GPU-32xCPU-384GB-3xL40S by UpCloud

GPU-32xCPU-384GB-3xL40S is a GPU (32 vCPUs, 384 GiB RAM, 3x L40S) server offered by UpCloud with 32 vCPUs, 384 GiB of memory and 0 GB of storage. The pricing starts at 4.0635 USD per hour.
32 vCPU
384 GiB Memory
3 GPU
Spare SCore
132,460
(All-cores)
4,190
(Single-core)

Specifications

Server Metadata

Vendor ID
upcloud
Name
GPU-32xCPU-384GB-3xL40S
Description
GPU (32 vCPUs, 384 GiB RAM, 3x L40S)
Family
GPU
Hw Virt
Average Time To Start
62
Status
active
Observed At
2026-10-29T00:09:38.848900

Availability

REGION / IDSPOTONDEMAND
Helsinki #2 (FI) / fi-hel2
4.0635 USD/h 5.4124 USD/h

Processor

vCPUs
32
Hypervisor
KVM
CPU Allocation
Shared
CPU Cores
32
CPU Speed
2 GHz
CPU Architecture
x86_64
CPU Manufacturer
AMD
CPU Family
EPYC
CPU Model
9575F
CPU L1D Cache
64 KiB
CPU L1D Cache Total
2 MiB
CPU L1I Cache
64 KiB
CPU L1I Cache Total
2 MiB
CPU L2 Cache
512 KiB
CPU L2 Cache Total
16 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, mmx, fxsr, sse, sse2, ht, syscall, nx, mmxext, fxsr_opt, pdpe1gb, rdtscp, lm, rep_good, nopl, cpuid, extd_apicid, tsc_known_freq, pni, pclmulqdq, ssse3, fma, cx16, pcid, sse4_1, sse4_2, x2apic, movbe, popcnt, tsc_deadline_timer, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, cmp_legacy, cr8_legacy, abm, sse4a, misalignsse, 3dnowprefetch, osvw, perfctr_core, ssbd, ibrs, ibpb, stibp, vmmcall, fsgsbase, tsc_adjust, bmi1, avx2, smep, bmi2, invpcid, avx512f, avx512dq, rdseed, adx, smap, avx512ifma, clflushopt, clwb, avx512cd, sha_ni, avx512bw, avx512vl, xsaveopt, xsavec, xgetbv1, xsaves, avx_vnni, avx512_bf16, clzero, xsaveerptr, wbnoinvd, arat, avx512vbmi, umip, pku, ospke, avx512_vbmi2, gfni, vaes, vpclmulqdq, avx512_vnni, avx512_bitalg, avx512_vpopcntdq, la57, rdpid, movdiri, movdir64b, avx512_vp2intersect, arch_capabilities
Ecpus
31.6
Scalability
98.75

System Resources and Accelerators

MEMORY
Memory Amount
384 GiB
Memory Amount Actual
384 GiB
ACCELERATORS
Accelerator Count
3
Accelerator Memory Min
48 GiB
Accelerator Memory Total
144 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ada Lovelace
Accelerator Model
L40S
Accelerators
  • NVIDIA Ada Lovelace NVIDIA L40S (Memory amount: 46068, Firmware version: 595.84, BIOS version: 95.02.66.00.02, Clock rate: 2520)
  • NVIDIA Ada Lovelace NVIDIA L40S (Memory amount: 46068, Firmware version: 595.84, BIOS version: 95.02.66.00.02, Clock rate: 2520)
  • NVIDIA Ada Lovelace NVIDIA L40S (Memory amount: 46068, Firmware version: 595.84, BIOS version: 95.02.66.00.02, Clock rate: 2520)
STORAGE
Storage Size
0 GB
Storages
    NETWORK
    Inbound Traffic
    0 GB/month
    Outbound Traffic
    49152 GB/month
    IPv4
    1

    CPU and System Topology

    Server Description

    A high-density GPU server featuring triple NVIDIA L40S accelerators and 384 GB RAM for demanding machine learning and parallel computing workloads.

