a3-highgpu-4g by Google Cloud Platform

a3-highgpu-4g is a Accelerator Optimized: 4 NVIDIA H100 GPU, 104 vCPUs, 936GB RAM server offered by Google Cloud Platform with 104 vCPUs, 936 GiB of memory and 3.22 TB of storage. The pricing starts at 4.5688 USD per hour.
104 vCPU
936 GiB Memory
3.22 TB Storage
4 GPU
Spare SCore
83,724
(All-cores)
1,985
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1720363
Name
a3-highgpu-4g
Description
Accelerator Optimized: 4 NVIDIA H100 GPU, 104 vCPUs, 936GB RAM
Family
a3
Hw Virt
Average Time To Start
41
Status
active
Observed At
2026-10-31T19:09:46.892581

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
26.3502 USD/h 44.245 USD/h
The Dalles (US) / us-west1
26.3502 USD/h 44.245 USD/h
Ashburn (US) / us-east4
26.5665 USD/h 44.2779 USD/h
Mumbai (IN) / asia-south1
22.3781 USD/h 46.0674 USD/h
St. Ghislain (BE) / europe-west1
22.6199 USD/h 48.6849 USD/h
Hamina (FI) / europe-north1
4.5688 USD/h 48.7178 USD/h
Columbus (US) / us-east5
22.1297 USD/h 49.7916 USD/h
Las Vegas (US) / us-west4
22.3025 USD/h 49.8245 USD/h
London (GB) / europe-west2
9.0054 USD/h 50.459 USD/h
Changhua County (TW) / asia-east1
30.7069 USD/h 51.1794 USD/h
Paris (FR) / europe-west9
19.4347 USD/h 51.3242 USD/h
Frankfurt (DE) / europe-west3
31.1017 USD/h 52.2157 USD/h
Delhi (IN) / asia-south2
17.0376 USD/h 53.1423 USD/h
Sydney (AU) / australia-southeast1
12.6584 USD/h 55.3391 USD/h
Eemshaven (NL) / europe-west4
33.7483 USD/h 56.2702 USD/h
Seoul (KR) / asia-northeast3
28.8556 USD/h 56.8246 USD/h
Jurong West (SG) / asia-southeast1
27.9394 USD/h 57.1391 USD/h
Melbourne (AU) / australia-southeast2
5.3863 USD/h 57.535 USD/h
Tokyo (JP) / asia-northeast1
10.4474 USD/h 63.3217 USD/h

Processor

vCPUs
104
CPU Allocation
Dedicated
CPU Cores
52
CPU Speed
2.7 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU Model
8481C
CPU L1D Cache
48 KiB
CPU L1D Cache Total
2 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
2 MiB
CPU L2 Cache
2 MiB
CPU L2 Cache Total
104 MiB
CPU L3 Cache
105 MiB
CPU L3 Cache Total
105 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, avx2, smep, bmi2, erms, invpcid, rtm, avx512f, avx512dq, rdseed, adx, smap, avx512ifma, clflushopt, clwb, avx512cd, sha_ni, avx512bw, avx512vl, xsaveopt, xsavec, xgetbv1, xsaves, avx_vnni, avx512_bf16, arat, avx512vbmi, umip, avx512_vbmi2, gfni, vaes, vpclmulqdq, avx512_vnni, avx512_bitalg, avx512_vpopcntdq, la57, rdpid, cldemote, movdiri, movdir64b, fsrm, md_clear, serialize, tsxldtrk, amx_bf16, avx512_fp16, amx_tile, amx_int8, arch_capabilities
Ecpus
42.2
Scalability
40.58

