GPU-96xCPU-960GB-4xB200 by UpCloud

GPU-96xCPU-960GB-4xB200 is a GPU (96 vCPUs, 960 GiB RAM, 4x B200) server offered by UpCloud with 96 vCPUs, 960 GiB of memory and 0 GB of storage. The last known price was 14.8392 USD per hour.
96 vCPU
960 GiB Memory
4 GPU
Spare SCore
193,394
(All-cores)
2,089
(Single-core)

Specifications #

Server Metadata #

Vendor ID
upcloud
Name
GPU-96xCPU-960GB-4xB200
Description
GPU (96 vCPUs, 960 GiB RAM, 4x B200)
Family
GPU
Series
B200
Hw Virt
Average Time To Start
89
Status
inactive
Observed At
2026-11-19T18:52:25.540150

Availability #

Region / IDSpotOndemand
Sydney #1 (AU) / au-syd1
14.8392 USD/h
19.7856 USD/h
Frankfurt #1 (DE) / de-fra1
14.8392 USD/h
19.7856 USD/h
Copenhagen #1 (DK) / dk-cph1
14.8392 USD/h
19.7856 USD/h
Madrid #1 (ES) / es-mad1
14.8392 USD/h
19.7856 USD/h
Helsinki #1 (FI) / fi-hel1
14.8392 USD/h
19.7856 USD/h
Amsterdam #1 (NL) / nl-ams1
14.8392 USD/h
19.7856 USD/h
Stavanger #1 (NO) / no-svg1
14.8392 USD/h
19.7856 USD/h
Warsaw #1 (PL) / pl-waw1
14.8392 USD/h
19.7856 USD/h
Stockholm #1 (SE) / se-sto1
14.8392 USD/h
19.7856 USD/h
Singapore #1 (SG) / sg-sin1
14.8392 USD/h
19.7856 USD/h
London #1 (GB) / uk-lon1
14.8392 USD/h
19.7856 USD/h
Chicago #1 (US) / us-chi1
14.8392 USD/h
19.7856 USD/h
New York #1 (US) / us-nyc1
14.8392 USD/h
19.7856 USD/h
San Jose #1 (US) / us-sjo1
14.8392 USD/h
19.7856 USD/h
Helsinki #2 (FI) / fi-hel2
16.1912 USD/h
23.1272 USD/h

Processor #

vCPUs
96
Hypervisor
KVM
CPU Allocation
Shared
CPU Cores
96
CPU Speed
2 GHz
CPU Architecture
x86_64
CPU Manufacturer
Intel
CPU Family
Xeon
CPU Model
6767P
CPU L1D Cache
32 KiB
CPU L1D Cache Total
3 MiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
3 MiB
CPU L2 Cache
4 MiB
CPU L2 Cache Total
384 MiB
CPU L3 Cache
16 MiB
CPU L3 Cache Total
16 MiB
CPU Flags
fpu, vme, de, pse, tsc, msr, pae, mce, cx8, apic, sep, mtrr, pge, mca, cmov, pat, pse36, clflush, dts, mmx, fxsr, sse, sse2, ss, ht, syscall, nx, pdpe1gb, rdtscp, lm, constant_tsc, arch_perfmon, pebs, bts, rep_good, nopl, xtopology, cpuid, tsc_known_freq, pni, pclmulqdq, dtes64, ssse3, fma, cx16, pdcm, pcid, sse4_1, sse4_2, x2apic, movbe, popcnt, tsc_deadline_timer, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, abm, 3dnowprefetch, cpuid_fault, ssbd, ibrs, ibpb, stibp, ibrs_enhanced, fsgsbase, tsc_adjust, bmi1, avx2, smep, bmi2, erms, invpcid, avx512f, avx512dq, rdseed, adx, smap, avx512ifma, clflushopt, clwb, avx512cd, sha_ni, avx512bw, avx512vl, xsaveopt, xsavec, xgetbv1, xsaves, avx_vnni, avx512_bf16, wbnoinvd, arat, avx512vbmi, umip, pku, ospke, waitpkg, avx512_vbmi2, gfni, vaes, vpclmulqdq, avx512_vnni, avx512_bitalg, avx512_vpopcntdq, la57, rdpid, bus_lock_detect, cldemote, movdiri, movdir64b, fsrm, md_clear, serialize, tsxldtrk, amx_bf16, avx512_fp16, amx_tile, amx_int8, flush_l1d, arch_capabilities
Ecpus
92.6
Scalability
96.46

