a3-highgpu-1g by Google Cloud Platform

a3-highgpu-1g is a Accelerator Optimized: 1 NVIDIA H100 GPU, 26 vCPUs, 234GB RAM server offered by Google Cloud Platform with 26 vCPUs, 234 GiB of memory and 804 GB of storage. The pricing starts at 1.1422 USD per hour.
26 vCPU
234 GiB Memory
804 GB Storage
1 GPU
Spare SCore
27,428
(All-cores)
2,019
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1720361
Name
a3-highgpu-1g
Description
Accelerator Optimized: 1 NVIDIA H100 GPU, 26 vCPUs, 234GB RAM
Family
a3
Hw Virt
Average Time To Start
30
Status
active
Observed At
2026-10-31T15:45:34.013299

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
6.5876 USD/h 11.0612 USD/h
The Dalles (US) / us-west1
6.5876 USD/h 11.0612 USD/h
Ashburn (US) / us-east4
6.6416 USD/h 11.0695 USD/h
Mumbai (IN) / asia-south1
5.5945 USD/h 11.5168 USD/h
St. Ghislain (BE) / europe-west1
5.655 USD/h 12.1712 USD/h
Hamina (FI) / europe-north1
1.1422 USD/h 12.1795 USD/h
Columbus (US) / us-east5
5.5324 USD/h 12.4479 USD/h
Las Vegas (US) / us-west4
5.5756 USD/h 12.4561 USD/h
London (GB) / europe-west2
2.2513 USD/h 12.6148 USD/h
Changhua County (TW) / asia-east1
7.6767 USD/h 12.7948 USD/h
Paris (FR) / europe-west9
4.8587 USD/h 12.8311 USD/h
Frankfurt (DE) / europe-west3
7.7754 USD/h 13.0539 USD/h
Delhi (IN) / asia-south2
4.2594 USD/h 13.2856 USD/h
Sydney (AU) / australia-southeast1
3.1646 USD/h 13.8348 USD/h
Eemshaven (NL) / europe-west4
8.4371 USD/h 14.0676 USD/h
Seoul (KR) / asia-northeast3
7.2139 USD/h 14.2062 USD/h
Jurong West (SG) / asia-southeast1
6.9849 USD/h 14.2848 USD/h
Melbourne (AU) / australia-southeast2
1.3466 USD/h 14.3837 USD/h
Tokyo (JP) / asia-northeast1
2.6118 USD/h 15.8304 USD/h

Processor

vCPUs
26
CPU Allocation
Dedicated
CPU Cores
13
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
624 KiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
416 KiB
CPU L2 Cache
2 MiB
CPU L2 Cache Total
26 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
13.6
Scalability
52.31

System Resources and Accelerators

MEMORY
Memory Amount
234 GiB
Memory Amount Actual
234 GiB
ACCELERATORS
Accelerator Count
1
Accelerator Memory Min
80 GiB
Accelerator Memory Total
80 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.A5.00.01, Clock rate: 1980)
STORAGE
Storage Size
804 GB
Storage Type
nvme ssd
Storages
  • 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 server featuring a single NVIDIA H100 GPU and dedicated Intel Xeon processors for high-performance machine learning workloads.

GPU AcceleratedMemory Optimized

Google Cloud Platform a3-highgpu-1g is an accelerator-optimized instance designed for machine learning and high-performance computing. It features 26 dedicated Intel Xeon 8481C vCPUs operating at 2.7 GHz, 234 GB of RAM, and an 804 GB NVMe SSD. The core of its performance is a single NVIDIA Hopper H100 GPU with 80 GB of VRAM. Benchmarks show strong single-core capabilities, with a Geekbench single-core score of 2072.0, and top-tier text generation speeds of 348.05 tokens per second for small 135M parameter LLMs. However, performance drops significantly to 2.49 tokens per second for large 70B models, highlighting a clear scaling tradeoff. This instance is qualitatively positioned as a high-density, single-GPU solution best suited for small-to-medium LLM inference, deep learning training, and GPU-accelerated data science workloads.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

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) 5,218,913 ops/sec 1,918,343 ops/sec 2.7 50.00% 50.00% +64.00%
Redis RPS (pipeline=16, SET) 24,063,433 ops/sec 11,704,734 ops/sec 2 20.00% 20.00% +15.00%
PassMark Memory Mark (composite) 2,194 2,416 0.92 10.00% 10.00% -0.83%
Memory bandwidth (read, 16 MB ~ L3) 144,753 MB/sec 108,466 MB/sec 1.3 10.00% 10.00% +2.70%
PassMark single-thread CPU 2,802 Mops/s 2,365 Mops/s 1.2 10.00% 10.00% +1.80%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
24.1 GB/s
Node GPU-to-host bandwidth
28.6 GB/s
Node host-to-GPU bandwidth
27.7 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
1,219 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Version (driver)
 12.6
nvbandwidth Version
 v0.5
GPU Count
 1
P2P Supported
No
Efficiency
SM/CE copy efficiency
0.91

