a3-ultragpu-8g by Google Cloud Platform

a3-ultragpu-8g is a Accelerator Optimized: 8 NVIDIA H200 GPU, 224 vCPUs, 2952GB RAM server offered by Google Cloud Platform with 224 vCPUs, 2952 GiB of memory and 12.6 TB of storage. The pricing starts at 50.5748 USD per hour.
224 vCPU
2952 GiB Memory
12.6 TB Storage
8 GPU
Spare SCore
180,093
(All-cores)
2,143
(Single-core)

Specifications

Server Metadata

Vendor ID
gcp
Server ID
1722000
Name
a3-ultragpu-8g
Description
Accelerator Optimized: 8 NVIDIA H200 GPU, 224 vCPUs, 2952GB RAM
Family
a3
Hw Virt
Average Time To Start
285
Status
active
Observed At
2026-10-31T05:36:32.308681

Availability

REGION / IDSPOTONDEMAND
Eemshaven (NL) / europe-west4
55.9602 USD/h -
Dallas (US) / us-south1
59.8917 USD/h -
Council Bluffs (US) / us-central1
50.8712 USD/h 84.8069 USD/h
The Dalles (US) / us-west1
50.8712 USD/h 84.8069 USD/h
Ashburn (US) / us-east4
50.5748 USD/h 84.9384 USD/h
St. Ghislain (BE) / europe-west1
55.9412 USD/h 93.2623 USD/h
Mumbai (IN) / asia-south1
60.3046 USD/h 101.7738 USD/h

Processor

vCPUs
224
CPU Allocation
Dedicated
CPU Cores
112
CPU Speed
2.1 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU Model
8581C
CPU L1D Cache
48 KiB
CPU L1D Cache Total
5 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
4 MiB
CPU L2 Cache
2 MiB
CPU L2 Cache Total
224 MiB
CPU L3 Cache
260 MiB
CPU L3 Cache Total
520 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, monitor, vmx, 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, tpr_shadow, flexpriority, ept, vpid, ept_ad, fsgsbase, tsc_adjust, bmi1, hle, 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, wbnoinvd, arat, vnmi, 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
84
Scalability
37.5

System Resources and Accelerators

MEMORY
Memory Amount
3 TB
Memory Amount Actual
3 TB
ACCELERATORS
Accelerator Count
8
Accelerator Memory Min
141 GiB
Accelerator Memory Total
1 TB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Hopper
Accelerator Model
H200
Accelerators
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
  • NVIDIA Hopper NVIDIA H200 (Memory amount: 143771, Firmware version: 595.71.05, BIOS version: 96.00.D9.00.02, Clock rate: 1980)
STORAGE
Storage Size
12599 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
  • 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
  • 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
  • 137 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 featuring eight high-performance GPUs, massive system memory, and high-speed local NVMe storage for intensive parallel computing.

GPU AcceleratedStorage & DatabaseCompute Optimized

Google Cloud Platform a3-ultragpu-8g is an accelerator-optimized server designed for high-performance computing and AI workloads. It features eight NVIDIA Hopper H200 GPUs with 1128 GB of total VRAM, paired with 224 dedicated vCPUs on an Intel Xeon 8581C x86_64 processor. The system includes 2952.0 GB of RAM and 12599 GB of local NVMe SSD storage. Benchmarks indicate top-tier multi-core CPU performance, memory bandwidth, and LLM inference speeds, though single-core performance remains average. This server is highly suited for large language model training, complex simulations, ray tracing, and high-throughput database operations.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

7.6Score
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) 28,732,061 ops/sec 1,918,343 ops/sec 15 50.00% 50.00% +290.00%
Redis RPS (pipeline=16, SET) 151,183,179 ops/sec 11,704,734 ops/sec 12 20.00% 20.00% +64.00%
PassMark Memory Mark (composite) 2,379 2,416 1 10.00% 10.00% 0.00%
Memory bandwidth (read, 16 MB ~ L3) 515,217 MB/sec 108,466 MB/sec 4.7 10.00% 10.00% +17.00%
PassMark single-thread CPU 3,208 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
153.0 GB/s
Node GPU-to-host bandwidth
223.2 GB/s
Node host-to-GPU bandwidth
222.8 GB/s
Peer-to-peer
GPU-to-GPU P2P duplex bandwidth
391.8 GB/s
GPU-to-GPU P2P gather bandwidth
399.6 GB/s
GPU-to-GPU P2P latency
827 ns
GPU-to-GPU P2P bandwidth
396.4 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
55.4 GB/s
Per-GPU host-to-GPU bandwidth
55.5 GB/s
Per-GPU host-device latency
1,188 ns
Per-GPU device memory copy bandwidth
2094.1 GB/s
Benchmark Environment
Driver Version
 595.71.05
CUDA Version (driver)
 13.2
nvbandwidth Version
 v0.8
GPU Count
 8
P2P Supported
Yes
Efficiency
SM/CE copy efficiency
0.93