System Resources and Accelerators #

Memory
Memory Amount
960 GiB
Memory Amount Actual
960 GiB
Accelerators
Accelerator Count
4
Accelerator Memory Min
192 GiB
Accelerator Memory Total
768 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Blackwell
Accelerator Model
B200
Accelerators
  • NVIDIA Blackwell NVIDIA B200 (Memory amount: 183359, Firmware version: 595.99.02, BIOS version: 97.00.FA.00.09, Clock rate: 1965)
  • NVIDIA Blackwell NVIDIA B200 (Memory amount: 183359, Firmware version: 595.99.02, BIOS version: 97.00.FA.00.09, Clock rate: 1965)
  • NVIDIA Blackwell NVIDIA B200 (Memory amount: 183359, Firmware version: 595.99.02, BIOS version: 97.00.FA.00.09, Clock rate: 1965)
  • NVIDIA Blackwell NVIDIA B200 (Memory amount: 183359, Firmware version: 595.99.02, BIOS version: 97.00.FA.00.09, Clock rate: 1965)
Storage
Storage Size
0 GB
Storages
    Compatible Storage IDs
    maxiops, standard, hdd
    Network
    Inbound Traffic
    0 GB/month
    Outbound Traffic
    81920 GB/month
    IPv4
    1

    CPU and System Topology #

    Server Description #

    A GPU-accelerated server featuring four NVIDIA Blackwell B200 GPUs and 960 GB of memory for parallel computing workloads.

    UpCloud GPU-96xCPU-960GB-4xB200 is a high-density GPU-accelerated server built on the KVM hypervisor with an x86_64 architecture. It features 96 shared vCPUs, 960.0 GB of system memory, and four NVIDIA Blackwell B200 GPUs delivering 768 GB of total VRAM. This configuration does not include local storage, requiring external storage attachments, and provides one complimentary public IPv4 address. While the 96 vCPUs use a shared allocation model, the massive system memory and GPU capacity make this server highly suitable for parallel processing tasks. It is designed for workloads such as large-scale machine learning, deep learning model training, and complex data science computations where GPU acceleration and large VRAM capacity are critical.

    Economics #

    Average Price per Region #

    Prices per Zone #

    Lowest Prices #

    Performance #

    Workload Profiles #

    12Score
    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.Partial coverage: missing component benchmark(s): PassMark Memory Mark (composite), PassMark single-thread CPU.
    Component Score Weight Impact
    Raw Reference Normalized Target Actual
    Redis RPS (pipeline=1, SET) 27,996,138 ops/sec 1,936,298 ops/sec 15 50.00% 62.00% +440.00%
    Redis RPS (pipeline=16, SET) 131,718,235 ops/sec 11,817,332 ops/sec 11 20.00% 25.00% +82.00%
    PassMark Memory Mark (composite) — 2,408 — 10.00% 0.00% —
    Memory bandwidth (read, 16 MB ~ L3) 671,917 MB/sec 108,674 MB/sec 6.1 10.00% 13.00% +27.00%
    PassMark single-thread CPU — 2,365 Mops/s — 10.00% 0.00% —

    GPU Bandwidth and Latency #

    Node (all GPUs)
    Node baseline directional bandwidth
    142.1 GB/s
    Node GPU-to-host bandwidth
    229.3 GB/s
    Node host-to-GPU bandwidth
    220.6 GB/s
    Peer-to-peer
    GPU-to-GPU P2P duplex bandwidth
    769.9 GB/s
    GPU-to-GPU P2P gather bandwidth
    795.3 GB/s
    GPU-to-GPU P2P latency
    1,607 ns
    GPU-to-GPU P2P bandwidth
    774.3 GB/s
    Single GPU slot
    Per-GPU GPU-to-host bandwidth
    56.9 GB/s
    Per-GPU host-to-GPU bandwidth
    55.4 GB/s
    Per-GPU host-device latency
    1,338 ns
    Per-GPU device memory copy bandwidth
    3199.0 GB/s
    Benchmark Environment
    Driver Version
    595.99.02
    CUDA Version (driver)
    13.2
    nvbandwidth Version
    v0.8
    GPU Count
    4
    P2P Supported
    Yes
    Efficiency
    SM/CE copy efficiency
    0.93