System Resources and Accelerators

MEMORY
Memory Amount
936 GiB
Memory Amount Actual
936 GiB
ACCELERATORS
Accelerator Count
4
Accelerator Memory Min
80 GiB
Accelerator Memory Total
320 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Hopper
Accelerator Model
H100
Accelerators
  • NVIDIA Hopper NVIDIA H100 80GB HBM3 (Memory amount: 81559, Firmware version: 590.48.01, BIOS version: 96.00.CF.00.01, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H100 80GB HBM3 (Memory amount: 81559, Firmware version: 590.48.01, BIOS version: 96.00.CF.00.01, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H100 80GB HBM3 (Memory amount: 81559, Firmware version: 590.48.01, BIOS version: 96.00.CF.00.01, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H100 80GB HBM3 (Memory amount: 81559, Firmware version: 590.48.01, BIOS version: 96.00.CF.00.01, Clock rate: 1980)
STORAGE
Storage Size
3216 GB
Storage Type
nvme ssd
Storages
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
  • 402 GB nvme ssd
NETWORK
Inbound Traffic
0 GB/month
Outbound Traffic
0 GB/month
IPv4
0

CPU and System Topology

Server Description

An accelerator-optimized architecture featuring four dedicated hardware accelerators and high-capacity local NVMe storage for parallel computing workloads.

GPU AcceleratedCompute OptimizedStorage & Database

Google Cloud Platform a3-highgpu-4g is an accelerator-optimized server featuring 104 vCPUs and 936 GB of system memory, powered by dedicated Intel Xeon 8481C processors. The hardware configuration includes four bundled NVIDIA Hopper H100 GPUs with 320 GB of total VRAM and 3216 GB of local NVMe SSD storage. Benchmark results place this instance in the top tier for CPU processing and image manipulation, with a PassMark CPU score of 86938.94 and strong multi-core performance. It features Intel Advanced Matrix Extensions and Advanced Vector Extensions to accelerate mathematical computations. This server is designed for demanding workloads such as large language model inference, machine learning training, ray tracing, and high-throughput database operations.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

4.2Score
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) 13,528,674 ops/sec 1,918,343 ops/sec 7.4 50.00% 50.00% +170.00%
Redis RPS (pipeline=16, SET) 62,066,679 ops/sec 11,704,734 ops/sec 5.2 20.00% 20.00% +39.00%
PassMark Memory Mark (composite) 2,360 2,416 1 10.00% 10.00% 0.00%
Memory bandwidth (read, 16 MB ~ L3) 232,668 MB/sec 108,466 MB/sec 2.1 10.00% 10.00% +7.70%
PassMark single-thread CPU 3,154 Mops/s 2,365 Mops/s 1.3 10.00% 10.00% +2.70%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
48.4 GB/s
Node GPU-to-host bandwidth
57.1 GB/s
Node host-to-GPU bandwidth
55.4 GB/s
Peer-to-peer
GPU-to-GPU P2P duplex bandwidth
383.4 GB/s
GPU-to-GPU P2P gather bandwidth
398.1 GB/s
GPU-to-GPU P2P latency
1,227 ns
GPU-to-GPU P2P bandwidth
388.8 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
28.6 GB/s
Per-GPU host-to-GPU bandwidth
27.7 GB/s
Per-GPU host-device latency
1,216 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Version (driver)
 12.6
nvbandwidth Version
 v0.5
GPU Count
 4
P2P Supported
Yes
Efficiency
SM/CE copy efficiency
0.91

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 2,100

Geekbench Multi-Core

Score: 19,675

Passmark CPU Scores

BENCHMARKSCORE
Mark
 86,939
Compression
 1,376,341
Encryption
 84,930
Extended Instructions
 95,800
Floating Point Maths
 285,267
Integer Maths
 400,329
Physics
 7,469
Prime Numbers
 520
Single Threaded
 3,154
String Sorting
 175,027

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,360
Database Operations
 29,920
Memory Latency
 71
Memory Read Cached
 27,843
Memory Read Uncached
 9,946
Memory Write
 8,671

