a3-highgpu-8g by Google Cloud Platform

a3-highgpu-8g is a Accelerator Optimized: 8 NVIDIA H100 GPU, 208 vCPUs, 1872GB RAM, 16 local SSD server offered by Google Cloud Platform with 208 vCPUs, 1872 GiB of memory and 6.17 TB of storage. The pricing starts at 9.1376 USD per hour.
208 vCPU
1872 GiB Memory
6.17 TB Storage
8 GPU
Spare SCore
167,332
(All-cores)
2,053
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1720364
Name
a3-highgpu-8g
Description
Accelerator Optimized: 8 NVIDIA H100 GPU, 208 vCPUs, 1872GB RAM, 16 local SSD
Family
a3
Hw Virt
Status
active
Observed At
2026-10-31T19:09:46.892637

Availability

REGION / IDSPOTONDEMAND
Council Bluffs (US) / us-central1
52.7005 USD/h 88.49 USD/h
The Dalles (US) / us-west1
52.7005 USD/h 88.49 USD/h
Ashburn (US) / us-east4
53.1331 USD/h 88.5558 USD/h
Mumbai (IN) / asia-south1
44.7563 USD/h 92.1348 USD/h
St. Ghislain (BE) / europe-west1
45.2399 USD/h 97.3698 USD/h
Hamina (FI) / europe-north1
9.1376 USD/h 97.4356 USD/h
Columbus (US) / us-east5
44.2594 USD/h 99.5832 USD/h
Las Vegas (US) / us-west4
44.6049 USD/h 99.649 USD/h
London (GB) / europe-west2
18.0107 USD/h 100.9181 USD/h
Changhua County (TW) / asia-east1
61.4139 USD/h 102.3588 USD/h
Paris (FR) / europe-west9
38.8694 USD/h 102.6484 USD/h
Frankfurt (DE) / europe-west3
62.2034 USD/h 104.4314 USD/h
Delhi (IN) / asia-south2
34.0753 USD/h 106.2846 USD/h
Sydney (AU) / australia-southeast1
25.3168 USD/h 110.6783 USD/h
Eemshaven (NL) / europe-west4
67.4966 USD/h 112.5405 USD/h
Seoul (KR) / asia-northeast3
57.7113 USD/h 113.6492 USD/h
Jurong West (SG) / asia-southeast1
55.8789 USD/h 114.2783 USD/h
Melbourne (AU) / australia-southeast2
10.7726 USD/h 115.0699 USD/h
Tokyo (JP) / asia-northeast1
20.8948 USD/h 126.6434 USD/h

Processor

vCPUs
208
CPU Allocation
Dedicated
CPU Cores
104
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
5 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
3 MiB
CPU L2 Cache
2 MiB
CPU L2 Cache Total
208 MiB
CPU L3 Cache
105 MiB
CPU L3 Cache Total
210 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, amx_bf16, avx512_fp16, amx_tile, amx_int8, arch_capabilities
Ecpus
81.5
Scalability
39.18

System Resources and Accelerators

MEMORY
Memory Amount
2 TB
Memory Amount Actual
2 TB
ACCELERATORS
Accelerator Count
8
Accelerator Memory Min
80 GiB
Accelerator Memory Total
640 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)
  • 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
6167 GB
Storage Type
nvme ssd
Storages
  • 137 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
  • 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 platform combining eight high-capacity GPUs, extensive system memory, and local NVMe storage for massive parallel computing workloads.

GPU AcceleratedCompute OptimizedStorage & Database

Google Cloud Platform a3-highgpu-8g is an accelerator-optimized server featuring eight NVIDIA H100 GPUs with 640 GB of total VRAM, 208 dedicated Intel Xeon vCPUs, 1872.0 GB of system memory, and 6167 GB of local NVMe SSD storage. Built on the x86_64 architecture, the server utilizes Intel Xeon 8481C processors supporting Advanced Matrix Extensions and AVX-512. Benchmark results place its multi-core CPU performance, memory bandwidth, and database operations in the top 10% tier, while single-core metrics sit in the middle 50% average. This resource density makes the server highly efficient for parallelized workloads, minimizing cluster overhead. It is engineered for demanding workloads such as large language model inference, deep learning training, database operations, and high-throughput data processing.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

7Score
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) 26,654,430 ops/sec 1,918,343 ops/sec 14 50.00% 50.00% +270.00%
Redis RPS (pipeline=16, SET) 124,358,458 ops/sec 11,704,734 ops/sec 10 20.00% 20.00% +58.00%
PassMark Memory Mark (composite) 2,359 2,416 1 10.00% 10.00% 0.00%
Memory bandwidth (read, 16 MB ~ L3) 460,829 MB/sec 108,466 MB/sec 4.2 10.00% 10.00% +15.00%
PassMark single-thread CPU 3,135 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
96.8 GB/s
Node GPU-to-host bandwidth
114.2 GB/s
Node host-to-GPU bandwidth
111.0 GB/s
Peer-to-peer
GPU-to-GPU P2P duplex bandwidth
383.5 GB/s
GPU-to-GPU P2P gather bandwidth
398.9 GB/s
GPU-to-GPU P2P latency
1,241 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
1,212 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
0.91

