NC96ads_A100_v4 by Microsoft Azure

NC96ads_A100_v4 is a NCADSA100v4 family [AMD processor] [Local Disk] [Premium Storage capable] 96 vCPUs server offered by Microsoft Azure with 96 vCPUs, 880 GiB of memory and 4.1 TB of storage. The pricing starts at 2.7151 USD per hour.
96 vCPU
880 GiB Memory
4.1 TB Storage
4 GPU
Spare SCore
247,032
(All-cores)
2,917
(Single-core)

Specifications

Server Metadata

Vendor ID
azure
Server ID
Standard_NC96ads_A100_v4
Name
NC96ads_A100_v4
API Reference
Standard_NC96ads_A100_v4
Description
NCADSA100v4 family [AMD processor] [Local Disk] [Premium Storage capable] 96 vCPUs
Family
NCADSA100v4
Hw Virt
Average Time To Start
104
Status
active
Observed At
2026-11-01T13:17:13.756880

Availability

REGION / IDSPOTONDEMAND
East US (US) / eastus
2.7151 USD/h 14.692 USD/h
East US 2 (US) / eastus2
2.9854 USD/h 14.692 USD/h
West US 2 (US) / westus2
2.9854 USD/h 14.692 USD/h
West US 3 (US) / westus3
2.9854 USD/h 14.692 USD/h
Central US (US) / centralus
3.3735 USD/h 16.602 USD/h
Canada Central (CA) / canadacentral
3.258 USD/h 17.63 USD/h
North Central US (US) / northcentralus
3.258 USD/h 17.63 USD/h
North Europe (IE) / northeurope
3.258 USD/h 17.63 USD/h
South Central US (US) / southcentralus
3.258 USD/h 17.63 USD/h
France Central (FR) / francecentral
3.7318 USD/h 18.365 USD/h
UK South (GB) / uksouth
3.7336 USD/h 18.365 USD/h
Germany West Central (DE) / germanywestcentral
3.5297 USD/h 19.1 USD/h
Spain Central (ES) / spaincentral
3.5297 USD/h 19.1 USD/h
Italy North (IT) / italynorth
3.8811 USD/h 19.1 USD/h
Poland Central (PL) / polandcentral
3.8811 USD/h 19.1 USD/h
Southeast Asia (SG) / southeastasia
3.8811 USD/h 19.1 USD/h
West Europe (NL) / westeurope
3.8811 USD/h 19.1 USD/h
West US (US) / westus
3.8811 USD/h 19.1 USD/h
Sweden Central (SE) / swedencentral
4.1237 USD/h 19.1 USD/h
Korea Central (KR) / koreacentral
3.6653 USD/h 19.834 USD/h
Central India (IN) / centralindia
3.8012 USD/h 20.569 USD/h
Jio India Central (IN) / jioindiacentral
- 20.569 USD/h
Jio India West (IN) / jioindiawest
- 20.569 USD/h
Switzerland North (CH) / switzerlandnorth
3.8826 USD/h 21.01 USD/h
Australia East (AU) / australiaeast
3.9368 USD/h 21.303 USD/h
Japan East (JP) / japaneast
3.9368 USD/h 21.303 USD/h
Brazil South (BR) / brazilsouth
5.4302 USD/h 29.384 USD/h

Processor

vCPUs
96
Hypervisor
Microsoft Hyper-V
CPU Allocation
Dedicated
CPU Cores
96
CPU Speed
3.7 GHz
CPU Architecture
x86_64
CPU Manufacturer
AMD
CPU Family
EPYC
CPU Model
7V13
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
512 KiB
CPU L2 Cache Total
48 MiB
CPU L3 Cache
32 MiB
CPU L3 Cache Total
384 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, constant_tsc, rep_good, nopl, tsc_reliable, nonstop_tsc, cpuid, extd_apicid, aperfmperf, pni, pclmulqdq, ssse3, fma, cx16, pcid, sse4_1, sse4_2, movbe, popcnt, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, cmp_legacy, cr8_legacy, abm, sse4a, misalignsse, 3dnowprefetch, osvw, topoext, perfctr_core, vmmcall, fsgsbase, bmi1, avx2, smep, bmi2, erms, invpcid, rdseed, adx, smap, clflushopt, clwb, sha_ni, xsaveopt, xsavec, xgetbv1, xsaves, user_shstk, clzero, xsaveerptr, rdpru, arat, umip, vaes, vpclmulqdq, rdpid, fsrm
Ecpus
84.7
Scalability
88.23

