a2-highgpu-8g by Google Cloud Platform

a2-highgpu-8g is a Accelerator Optimized: 8 NVIDIA Tesla A100 GPUs, 96 vCPUs, 680GB RAM server offered by Google Cloud Platform with 96 vCPUs, 680 GiB of memory and 0 GB of storage. The pricing starts at 13.918 USD per hour.
96 vCPU
680 GiB Memory
8 GPU
Spare SCore
75,800
(All-cores)
1,201
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1000096
Name
a2-highgpu-8g
Description
Accelerator Optimized: 8 NVIDIA Tesla A100 GPUs, 96 vCPUs, 680GB RAM
Family
a2
Hw Virt
Average Time To Start
78
Status
active
Observed At
2026-11-05T11:58:09.218270

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
15.913 USD/h 29.3871 USD/h
The Dalles (US) / us-west1
15.913 USD/h 29.3871 USD/h
Moncks Corner (US) / us-east1
15.913 USD/h 29.3871 USD/h
Eemshaven (NL) / europe-west4
13.918 USD/h 29.9838 USD/h
Las Vegas (US) / us-west4
14.6123 USD/h 31.4748 USD/h
Jurong West (SG) / asia-southeast1
19.2181 USD/h 32.1101 USD/h
Tel Aviv (IL) / me-west1
19.395 USD/h 32.3258 USD/h
Seoul (KR) / asia-northeast3
19.2958 USD/h 32.3967 USD/h
Tokyo (JP) / asia-northeast1
19.3398 USD/h 32.3967 USD/h
Salt Lake City (US) / us-west3
18.643 USD/h 34.6005 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
680 GiB
Memory Amount Actual
680 GiB
ACCELERATORS
Accelerator Count
8
Accelerator Memory Min
40 GiB
Accelerator Memory Total
320 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: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100-SXM4-40GB (Memory amount: 40960, Firmware version: 590.48.01, BIOS version: 92.00.81.00.01, 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 architecture featuring eight dedicated GPUs and high memory bandwidth for intensive parallel computing and language model inference.

    GPU AcceleratedMemory OptimizedCompute Optimized

    Google Cloud Platform a2-highgpu-8g is an accelerator-optimized server designed for parallel computing and machine learning workloads. It features eight NVIDIA Ampere A100 GPUs with 320 GB of total VRAM, paired with an Intel Xeon x86_64 processor providing 96 dedicated vCPUs and 680.0 GB of system memory. While single-core CPU benchmarks show weak performance, the server demonstrates strong multi-core capabilities and top-tier performance in large language model inference prompt processing and text generation. It lacks local storage and complimentary public IPv4 addresses. This server is structured for GPU-intensive applications, including deep learning, ray tracing, and high-throughput multi-threaded processing, where the high density of accelerators justifies the hardware class investment.

    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) 6,771,021 ops/sec 1,937,246 ops/sec 3.6 50.00% 50.00% +90.00%
    Redis RPS (pipeline=16, SET) 39,929,924 ops/sec 11,803,344 ops/sec 3.3 20.00% 20.00% +27.00%
    PassMark Memory Mark (composite) 2,509 2,409 1 10.00% 10.00% 0.00%
    Memory bandwidth (read, 16 MB ~ L3) 231,594 MB/sec 108,490 MB/sec 2.1 10.00% 10.00% +7.70%
    PassMark single-thread CPU 1,699 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
    		41.5 GB/s
    Node GPU-to-host bandwidth
    		52.8 GB/s
    Node host-to-GPU bandwidth
    		49.5 GB/s
    Peer-to-peer
    GPU-to-GPU P2P duplex bandwidth
    		204.0 GB/s
    GPU-to-GPU P2P gather bandwidth
    		264.6 GB/s
    GPU-to-GPU P2P latency
    		1,261 ns
    GPU-to-GPU P2P bandwidth
    		264.2 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
    		1,221 ns
    Benchmark Environment
    Driver Version
    		 535.183.01
    CUDA Version (driver)
    		 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: 1,063

    Geekbench Multi-Core

    Score: 11,752

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 49,728
    Compression
    		 1,059,158
    Encryption
    		 31,839
    Extended Instructions
    		 67,287
    Floating Point Maths
    		 136,204
    Integer Maths
    		 253,339
    Physics
    		 5,048
    Prime Numbers
    		 333
    Single Threaded
    		 1,699
    String Sorting
    		 118,795

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 2,509
    Database Operations
    		 20,510
    Memory Latency
    		 54
    Memory Read Cached
    		 20,819
    Memory Read Uncached
    		 9,785
    Memory Write
    		 9,390