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 1,978

Geekbench Multi-Core

Score: 19,316

Passmark CPU Scores

BENCHMARKSCORE
Mark
 134,560
Compression
 2,988,732
Encryption
 188,470
Extended Instructions
 204,838
Floating Point Maths
 609,670
Integer Maths
 861,641
Physics
 23,487
Prime Numbers
 1,611
Single Threaded
 3,208
String Sorting
 393,746

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,379
Database Operations
 34,910
Memory Latency
 59
Memory Read Cached
 27,814
Memory Read Uncached
 7,651
Memory Write
 7,542

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 223.28 37.46 0.17×
bitops
 163,388 16,340,479 100.01×
branch
 155,188 19,711,809 127.02×
funccall
 838.75 88,810 105.88×
funcret
 13,864 1,786,079 128.83×
hash
 197,191 23,585,327 119.61×
intmath
 71,499 6,903,286 96.55×
nop
 291,698 19,267,107 66.05×
prime
 519,621 50,361,830 96.92×
rotate
 109,354 9,612,814 87.91×
str
 988,849 115,131,322 116.43×
wcs
 760,282 70,239,267 92.39×
CryptographySingleMultiScaling
crypt
 3,650 354,634 97.17×
CompressionSingleMultiScaling
jpeg
 941.5 91,669 97.37×
zlib
 64.88 8,809 135.78×
HypervisorSingleMultiScaling
clock
 23,676 163,849 6.92×
clone
 457.91 15,890 34.70×
context
 5,758 782,945 135.98×
cpu-sched
 60.14 13,466 223.92×
eventfd
 164,697 2,770,115 16.82×
fault
 34,138 184,738 5.41×
fork
 3,913 61,628 15.75×
getrandom
 65,399 7,897,583 120.76×
hrtimers
 880,847 18,495,867 21.00×
mprotect
 722,016 17,068,310 23.64×
mtx
 16,198,347 5,019,211,862 309.86×
munmap
 3,277 72,050 21.98×
null
 991,802 129,062,890 130.13×
pipe
 755,464 134,639,698 178.22×
priv-instr
 926,093 8,932,914 9.65×
resched
 10,470,830 26,058,087 2.49×
switch
 292,128 15,195,291 52.02×
timermix
 224,128 33,501,724 149.48×
tlb-shootdown
 11,768 228,413 19.41×
tsc
 644,925 86,781,656 134.56×
usersyscall
 792,737 8,938,249 11.28×
vdso
 95,247,041 11,880,046,173 124.73×
vfork
 27,291 120,441 4.41×
x86syscall
 11,152,325 1,531,712,134 137.34×
yield
 222,094,167 242,027,137 1.09×
zero
 7,196,262 949,621,069 131.96×
CPU float/SIMDSingleMultiScaling
eigen
 17,976 1,894,350 105.38×
expmath
 3,909 467,418 119.57×
fma
 1,451,001 152,894,601 105.37×
fp
 231.26 27,674 119.67×
fractal
 7.2 851.96 118.33×
hyperbolic
 4,480 534,351 119.29×
logmath
 5,788 750,190 129.60×
matrix
 6,039 516,873 85.59×
matrix-3d
 798.1 20,787 26.05×
monte-carlo
 313.16 43,198 137.94×
mpfr
 1,938 175,867 90.75×
powmath
 2,179 270,882 124.32×
trig
 2,979 310,174 104.13×
veccmp
 27,771 2,515,692 90.59×
vecfp
 1,709 179,297 104.94×
vecmath
 4,366 500,974 114.76×
vecshuf
 918.47 85,832 93.45×
vecwide
 38,811 3,615,969 93.17×
AlgorithmsSingleMultiScaling
bitonicsort
 9.8 951.17 97.06×
bsearch
 320.28 36,753 114.75×
bubblesort
 1.2 89.6 74.67×
fibsearch
 363.5 36,880 101.46×
heapsort
 11.2 1,210 108.02×
hsearch
 9,169 895,047 97.62×
insertionsort
 14.2 1,050 73.91×
malloc
 3,048,625 154,346,450 50.63×
mcontend
 55.31 28,992 524.17×
mergesort
 34 3,822 112.42×
qsort
 24.2 2,641 109.13×
skiplist
 6,050 782,808 129.39×
tree
 1.8 178.79 99.33×
MemorySingleMultiScaling
cache
 193,442 8,693,650 44.94×
cachehammer
 196,597 7,713,335 39.23×
cacheline
 54.16 2,564 47.33×
l1cache
 1,227 24,722 20.15×
memcpy
 487.7 40,110 82.24×
memrate
 0.12 7.78 64.83×
memthrash
 522.16 1,052 2.02×
ptr-chase
 30,734,661 1,694,801,982 55.14×
spinmem
 254.19 9,634 37.90×
stream
 2.64 16,964 6425.89×
vm
 189,903 388,964 2.05×

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-highgpu-8g 2081872 GiB8
a3-megagpu-8g 2081872 GiB8
a3-edgegpu-8g 2081872 GiB8
a3-edgegpu-8g-nolssd 2081872 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

a3-ultragpu-8g FAQs