    GPU AcceleratedMemory OptimizedCompute Optimized

    UpCloud GPU-32xCPU-384GB-3xL40S is a GPU-accelerated server featuring 32 vCPUs powered by AMD EPYC 9575F processors and 384.0 GB of RAM. The system integrates three NVIDIA L40S GPUs from the Ada Lovelace architecture, providing a total of 144 GB of VRAM. Operating on the KVM hypervisor, the CPU utilizes a shared allocation model and supports AVX-512 instruction sets. Benchmark results place this server in the top 10% for single-core and multi-core CPU performance, as well as LLM inference speeds, where it processes 6061.83 tokens per second on a 7B model. A key hardware tradeoff is the absence of local storage, meaning all data persistence must be handled externally. This server is optimized for machine learning inference, data science, image processing, and parallel scientific computing.

    Economics

    Average Price per Region

    Prices per Zone

    Lowest Prices

    Performance

    Workload Profiles

    4.8Score
    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) 12,721,752 ops/sec 1,919,289 ops/sec 6.8 50.00% 50.00% +160.00%
    Redis RPS (pipeline=16, SET) 80,088,309 ops/sec 11,728,058 ops/sec 6.7 20.00% 20.00% +46.00%
    PassMark Memory Mark (composite) 2,859 2,416 1.2 10.00% 10.00% +1.80%
    Memory bandwidth (read, 16 MB ~ L3) 496,469 MB/sec 108,466 MB/sec 4.5 10.00% 10.00% +16.00%
    PassMark single-thread CPU 4,232 Mops/s 2,365 Mops/s 1.8 10.00% 10.00% +6.10%

    GPU Bandwidth and Latency

    Node (all GPUs)
    Node baseline directional bandwidth
    		67.9 GB/s
    Node GPU-to-host bandwidth
    		79.1 GB/s
    Node host-to-GPU bandwidth
    		80.6 GB/s
    Single GPU slot
    Per-GPU GPU-to-host bandwidth
    		26.4 GB/s
    Per-GPU host-to-GPU bandwidth
    		26.9 GB/s
    Per-GPU host-device latency
    		559 ns
    Benchmark Environment
    Driver Version
    		 595.84
    CUDA Runtime Version
    		 13.2
    nvbandwidth Version
    		 v0.8
    GPU Count
    		 3
    P2P Supported
    No
    Efficiency
    SM/CE copy efficiency
    		1.00

    Memory Bandwidth

    Compression

    OpenSSL

    Geekbench Single-Core

    Score: 3,003

    Geekbench Multi-Core

    Score: 27,001

    Passmark CPU Scores

    BENCHMARKSCORE
    Mark
    		 87,724
    Compression
    		 1,105,911
    Encryption
    		 56,047
    Extended Instructions
    		 97,809
    Floating Point Maths
    		 270,688
    Integer Maths
    		 272,400
    Physics
    		 16,555
    Prime Numbers
    		 1,047
    Single Threaded
    		 4,232
    String Sorting
    		 153,718

    Passmark Memory Scores

    BENCHMARKSCORE
    Memory Mark
    		 2,859
    Database Operations
    		 42,186
    Memory Latency
    		 88
    Memory Read Cached
    		 38,150
    Memory Read Uncached
    		 35,015
    Memory Write
    		 35,376