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 2,123

Geekbench Multi-Core

Score: 19,014

Passmark CPU Scores

BENCHMARKSCORE
Mark
 124,500
Compression
 2,653,996
Encryption
 166,857
Extended Instructions
 185,071
Floating Point Maths
 569,337
Integer Maths
 795,827
Physics
 14,781
Prime Numbers
 1,017
Single Threaded
 3,135
String Sorting
 338,535

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,359
Database Operations
 34,384
Memory Latency
 71
Memory Read Cached
 27,762
Memory Read Uncached
 9,821
Memory Write
 8,556

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 196.99 33.91 0.17×
bitops
 156,993 15,023,312 95.69×
branch
 146,574 18,051,411 123.16×
funccall
 788.34 75,621 95.92×
funcret
 13,133 1,626,356 123.83×
hash
 189,012 21,665,039 114.62×
intmath
 68,519 6,393,264 93.31×
nop
 281,664 18,197,590 64.61×
prime
 498,697 42,363,642 84.95×
rotate
 105,616 8,948,679 84.73×
str
 990,004 93,171,205 94.11×
wcs
 728,642 59,638,250 81.85×
CryptographySingleMultiScaling
crypt
 3,578 261,854 73.18×
CompressionSingleMultiScaling
jpeg
 881.82 77,535 87.93×
zlib
 69.26 7,918 114.32×
HypervisorSingleMultiScaling
clock
 22,026 202,467 9.19×
clone
 421.26 13,913 33.03×
context
 5,492 723,606 131.75×
cpu-sched
 53.91 12,322 228.57×
eventfd
 66,724 2,317,781 34.74×
fault
 32,261 160,964 4.99×
fork
 4,036 52,949 13.12×
getrandom
 62,834 6,784,813 107.98×
hrtimers
 132,072 15,747,116 119.23×
mprotect
 633,439 13,903,313 21.95×
mtx
 27,722,000 4,920,043,402 177.48×
munmap
 3,146 64,209 20.41×
null
 946,905 114,530,641 120.95×
pipe
 655,333 80,132,911 122.28×
priv-instr
 879,226 6,887,536 7.83×
resched
 10,140,060 25,036,395 2.47×
switch
 102,781 6,991,268 68.02×
timermix
 235,697 28,718,768 121.85×
tlb-shootdown
 7,182 170,224 23.70×
tsc
 617,758 81,283,722 131.58×
usersyscall
 753,561 6,958,627 9.23×
vdso
 95,602,709 11,726,229,489 122.66×
vfork
 23,053 115,463 5.01×
x86syscall
 10,567,912 1,407,814,044 133.22×
yield
 154,079,819 193,194,937 1.25×
zero
 6,819,184 857,421,678 125.74×
CPU float/SIMDSingleMultiScaling
eigen
 17,130 1,682,140 98.20×
expmath
 3,739 412,856 110.42×
fma
 1,463,675 145,878,235 99.67×
fp
 326.39 37,967 116.32×
fractal
 7.4 818.26 110.58×
hyperbolic
 4,310 472,916 109.73×
logmath
 5,531 685,340 123.91×
matrix
 6,632 493,835 74.47×
matrix-3d
 644.26 15,912 24.70×
monte-carlo
 311.5 37,869 121.57×
mpfr
 1,834 144,868 79.00×
powmath
 2,044 240,898 117.84×
trig
 2,815 263,369 93.57×
veccmp
 27,247 2,138,488 78.49×
vecfp
 2,409 210,244 87.29×
vecmath
 4,166 441,972 106.09×
vecshuf
 4,074 606,855 148.95×
vecwide
 86,448 7,600,179 87.92×
AlgorithmsSingleMultiScaling
bitonicsort
 9.4 815.58 86.76×
bsearch
 296.02 28,617 96.67×
bubblesort
 1 83.2 83.20×
fibsearch
 321.87 31,148 96.77×
heapsort
 10.6 1,023 96.51×
hsearch
 8,751 772,312 88.25×
insertionsort
 13.4 880.99 65.75×
malloc
 2,775,538 124,542,853 44.87×
mcontend
 53.05 28,637 539.81×
mergesort
 32.2 3,201 99.41×
qsort
 23.2 2,223 95.82×
skiplist
 5,790 698,803 120.70×
tree
 1.6 124.73 77.96×
MemorySingleMultiScaling
cache
 220,284 8,605,118 39.06×
cachehammer
 203,982 8,556,808 41.95×
cacheline
 27.58 2,420 87.75×
l1cache
 1,172 24,735 21.10×
memcpy
 463.97 33,422 72.04×
memrate
 0.17 24.25 142.65×
memthrash
 475.3 941.06 1.98×
ptr-chase
 26,082,291 1,547,797,792 59.34×
spinmem
 253.99 8,834 34.78×
stream
 7.07 40,828 5774.80×
vm
 127,926 507,833 3.97×

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-4g 104936 GiB4
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-8g FAQs