g5.12xlarge by Amazon Web Services

g5.12xlarge is a Graphics intensive Gen5 12xlarge server offered by Amazon Web Services with 48 vCPUs, 192 GiB of memory and 3.8 TB of storage. The pricing starts at 0.8238 USD per hour.
48 vCPU
192 GiB Memory
3.8 TB Storage
4 GPU
Spare SCore
18,431
(All-cores)
765
(Single-core)

Specifications

Server Metadata

Vendor ID
aws
Name
g5.12xlarge
Description
Graphics intensive Gen5 12xlarge
Family
g5
Hw Virt
Status
active
Observed At
2026-11-04T10:36:34.701988

Availability

REGION / IDSPOTONDEMAND
Oregon (US) / us-west-2
2.973767 USD/h 5.672 USD/h
Northern Virgina (US) / us-east-1
3.7072 USD/h 5.672 USD/h
Ohio (US) / us-east-2
3.153067 USD/h 5.672 USD/h
Aragón (ES) / eu-south-2
4.52875 USD/h 5.9765 USD/h
Stockholm (SE) / eu-north-1
1.3575 USD/h 6.0159 USD/h
Quebec (CA) / ca-central-1
3.64475 USD/h 6.2978 USD/h
Dublin (IE) / eu-west-1
2.748367 USD/h 6.3317 USD/h
Tel Aviv (IL) / il-central-1
5.205833 USD/h 6.6482 USD/h
Mumbai (IN) / ap-south-1
4.9411 USD/h 6.8109 USD/h
United Arab Emirates / me-central-1
2.7251 USD/h 6.9575 USD/h
Seoul (KR) / ap-northeast-2
3.572767 USD/h 6.9744 USD/h
Frankfurt (DE) / eu-central-1
4.260467 USD/h 7.0928 USD/h
London (GB) / eu-west-2
4.23775 USD/h 7.1999 USD/h
Paris (FR) / eu-west-3
7.1999 USD/h 7.1999 USD/h
Sydney (AU) / ap-southeast-2
7.3747 USD/h 7.3747 USD/h
Jakarta (ID) / ap-southeast-3
3.6567 USD/h 7.9385 USD/h
Tokyo (JP) / ap-northeast-1
5.7215 USD/h 8.2261 USD/h
Hong Kong (HK) / ap-east-1
4.91145 USD/h 8.7324 USD/h
Sao Paulo (BR) / sa-east-1
7.807 USD/h 9.6413 USD/h

Processor

vCPUs
48
Hypervisor
nitro
CPU Allocation
Dedicated
CPU Cores
24
CPU Speed
3.3 GHz
CPU Architecture
x86_64
CPU Manufacturer
AMD
CPU Family
EPYC
CPU Model
7R32
CPU L1D Cache
32 KiB
CPU L1D Cache Total
768 KiB
CPU L1I Cache
32 KiB
CPU L1I Cache Total
768 KiB
CPU L2 Cache
512 KiB
CPU L2 Cache Total
12 MiB
CPU L3 Cache
16 MiB
CPU L3 Cache Total
96 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, nonstop_tsc, cpuid, extd_apicid, aperfmperf, tsc_known_freq, pni, pclmulqdq, ssse3, fma, cx16, sse4_1, sse4_2, movbe, popcnt, aes, xsave, avx, f16c, rdrand, hypervisor, lahf_lm, cmp_legacy, cr8_legacy, abm, sse4a, misalignsse, 3dnowprefetch, topoext, ssbd, ibrs, ibpb, stibp, vmmcall, fsgsbase, bmi1, avx2, smep, bmi2, rdseed, adx, smap, clflushopt, clwb, sha_ni, xsaveopt, xsavec, xgetbv1, clzero, xsaveerptr, rdpru, wbnoinvd, arat, npt, nrip_save, rdpid
Ecpus
24.1
Scalability
50.21