System Resources and Accelerators

MEMORY
Memory Amount
880 GiB
Memory Amount Actual
880 GiB
ACCELERATORS
Accelerator Count
4
Accelerator Memory Min
80 GiB
Accelerator Memory Total
320 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ampere
Accelerator Model
A100
Accelerators
  • NVIDIA Ampere NVIDIA A100 80GB PCIe (Memory amount: 81920, Firmware version: 595.71.05, BIOS version: 92.00.68.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100 80GB PCIe (Memory amount: 81920, Firmware version: 595.71.05, BIOS version: 92.00.68.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100 80GB PCIe (Memory amount: 81920, Firmware version: 595.71.05, BIOS version: 92.00.68.00.01, Clock rate: 1410)
  • NVIDIA Ampere NVIDIA A100 80GB PCIe (Memory amount: 81920, Firmware version: 595.71.05, BIOS version: 92.00.68.00.01, Clock rate: 1410)
STORAGE
Storage Size
4102 GB
Storage Type
nvme ssd
Storages
  • 262 GB ssd (temp disk)
  • 3840 GB nvme ssd
NETWORK
Inbound Traffic
0 GB/month
Outbound Traffic
0 GB/month
IPv4
0

CPU and System Topology

Server Description

A quad-GPU accelerated platform featuring dedicated AMD EPYC processors, massive NVMe storage, and high-capacity memory for parallel computing workloads.

GPU AcceleratedMemory OptimizedStorage & Database

Microsoft Azure NC96ads_A100_v4 is a GPU-accelerated server featuring 96 dedicated AMD EPYC 7V13 physical cores operating at 3.7 GHz, 880.0 GB of system memory, and four NVIDIA Ampere A100 GPUs with 320 GB of total VRAM. It includes 4102 GB of local NVMe SSD storage. Benchmarks show top-tier multi-core CPU performance, high memory bandwidth, and top-tier LLM inference speeds for 7B and 70B models, though single-threaded performance remains average. This hardware profile is optimized for parallel computing, large-scale machine learning, database operations, and ray tracing.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

3.9Score
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) 9,941,423 ops/sec 1,918,343 ops/sec 5.2 50.00% 50.00% +130.00%
Redis RPS (pipeline=16, SET) 78,543,415 ops/sec 11,704,734 ops/sec 6.6 20.00% 20.00% +46.00%
PassMark Memory Mark (composite) 3,339 2,416 1.4 10.00% 10.00% +3.40%
Memory bandwidth (read, 16 MB ~ L3) 390,270 MB/sec 108,466 MB/sec 3.5 10.00% 10.00% +13.00%
PassMark single-thread CPU 2,511 Mops/s 2,365 Mops/s 1 10.00% 10.00% 0.00%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
78.2 GB/s
Node GPU-to-host bandwidth
105.3 GB/s
Node host-to-GPU bandwidth
106.8 GB/s
Peer-to-peer
GPU-to-GPU P2P duplex bandwidth
260.5 GB/s
GPU-to-GPU P2P gather bandwidth
277.3 GB/s
GPU-to-GPU P2P latency
858 ns
GPU-to-GPU P2P bandwidth
277.3 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
26.3 GB/s
Per-GPU host-to-GPU bandwidth
26.7 GB/s
Per-GPU host-device latency
812 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Version (driver)
 12.6
nvbandwidth Version
 v0.5
GPU Count
 4
P2P Supported
Yes
Efficiency
SM/CE copy efficiency
1.00

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 1,731

Geekbench Multi-Core

Score: 17,610

Passmark CPU Scores

BENCHMARKSCORE
Mark
 94,593
Compression
 1,845,778
Encryption
 102,482
Extended Instructions
 130,957
Floating Point Maths
 386,385
Integer Maths
 477,337
Physics
 18,410
Prime Numbers
 1,401
Single Threaded
 2,511
String Sorting
 252,002

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 3,339
Database Operations
 29,092
Memory Latency
 55
Memory Read Cached
 25,789
Memory Read Uncached
 23,045
Memory Write
 20,496