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 153.72 104.95 0.68×
    bitops
    		 92,273 5,032,775 54.54×
    branch
    		 131,431 9,529,748 72.51×
    funccall
    		 463.41 21,778 47.00×
    funcret
    		 9,638 629,715 65.33×
    hash
    		 107,294 7,334,135 68.36×
    intmath
    		 19,001 1,357,668 71.45×
    nop
    		 163,183 6,755,467 41.40×
    prime
    		 271,631 13,931,510 51.29×
    rotate
    		 82,434 5,439,135 65.98×
    str
    		 585,724 28,708,335 49.01×
    wcs
    		 401,978 18,415,903 45.81×
    CryptographySingleMultiScaling
    crypt
    		 2,082 94,959 45.60×
    CompressionSingleMultiScaling
    jpeg
    		 501.34 26,301 52.46×
    zlib
    		 41.42 3,051 73.66×
    HypervisorSingleMultiScaling
    clock
    		 21,939 173,431 7.91×
    clone
    		 293.51 6,994 23.83×
    context
    		 2,249 116,351 51.73×
    cpu-sched
    		 60.74 5,793 95.38×
    eventfd
    		 46,427 1,832,714 39.47×
    fault
    		 9,376 137,411 14.65×
    fork
    		 3,422 46,586 13.61×
    getrandom
    		 38,111 2,163,008 56.76×
    hrtimers
    		 134,610 9,357,178 69.51×
    mprotect
    		 590,465 11,517,943 19.51×
    mtx
    		 9,748,710 1,721,376,095 176.57×
    munmap
    		 2,231 54,405 24.39×
    null
    		 254,359 11,257,218 44.26×
    pipe
    		 855,262 21,577,419 25.23×
    priv-instr
    		 484,929 9,665,392 19.93×
    resched
    		 4,370,365 5,289,510 1.21×
    switch
    		 79,296 12,835,622 161.87×
    timermix
    		 158,873 10,985,663 69.15×
    tlb-shootdown
    		 5,732 94,112 16.42×
    tsc
    		 849,632 56,632,736 66.66×
    usersyscall
    		 328,416 8,490,888 25.85×
    vdso
    		 70,082,275 4,416,062,051 63.01×
    vfork
    		 20,019 104,002 5.20×
    x86syscall
    		 2,895,300 126,794,449 43.79×
    yield
    		 43,262,266 47,330,271 1.09×
    zero
    		 1,729,473 86,616,563 50.08×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 13,753 773,265 56.23×
    expmath
    		 2,280 159,511 69.95×
    fma
    		 509,954 39,168,486 76.81×
    fp
    		 110.4 6,135 55.57×
    fractal
    		 4.2 282.67 67.30×
    hyperbolic
    		 2,372 173,616 73.21×
    logmath
    		 3,865 261,536 67.67×
    matrix
    		 2,592 190,771 73.59×
    matrix-3d
    		 596.98 6,780 11.36×
    monte-carlo
    		 118.31 6,139 51.89×
    mpfr
    		 1,034 49,925 48.30×
    powmath
    		 1,353 94,048 69.50×
    trig
    		 1,461 84,668 57.95×
    veccmp
    		 13,307 614,545 46.18×
    vecfp
    		 1,151 67,336 58.52×
    vecmath
    		 1,873 134,289 71.69×
    vecshuf
    		 481.55 24,441 50.75×
    vecwide
    		 21,976 1,083,069 49.28×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 6.2 316.39 51.03×
    bsearch
    		 203.89 11,302 55.43×
    bubblesort
    		 0.6 19.2 32.00×
    fibsearch
    		 208.4 11,480 55.09×
    heapsort
    		 7.4 403.39 54.51×
    hsearch
    		 4,188 240,511 57.42×
    insertionsort
    		 7 316.8 45.26×
    malloc
    		 2,026,343 56,234,290 27.75×
    mcontend
    		 69.1 17,166 248.43×
    mergesort
    		 19.8 1,144 57.77×
    qsort
    		 12.6 643.58 51.08×
    skiplist
    		 4,154 273,039 65.73×
    tree
    		 1.2 38.38 31.98×
    MemorySingleMultiScaling
    cache
    		 288,728 1,715,755 5.94×
    cachehammer
    		 351,818 14,223,922 40.43×
    cacheline
    		 20.74 1,408 67.88×
    l1cache
    		 192.92 9,391 48.68×
    memcpy
    		 295.75 14,291 48.32×
    memrate
    		 1.17 14.87 12.71×
    memthrash
    		 283.24 1,075 3.79×
    ptr-chase
    		 30,614,017 866,089,846 28.29×
    spinmem
    		 190.18 3,535 18.59×
    stream
    		 13.78 24,065 1746.34×
    vm
    		 52,087 524,754 10.07×

    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-highgpu-4g 48340 GiB4
    a2-ultragpu-4g 48680 GiB4
    a2-ultragpu-8g 961360 GiB8

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    a2-highgpu-8g FAQs