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 2,072

Geekbench Multi-Core

Score: 14,893

Passmark CPU Scores

BENCHMARKSCORE
Mark
 35,694
Compression
 413,228
Encryption
 23,911
Extended Instructions
 29,708
Floating Point Maths
 76,555
Integer Maths
 112,126
Physics
 4,676
Prime Numbers
 223
Single Threaded
 2,802
String Sorting
 54,325

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,194
Database Operations
 12,845
Memory Latency
 75
Memory Read Cached
 25,869
Memory Read Uncached
 9,304
Memory Write
 8,245

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 184.29 165.17 0.90×
bitops
 125,498 1,892,652 15.08×
branch
 118,446 2,271,000 19.17×
funccall
 630.67 9,984 15.83×
funcret
 10,615 207,897 19.58×
hash
 152,316 2,729,864 17.92×
intmath
 55,788 811,668 14.55×
nop
 231,165 2,319,621 10.03×
prime
 405,990 5,949,542 14.65×
rotate
 84,569 1,121,779 13.26×
str
 773,412 14,365,734 18.57×
wcs
 570,494 8,893,240 15.59×
CryptographySingleMultiScaling
crypt
 2,889 34,226 11.85×
CompressionSingleMultiScaling
jpeg
 717.88 11,402 15.88×
zlib
 47.17 1,061 22.50×
HypervisorSingleMultiScaling
clock
 40,004 361,021 9.02×
clone
 396.71 4,251 10.71×
context
 4,418 91,015 20.60×
cpu-sched
 66.88 1,642 24.55×
eventfd
 55,138 876,219 15.89×
fault
 26,502 266,558 10.06×
fork
 5,261 68,554 13.03×
getrandom
 50,469 910,570 18.04×
hrtimers
 186,703 2,595,936 13.90×
mprotect
 951,995 6,925,186 7.27×
mtx
 10,654,392 738,332,112 69.30×
munmap
 3,966 53,430 13.47×
null
 761,803 14,900,001 19.56×
pipe
 541,868 18,998,559 35.06×
priv-instr
 703,193 11,253,529 16.00×
resched
 9,869,926 1,638,451 0.17×
switch
 79,838 6,798,493 85.15×
timermix
 198,538 4,181,792 21.06×
tlb-shootdown
 8,958 43,104 4.81×
tsc
 492,407 10,292,796 20.90×
usersyscall
 578,808 9,882,517 17.07×
vdso
 76,660,955 1,480,356,848 19.31×
vfork
 34,042 249,724 7.34×
x86syscall
 8,506,733 177,316,917 20.84×
yield
 22,106,277 23,852,808 1.08×
zero
 5,539,218 110,162,600 19.89×
CPU float/SIMDSingleMultiScaling
eigen
 13,879 219,957 15.85×
expmath
 3,014 54,045 17.93×
fma
 1,187,818 20,529,282 17.28×
fp
 265.48 5,056 19.04×
fractal
 6 106.41 17.73×
hyperbolic
 3,493 61,841 17.70×
logmath
 4,436 86,865 19.58×
matrix
 5,296 70,286 13.27×
matrix-3d
 617.6 5,140 8.32×
monte-carlo
 247.67 4,988 20.14×
mpfr
 1,490 20,454 13.72×
powmath
 1,646 31,156 18.93×
trig
 2,248 36,900 16.41×
veccmp
 22,087 292,465 13.24×
vecfp
 1,855 26,131 14.09×
vecmath
 3,399 58,837 17.31×
vecshuf
 3,259 76,103 23.35×
vecwide
 65,975 1,022,201 15.49×
AlgorithmsSingleMultiScaling
bitonicsort
 7.4 109.4 14.78×
bsearch
 238.4 3,927 16.47×
bubblesort
 0.8 10.4 13.00×
fibsearch
 265.27 3,840 14.48×
heapsort
 8.6 141.4 16.44×
hsearch
 7,099 110,660 15.59×
insertionsort
 10.8 119 11.02×
malloc
 2,448,213 36,832,843 15.04×
mcontend
 92.66 4,157 44.87×
mergesort
 26.2 462.99 17.67×
qsort
 18.6 319.59 17.18×
skiplist
 4,676 91,344 19.54×
tree
 1.4 20.78 14.84×
MemorySingleMultiScaling
cache
 313,174 999,119 3.19×
cachehammer
 206,036 4,348,495 21.11×
cacheline
 52.76 1,119 21.20×
l1cache
 941.74 4,454 4.73×
memcpy
 375.49 4,276 11.39×
memrate
 0.17 3.99 23.47×
memthrash
 207.2 917.16 4.43×
ptr-chase
 24,288,990 244,340,627 10.06×
spinmem
 211.99 1,054 4.97×
stream
 14.07 4,158 295.55×
vm
 102,477 295,207 2.88×

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-2g 52468 GiB2
a3-highgpu-4g 104936 GiB4
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-1g FAQs