System Resources and Accelerators

MEMORY
Memory Amount
192 GiB
Memory Amount Actual
192 GiB
Memory Generation
DDR4
Memory Speed
3200 Mhz
ACCELERATORS
Accelerator Count
4
Accelerator Memory Min
22 GiB
Accelerator Memory Total
89 GiB
Accelerator Type
GPU
Accelerator Manufacturer
NVIDIA
Accelerator Family
Ampere
Accelerator Model
A10G
Accelerators
  • NVIDIA Ampere NVIDIA A10G (Memory amount: 23028, Firmware version: 590.48.01, BIOS version: 94.02.75.00.01, Clock rate: 1710)
  • NVIDIA Ampere NVIDIA A10G (Memory amount: 23028, Firmware version: 590.48.01, BIOS version: 94.02.75.00.01, Clock rate: 1710)
  • NVIDIA Ampere NVIDIA A10G (Memory amount: 23028, Firmware version: 590.48.01, BIOS version: 94.02.75.00.01, Clock rate: 1710)
  • NVIDIA Ampere NVIDIA A10G (Memory amount: 23028, Firmware version: 590.48.01, BIOS version: 94.02.75.00.01, Clock rate: 1710)
STORAGE
Storage Size
3800 GB
Storage Type
nvme ssd
Storages
  • 3800 GB nvme ssd
NETWORK
Network Speed Baseline
40 Gbps
Network Speed Max
40 Gbps
Network Storage Speed Baseline
16 Gbps
Network Storage Speed Max
16 Gbps
Inbound Traffic
0 GB/month
Outbound Traffic
0 GB/month
IPv4
0

CPU and System Topology

Server Description

A GPU-accelerated platform combining four NVIDIA Ampere GPUs and local NVMe storage for demanding parallel computing and machine learning inference.

GPU AcceleratedStorage & DatabaseGeneral Purpose

Amazon Web Services g5.12xlarge is a graphics-intensive server powered by the Nitro hypervisor and an AMD EPYC 7R32 x86_64 processor with 48 vCPUs and 192.0 GB of DDR4 memory. It integrates four NVIDIA Ampere A10G GPUs with 89 GB of total VRAM and 3800 GB of local NVMe SSD storage, backed by 40 Gbps baseline network bandwidth. Benchmark results show a clear tradeoff: single-core CPU performance ranks in the bottom 10%, whereas GPU-driven LLM inference for prompt processing and text generation consistently reaches top-tier levels. Passmark memory and CPU encryption performance also rank in the top 25%. This configuration provides high resource density for workloads like large language model inference, multi-threaded data processing, and graphics rendering, offering a consolidated alternative to multi-node clusters.

Economics

Average Price per Region

Prices per Zone

Lowest Prices

Performance

Workload Profiles

1.7Score
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) 3,865,784 ops/sec 1,933,252 ops/sec 2.1 50.00% 50.00% +45.00%
Redis RPS (pipeline=16, SET) 25,860,205 ops/sec 11,794,699 ops/sec 2.2 20.00% 20.00% +17.00%
PassMark Memory Mark (composite) 2,887 2,412 1.2 10.00% 10.00% +1.80%
Memory bandwidth (read, 16 MB ~ L3) 101,115 MB/sec 108,490 MB/sec 0.91 10.00% 10.00% -0.94%
PassMark single-thread CPU 2,124 Mops/s 2,365 Mops/s 0.88 10.00% 10.00% -1.30%

GPU Bandwidth and Latency

Node (all GPUs)
Node baseline directional bandwidth
44.0 GB/s
Node GPU-to-host bandwidth
52.9 GB/s
Node host-to-GPU bandwidth
53.9 GB/s
Single GPU slot
Per-GPU GPU-to-host bandwidth
13.2 GB/s
Per-GPU host-to-GPU bandwidth
13.5 GB/s
Per-GPU host-device latency
925 ns
Benchmark Environment
Driver Version
 535.183.01
CUDA Version (driver)
 12.6
nvbandwidth Version
 v0.5
GPU Count
 4
P2P Supported
No
Efficiency
SM/CE copy efficiency
1.00

Memory Bandwidth

Compression

OpenSSL

Geekbench Single-Core

Score: 1,324

Geekbench Multi-Core

Score: 13,002

Passmark CPU Scores

BENCHMARKSCORE
Mark
 46,162
Compression
 711,299
Encryption
 51,146
Extended Instructions
 44,587
Floating Point Maths
 96,700
Integer Maths
 162,110
Physics
 4,566
Prime Numbers
 234
Single Threaded
 2,124
String Sorting
 92,793

Passmark Memory Scores

BENCHMARKSCORE
Memory Mark
 2,887
Database Operations
 16,787
Memory Latency
 52
Memory Read Cached
 24,631
Memory Read Uncached
 12,721
Memory Write
 13,501

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
g5.xlarge 416 GiB1
g5.2xlarge 832 GiB1
g5.4xlarge 1664 GiB1
g5.8xlarge 32128 GiB1
g5.16xlarge 64256 GiB1
g5.24xlarge 96384 GiB4
g5.48xlarge 192768 GiB8

Similar Servers

INSTANCEVENDORvCPUsMEMORYAccelerators

g5.12xlarge FAQs