedicated
CPU Cores
24
CPU Speed
2.2 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU L1D Cache
32 KiB
CPU L1D Cache Total
768 KiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
768 KiB
CPU L2 Cache
1 MiB
CPU L2 Cache Total
24 MiB
CPU L3 Cache
39 MiB
CPU L3 Cache Total
39 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
31.6
Scalability
65.83

System Resources and Accelerators

MEMORY
Memory Amount
340 GiB
Memory Amount Actual
340 GiB
ACCELERATORS
Accelerator Count
4
Accelerator Memory Min
40 GiB
Accelerator Memory Total
160 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)
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 configuration pairing four dedicated GPUs with massive memory for high-throughput parallel computing and large language model inference.

    GPU AcceleratedMemory Optimized

    Google Cloud Platform a2-highgpu-4g is an accelerator-optimized server featuring 48 vCPUs, 24 physical Intel Xeon cores, and 340.0 GB of RAM. The hardware configuration is defined by its integration of four NVIDIA Ampere A100 GPUs with 160 GB of total VRAM, while offering no local storage. Benchmark results indicate a significant performance tradeoff: single-core CPU execution is weak, falling in the bottom 25%, whereas multi-core CPU and memory bandwidth benchmarks perform in the average tier. However, the server excels in GPU-driven tasks, achieving top-tier results in the top 10% for large language model inference prompt processing and text generation across medium and large models. This server is structured for workloads that heavily leverage GPU acceleration, such as deep learning and LLM inference, rather than CPU-bound or single-threaded applications.

    Economics

    Average Price per Region

    Prices per Zone

    Lowest Prices

    Performance

    Workload Profiles

    1.4Score
    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) 3,221,348 ops/sec 1,918,343 ops/sec 1.7 50.00% 50.00% +30.00%
    Redis RPS (pipeline=16, SET) 19,056,588 ops/sec 11,704,734 ops/sec 1.6 20.00% 20.00% +9.90%
    PassMark Memory Mark (composite) 2,511 2,416 1 10.00% 10.00% 0.00%
    Memory bandwidth (read, 16 MB ~ L3) 115,554 MB/sec 108,420 MB/sec 1.1 10.00% 10.00% +0.96%
    PassMark single-thread CPU 1,702 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
    		20.3 GB/s
    Node GPU-to-host bandwidth
    		26.4 GB/s
    Node host-to-GPU bandwidth
    		24.7 GB/s
    Peer-to-peer
    GPU-to-GPU P2P duplex bandwidth
    		207.2 GB/s
    GPU-to-GPU P2P gather bandwidth
    		277.5 GB/s
    GPU-to-GPU P2P latency
    		1,231 ns
    GPU-to-GPU P2P bandwidth
    		266.9 GB/s
    Single GPU slot
    Per-GPU GPU-to-host bandwidth
    		13.2 GB/s
    Per-GPU host-to-GPU bandwidth
    		12.3 GB/s
    Per-GPU host-device latency
    		1,223 ns
    Benchmark Environment
    Driver Version
    		 535.183.01
    CUDA Runtime Version
    		 12.6
    nvbandwidth Version
    		 v0.5
    GPU Count
    		 4
    P2P Supported
    Yes
    Efficiency
    SM/CE copy efficiency
    		1.00

    Memory Bandwidth

    Compression

    OpenSSL

    Geekbench Single-Core

    Score: 1,062

    Geekbench Multi-Core

    Score: 10,469

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 31,203
    Compression
    		 533,487
    Encryption
    		 15,898
    Extended Instructions
    		 33,690
    Floating Point Maths
    		 68,173
    Integer Maths
    		 126,886
    Physics
    		 2,568
    Prime Numbers
    		 168
    Single Threaded
    		 1,702
    String Sorting
    		 59,439

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 2,511
    Database Operations
    		 14,784
    Memory Latency
    		 53
    Memory Read Cached
    		 20,833
    Memory Read Uncached
    		 9,791
    Memory Write
    		 9,432