Stress-ng Stressors

CPU integerSingleMultiScaling
atomic
 597.92 158.86 0.27×
bitops
 154,360 11,980,002 77.61×
branch
 180,769 14,626,091 80.91×
funccall
 578.48 45,636 78.89×
funcret
 18,327 1,436,293 78.37×
hash
 145,629 11,850,044 81.37×
intmath
 61,970 5,143,471 83.00×
nop
 305,565 25,415,133 83.17×
prime
 422,724 32,164,381 76.09×
rotate
 83,414 6,922,712 82.99×
str
 901,788 63,841,788 70.79×
wcs
 620,763 41,301,789 66.53×
CryptographySingleMultiScaling
crypt
 3,559 252,757 71.02×
CompressionSingleMultiScaling
jpeg
 757.16 54,028 71.36×
zlib
 50.59 4,634 91.60×
HypervisorSingleMultiScaling
clock
 3,704 98,116 26.49×
clone
 912.87 27,613 30.25×
context
 3,618 307,920 85.10×
cpu-sched
 59.89 5,939 99.16×
eventfd
 26,410 4,546,073 172.13×
fault
 8,943 138,696 15.51×
fork
 3,329 43,356 13.02×
getrandom
 48,372 3,778,104 78.11×
hrtimers
 68,654 1,960,314 28.55×
mprotect
 445,221 15,751,762 35.38×
mtx
 46,258,144 7,489,445,138 161.91×
munmap
 2,524 45,424 18.00×
null
 464,295 38,181,143 82.23×
pipe
 1,823,828 33,355,183 18.29×
priv-instr
 421,729 14,557,042 34.52×
resched
 5,604,951 8,283,990 1.48×
switch
 57,145 16,933,167 296.32×
timermix
 42,481 2,922,788 68.80×
tlb-shootdown
 8,591 247,353 28.79×
tsc
 729,863 59,596,860 81.65×
usersyscall
 335,179 12,617,874 37.65×
vdso
 85,760,825 6,860,877,433 80.00×
vfork
 18,022 101,793 5.65×
x86syscall
 5,910,676 484,059,687 81.90×
yield
 69,752,603 66,028,499 0.95×
zero
 3,083,507 255,147,135 82.75×
CPU float/SIMDSingleMultiScaling
eigen
 15,157 1,073,937 70.85×
expmath
 2,731 214,353 78.48×
fma
 834,947 65,007,645 77.86×
fp
 228.77 19,864 86.83×
fractal
 7.2 551.72 76.63×
hyperbolic
 3,504 291,619 83.22×
logmath
 4,094 330,583 80.75×
matrix
 2,821 219,184 77.70×
matrix-3d
 91.29 6,547 71.72×
monte-carlo
 215.57 17,586 81.58×
mpfr
 1,512 113,444 75.02×
powmath
 1,281 106,143 82.85×
trig
 1,597 127,288 79.71×
veccmp
 17,242 1,306,029 75.75×
vecfp
 1,584 122,459 77.33×
vecmath
 3,692 299,459 81.10×
vecshuf
 574.2 43,535 75.82×
vecwide
 25,519 1,759,387 68.94×
AlgorithmsSingleMultiScaling
bitonicsort
 9.4 665.17 70.76×
bsearch
 289.64 21,802 75.27×
bubblesort
 0.8 64.6 80.75×
fibsearch
 337.38 26,534 78.65×
heapsort
 9.8 756.37 77.18×
hsearch
 8,033 593,570 73.89×
insertionsort
 12 786.78 65.56×
malloc
 2,716,578 115,161,988 42.39×
mcontend
 56.91 10,403 182.79×
mergesort
 30.2 2,279 75.46×
qsort
 19.2 1,504 78.35×
skiplist
 6,779 519,688 76.66×
tree
 2 114.59 57.30×
MemorySingleMultiScaling
cache
 321,164 2,552,369 7.95×
cachehammer
 275,649 10,604,496 38.47×
cacheline
 17.79 2,292 128.81×
l1cache
 153.23 12,855 83.89×
memcpy
 390.69 27,936 71.50×
memrate
 3.15 36.71 11.65×
memthrash
 167.61 1,436 8.57×
ptr-chase
 69,137,843 975,027,043 14.10×
spinmem
 250.59 1,783 7.11×
stream
 27.86 28,342 1017.29×
vm
 195,387 305,486 1.56×

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
NC24ads_A100_v4 24220 GiB1
NC48ads_A100_v4 48440 GiB2

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

NC96ads_A100_v4 FAQs