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 211.1 81.01 0.38×
bitops
 156,547 7,521,668 48.05×
branch
 146,008 9,042,663 61.93×
funccall
 783.82 38,818 49.52×
funcret
 13,236 840,778 63.52×
hash
 188,775 10,888,208 57.68×
intmath
 68,320 3,212,347 47.02×
nop
 283,790 9,100,243 32.07×
prime
 497,679 22,000,394 44.21×
rotate
 105,640 4,474,558 42.36×
str
 954,993 50,783,325 53.18×
wcs
 722,809 30,990,317 42.87×
CryptographySingleMultiScaling
crypt
 3,582 135,794 37.91×
CompressionSingleMultiScaling
jpeg
 858.13 40,187 46.83×
zlib
 64.39 3,942 61.22×
HypervisorSingleMultiScaling
clock
 32,027 388,237 12.12×
clone
 428.41 8,509 19.86×
context
 5,524 361,987 65.53×
cpu-sched
 70.36 6,529 92.79×
eventfd
 67,983 2,560,442 37.66×
fault
 31,967 253,138 7.92×
fork
 4,440 100,215 22.57×
getrandom
 62,655 3,504,009 55.93×
hrtimers
 155,396 12,419,166 79.92×
mprotect
 1,071,510 15,660,462 14.62×
mtx
 16,445,374 2,802,729,709 170.43×
munmap
 3,380 105,967 31.35×
null
 946,516 59,174,960 62.52×
pipe
 585,957 45,991,091 78.49×
priv-instr
 870,412 11,131,848 12.79×
resched
 14,129,194 9,007,432 0.64×
switch
 88,525 4,000,296 45.19×
timermix
 235,997 15,711,511 66.58×
tlb-shootdown
 9,039 114,634 12.68×
tsc
 614,707 40,998,951 66.70×
usersyscall
 744,850 11,077,956 14.87×
vdso
 95,284,581 5,866,739,398 61.57×
vfork
 28,367 163,957 5.78×
x86syscall
 10,536,819 704,388,191 66.85×
yield
 76,946,465 103,482,702 1.34×
zero
 6,798,011 435,885,625 64.12×
CPU float/SIMDSingleMultiScaling
eigen
 17,115 881,770 51.52×
expmath
 3,733 215,266 57.67×
fma
 1,466,676 75,647,600 51.58×
fp
 325.82 19,691 60.43×
fractal
 7.4 422.73 57.13×
hyperbolic
 4,305 245,843 57.10×
logmath
 5,518 345,475 62.61×
matrix
 6,707 261,348 38.97×
matrix-3d
 660.39 7,851 11.89×
monte-carlo
 311.58 19,359 62.13×
mpfr
 1,845 75,420 40.88×
powmath
 2,050 123,379 60.18×
trig
 2,808 136,704 48.69×
veccmp
 27,260 1,106,904 40.61×
vecfp
 2,403 105,249 43.80×
vecmath
 4,156 229,663 55.26×
vecshuf
 4,072 303,851 74.62×
vecwide
 86,301 3,882,502 44.99×
AlgorithmsSingleMultiScaling
bitonicsort
 9.4 416.39 44.30×
bsearch
 296.26 15,287 51.60×
bubblesort
 1 41.6 41.60×
fibsearch
 324.9 15,471 47.62×
heapsort
 10.6 528.19 49.83×
hsearch
 8,776 390,984 44.55×
insertionsort
 13.2 457 34.62×
malloc
 2,824,551 108,906,964 38.56×
mcontend
 90.09 16,530 183.48×
mergesort
 32.4 1,658 51.16×
qsort
 23.2 1,162 50.07×
skiplist
 5,783 362,643 62.71×
tree
 1.6 62.23 38.89×
MemorySingleMultiScaling
cache
 327,192 2,628,154 8.03×
cachehammer
 211,344 10,516,295 49.76×
cacheline
 32.38 3,515 108.57×
l1cache
 1,156 16,674 14.43×
memcpy
 467.89 17,623 37.66×
memrate
 0.17 12.39 72.88×
memthrash
 378 1,671 4.42×
ptr-chase
 26,013,429 774,440,308 29.77×
spinmem
 253.99 4,434 17.46×
stream
 15 22,673 1511.52×
vm
 125,939 306,612 2.43×

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
a3-highgpu-1g 26234 GiB1
a3-highgpu-2g 52468 GiB2
a3-highgpu-8g 2081872 GiB8
a3-megagpu-8g 2081872 GiB8
a3-edgegpu-8g 2081872 GiB8
a3-edgegpu-8g-nolssd 2081872 GiB8
a3-ultragpu-8g 2242952 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

a3-highgpu-4g FAQs