    Stress-ng Stressors

    CPU integerSingleMultiScaling
    atomic
    		 589.09 405.92 0.69×
    bitops
    		 253,958 7,919,321 31.18×
    branch
    		 200,522 6,174,836 30.79×
    funccall
    		 995.41 30,465 30.61×
    funcret
    		 29,367 901,742 30.71×
    hash
    		 240,693 7,469,389 31.03×
    intmath
    		 88,024 2,749,621 31.24×
    nop
    		 473,177 14,657,109 30.98×
    prime
    		 669,323 19,948,640 29.80×
    rotate
    		 157,115 4,930,636 31.38×
    str
    		 1,704,054 49,342,930 28.96×
    wcs
    		 1,150,876 32,254,177 28.03×
    CryptographySingleMultiScaling
    crypt
    		 6,019 178,488 29.66×
    CompressionSingleMultiScaling
    jpeg
    		 1,429 41,034 28.71×
    zlib
    		 85.87 2,432 28.32×
    HypervisorSingleMultiScaling
    clock
    		 81,779 251,155 3.07×
    clone
    		 438.19 10,218 23.32×
    context
    		 9,569 303,115 31.68×
    cpu-sched
    		 88.79 2,069 23.30×
    eventfd
    		 94,588 1,517,775 16.05×
    fault
    		 41,839 184,276 4.40×
    fork
    		 6,600 74,784 11.33×
    getrandom
    		 105,900 3,181,958 30.05×
    hrtimers
    		 134,944 4,566,679 33.84×
    mprotect
    		 697,793 14,368,027 20.59×
    mtx
    		 42,182,813 2,974,666,982 70.52×
    munmap
    		 5,199 56,943 10.95×
    null
    		 1,564,114 48,237,433 30.84×
    pipe
    		 695,742 21,478,537 30.87×
    priv-instr
    		 1,512,580 17,860,592 11.81×
    resched
    		 14,840,859 18,231,328 1.23×
    switch
    		 140,761 15,320,148 108.84×
    timermix
    		 365,603 10,961,063 29.98×
    tlb-shootdown
    		 24,830 243,144 9.79×
    tsc
    		 983,419 30,761,779 31.28×
    usersyscall
    		 1,186,473 17,595,309 14.83×
    vdso
    		 108,159,397 3,382,934,379 31.28×
    vfork
    		 39,432 173,686 4.40×
    x86syscall
    		 20,689,637 642,209,863 31.04×
    yield
    		 63,593,159 70,587,384 1.11×
    zero
    		 8,852,388 268,813,370 30.37×
    CPU float/SIMDSingleMultiScaling
    eigen
    		 19,416 601,827 31.00×
    expmath
    		 4,438 138,134 31.12×
    fma
    		 1,741,252 53,085,655 30.49×
    fp
    		 205.61 7,110 34.58×
    fractal
    		 9.8 303.43 30.96×
    hyperbolic
    		 5,938 187,544 31.58×
    logmath
    		 6,353 195,993 30.85×
    matrix
    		 10,182 292,953 28.77×
    matrix-3d
    		 1,656 12,677 7.65×
    monte-carlo
    		 505.8 15,279 30.21×
    mpfr
    		 2,652 81,024 30.55×
    powmath
    		 2,039 63,781 31.28×
    trig
    		 2,697 82,979 30.77×
    veccmp
    		 33,327 997,459 29.93×
    vecfp
    		 2,675 81,103 30.31×
    vecmath
    		 5,184 163,031 31.45×
    vecshuf
    		 1,106 33,237 30.05×
    vecwide
    		 43,758 1,289,842 29.48×
    AlgorithmsSingleMultiScaling
    bitonicsort
    		 14 416.79 29.77×
    bsearch
    		 404.34 12,141 30.03×
    bubblesort
    		 1.4 42.8 30.57×
    fibsearch
    		 475.03 14,648 30.84×
    heapsort
    		 15.8 464.79 29.42×
    hsearch
    		 14,443 428,566 29.67×
    insertionsort
    		 17.6 520.8 29.59×
    malloc
    		 5,256,287 64,536,719 12.28×
    mcontend
    		 109.46 5,639 51.52×
    mergesort
    		 42.6 1,290 30.29×
    qsort
    		 29 875.77 30.20×
    skiplist
    		 9,293 278,587 29.98×
    tree
    		 2.8 73.92 26.40×
    MemorySingleMultiScaling
    cache
    		 1,300,742 1,675,859 1.29×
    cachehammer
    		 281,670 7,687,930 27.29×
    cacheline
    		 68.95 3,103 45.00×
    l1cache
    		 754.57 22,387 29.67×
    memcpy
    		 903.89 27,112 29.99×
    memrate
    		 1.16 17.91 15.44×
    memthrash
    		 262.51 1,619 6.17×
    ptr-chase
    		 84,103,861 440,778,676 5.24×
    spinmem
    		 459.58 3,381 7.36×
    stream
    		 436.56 76,377 174.95×
    vm
    		 243,135 296,088 1.22×

    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-128GB-2xL40S 12128 GiB2
    GPU-16xCPU-80GB-1xRTXPRO6000 1680 GiB1
    GPU-16xCPU-192GB-1xL40S 16192 GiB1
    GPU-16xCPU-192GB-2xL40S 16192 GiB2
    GPU-16xCPU-192GB-3xL40S 16192 GiB3

    Similar Servers

    INSTANCEVENDORvCPUsMEMORYAccelerators

    GPU-32xCPU-384GB-3xL40S FAQs