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 196.04 168.92 0.86×
    bitops
    		 92,134 2,538,403 27.55×
    branch
    		 131,218 4,771,374 36.36×
    funccall
    		 463.7 11,387 24.56×
    funcret
    		 9,652 328,226 34.01×
    hash
    		 107,530 3,669,727 34.13×
    intmath
    		 19,059 680,664 35.71×
    nop
    		 163,091 3,377,232 20.71×
    prime
    		 271,698 7,260,388 26.72×
    rotate
    		 82,435 2,706,626 32.83×
    str
    		 643,204 15,246,161 23.70×
    wcs
    		 403,053 9,643,415 23.93×
    CryptographySingleMultiScaling
    crypt
    		 2,082 48,579 23.33×
    CompressionSingleMultiScaling
    jpeg
    		 498.96 13,745 27.55×
    zlib
    		 40.82 1,613 39.52×
    HypervisorSingleMultiScaling
    clock
    		 24,343 323,906 13.31×
    clone
    		 297.96 4,844 16.26×
    context
    		 2,238 60,093 26.86×
    cpu-sched
    		 71.64 3,056 42.65×
    eventfd
    		 46,072 972,489 21.11×
    fault
    		 9,401 221,722 23.59×
    fork
    		 3,437 69,266 20.15×
    getrandom
    		 38,170 1,112,969 29.16×
    hrtimers
    		 138,269 4,278,610 30.94×
    mprotect
    		 1,045,708 6,499,097 6.22×
    mtx
    		 8,534,480 1,077,181,469 126.22×
    munmap
    		 2,303 54,104 23.49×
    null
    		 253,543 5,885,930 23.21×
    pipe
    		 838,434 13,315,492 15.88×
    priv-instr
    		 482,566 11,032,081 22.86×
    resched
    		 5,534,427 2,941,365 0.53×
    switch
    		 73,385 7,008,612 95.50×
    timermix
    		 168,607 5,570,881 33.04×
    tlb-shootdown
    		 10,269 57,683 5.62×
    tsc
    		 850,352 28,381,270 33.38×
    usersyscall
    		 331,626 8,620,435 25.99×
    vdso
    		 70,609,085 2,340,473,390 33.15×
    vfork
    		 22,823 141,813 6.21×
    x86syscall
    		 2,914,942 66,724,691 22.89×
    yield
    		 21,246,445 25,595,944 1.20×
    zero
    		 1,731,867 44,033,632 25.43×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 13,773 396,052 28.75×
    expmath
    		 2,291 83,213 36.32×
    fma
    		 510,843 19,936,587 39.03×
    fp
    		 102.97 3,157 30.66×
    fractal
    		 4.2 147.08 35.02×
    hyperbolic
    		 2,380 89,175 37.47×
    logmath
    		 3,863 136,246 35.27×
    matrix
    		 2,600 99,551 38.29×
    matrix-3d
    		 618.11 3,345 5.41×
    monte-carlo
    		 118.63 3,088 26.03×
    mpfr
    		 1,032 26,209 25.39×
    powmath
    		 1,352 49,249 36.43×
    trig
    		 1,456 44,392 30.48×
    veccmp
    		 13,273 320,011 24.11×
    vecfp
    		 1,151 34,016 29.56×
    vecmath
    		 1,870 70,277 37.58×
    vecshuf
    		 481.07 12,356 25.68×
    vecwide
    		 22,270 570,478 25.62×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 6.2 163.19 26.32×
    bsearch
    		 206.02 5,587 27.12×
    bubblesort
    		 0.6 9.6 16.00×
    fibsearch
    		 208.31 5,737 27.54×
    heapsort
    		 7.4 211.19 28.54×
    hsearch
    		 4,292 125,378 29.21×
    insertionsort
    		 7 163.2 23.31×
    malloc
    		 1,983,283 49,498,312 24.96×
    mcontend
    		 104.48 9,349 89.48×
    mergesort
    		 19.8 597.98 30.20×
    qsort
    		 12.6 336.19 26.68×
    skiplist
    		 4,153 140,790 33.90×
    tree
    		 1.2 19.16 15.97×
    MemorySingleMultiScaling
    cache
    		 343,834 721,299 2.10×
    cachehammer
    		 364,598 12,915,117 35.42×
    cacheline
    		 22.57 2,143 94.96×
    l1cache
    		 193.68 4,963 25.63×
    memcpy
    		 295.87 7,484 25.29×
    memrate
    		 1.17 8.35 7.14×
    memthrash
    		 229.91 908.04 3.95×
    ptr-chase
    		 29,117,022 431,120,951 14.81×
    spinmem
    		 190.59 1,734 9.10×
    stream
    		 28.67 12,416 433.07×
    vm
    		 52,762 300,404 5.69×

    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
    a2-highgpu-1g 1285 GiB1
    a2-ultragpu-1g 12170 GiB1
    a2-highgpu-2g 24170 GiB2
    a2-ultragpu-2g 24340 GiB2
    a2-ultragpu-4g 48680 GiB4
    a2-highgpu-8g 96680 GiB8
    a2-ultragpu-8g 961360 GiB8

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    a2-highgpu-4g FAQs