    Memory Bandwidth#

    Compression #

    OpenSSL #

    Stress-ng Stressors #

    CPU integerSingleMultiScaling
    atomic
    191.12 151.97 0.80×
    bitops
    164,225 14,976,548 91.20×
    branch
    153,953 14,083,890 91.48×
    funccall
    852.4 78,796 92.44×
    funcret
    13,982 1,294,356 92.58×
    hash
    189,959 17,349,814 91.33×
    intmath
    70,140 6,372,883 90.86×
    nop
    277,619 26,060,499 93.87×
    prime
    523,816 47,834,671 91.32×
    rotate
    108,588 9,959,093 91.71×
    str
    1,052,734 99,779,927 94.78×
    wcs
    799,233 66,771,773 83.54×
    CryptographySingleMultiScaling
    crypt
    3,717 339,395 91.32×
    CompressionSingleMultiScaling
    jpeg
    876.41 20,726 23.65×
    zlib
    65.06 5,725 87.99×
    HypervisorSingleMultiScaling
    clock
    31,506 85,964 2.73×
    clone
    242.85 9,849 40.56×
    context
    5,575 507,640 91.06×
    cpu-sched
    57.73 6,102 105.70×
    eventfd
    57,511 1,110,095 19.30×
    fault
    30,228 45,110 1.49×
    fork
    1,617 24,200 14.96×
    getrandom
    60,359 5,575,266 92.37×
    hrtimers
    361,296 3,939,956 10.91×
    mprotect
    310,494 5,028,229 16.19×
    mtx
    33,576,551 6,498,250,573 193.54×
    munmap
    1,346 29,633 22.02×
    null
    972,799 87,944,475 90.40×
    pipe
    213,535 27,234,657 127.54×
    priv-instr
    854,904 2,458,544 2.88×
    resched
    4,647,297 13,863,015 2.98×
    switch
    63,575 3,238,338 50.94×
    timermix
    260,972 16,881,723 64.69×
    tlb-shootdown
    3,704 191,369 51.67×
    tsc
    632,714 58,029,622 91.72×
    usersyscall
    764,920 2,692,964 3.52×
    vdso
    86,393,760 7,968,081,968 92.23×
    vfork
    4,253 49,088 11.54×
    x86syscall
    10,789,732 993,044,027 92.04×
    yield
    99,204,676 133,218,923 1.34×
    zero
    7,119,338 654,455,232 91.93×
    CPU float/SIMDSingleMultiScaling
    eigen
    17,260 1,648,488 95.51×
    expmath
    3,871 352,885 91.15×
    fma
    1,314,396 120,468,750 91.65×
    fp
    222.51 20,756 93.28×
    fractal
    7.6 680.88 89.59×
    hyperbolic
    4,485 411,469 91.74×
    logmath
    5,693 526,552 92.49×
    matrix
    5,467 529,183 96.80×
    matrix-3d
    466.33 23,167 49.68×
    monte-carlo
    313.06 28,750 91.83×
    mpfr
    1,915 175,516 91.66×
    powmath
    2,165 197,979 91.44×
    trig
    3,013 275,565 91.46×
    veccmp
    26,977 2,502,627 92.77×
    vecfp
    1,659 148,698 89.64×
    vecmath
    4,298 396,183 92.17×
    vecshuf
    892.49 82,118 92.01×
    vecwide
    36,255 3,036,883 83.76×
    AlgorithmsSingleMultiScaling
    bitonicsort
    10.6 960.78 90.64×
    bsearch
    339.26 31,247 92.10×
    bubblesort
    1.2 115.2 96.00×
    fibsearch
    329.05 30,019 91.23×
    heapsort
    11 1,010 91.78×
    hsearch
    9,300 880,414 94.67×
    insertionsort
    14 1,286 91.83×
    malloc
    1,770,932 38,032,182 21.48×
    mcontend
    23.29 9,284 398.61×
    mergesort
    33.2 3,061 92.20×
    qsort
    24 2,208 91.98×
    skiplist
    7,024 649,018 92.39×
    tree
    1.2 100.87 84.06×
    MemorySingleMultiScaling
    cache
    576,842 934,129 1.62×
    cachehammer
    112,001 5,102,059 45.55×
    cacheline
    25.96 1,833 70.62×
    l1cache
    338.8 26,737 78.92×
    memcpy
    764.43 68,500 89.61×
    memthrash
    190.33 2,055 10.80×
    ptr-chase
    14,823,170 669,463,042 45.16×
    spinmem
    219.59 5,957 27.13×
    stream
    101.99 320,551 3142.97×
    vm
    56,351 298,394 5.30×

    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#

    PostgreSQL heavy read-only throughput #

    Alternatives #

    Servers of the Same Family #

    INSTANCEvCPUsMEMORYAccelerators
    GPU-8xCPU-64GB-1xL40S 864 GiB1
    GPU-12xCPU-128GB-1xL40S 12128 GiB1
    GPU-12xCPU-240GB-1xH100 12240 GiB1
    GPU-12xCPU-128GB-2xL40S 12128 GiB2
    GPU-16xCPU-80GB-1xRTXPRO6000 1680 GiB1
    GPU-16xCPU-192GB-1xL40S 16192 GiB1
    GPU-16xCPU-192GB-2xL40S 16192 GiB2

    Similar Servers #

    INSTANCEVENDORvCPUsMEMORYAccelerators

    GPU-96xCPU-960GB-4xB200 FAQs #