Value Analysis

Pricing & Price-Performance

Compare on-demand, spot, and committed-use pricing across all VM series. Find the best value for your workload.

Cheapest per vCPU

E2

$0.017/hr per vCPU (on-demand)

Best Spot Discount

~70%

Average spot savings across all series

3yr CUD Savings

~55%

Average committed-use discount vs on-demand

All VM Pricing

Representative machine types for each series. Region: us-central1. Click column headers to sort.

Machine Type
Series
vCPUs
Memory (GB)
On-Demand/hr
Spot/hr
1yr CUD/hr
3yr CUD/hr
e2-standard-4E2416$0.134$0.040$0.084$0.060
t2d-standard-4T2D416$0.154$0.046$0.097$0.069
n2d-standard-4N2D416$0.169$0.051$0.106$0.076
c3d-standard-4C3D416$0.192$0.058$0.121$0.086
n2-standard-4N2416$0.194$0.058$0.122$0.087
c3-standard-4C3416$0.209$0.063$0.132$0.094
e2-standard-8E2832$0.268$0.080$0.168$0.120
t2d-standard-8T2D832$0.308$0.092$0.194$0.139
n2d-standard-8N2D832$0.338$0.101$0.213$0.152
c3d-standard-8C3D832$0.384$0.115$0.242$0.173
n2-standard-8N2832$0.388$0.116$0.244$0.175
c3-standard-8C3832$0.418$0.125$0.263$0.188
e2-standard-16E21664$0.536$0.161$0.337$0.241
t2d-standard-16T2D1664$0.616$0.185$0.388$0.277
n2d-standard-16N2D1664$0.676$0.203$0.426$0.304
c3d-standard-16C3D1664$0.768$0.230$0.484$0.346
n2-standard-16N21664$0.776$0.233$0.489$0.349
g2-standard-4G2416$0.784$0.235$0.494$0.353
g2-standard-8G2832$0.918$0.275$0.578$0.413
c3-standard-22C32288$1.150$0.345$0.724$0.518
n2-standard-32N232128$1.553$0.466$0.978$0.699
g2-standard-24G22496$1.853$0.556$1.167$0.834
c3-standard-44C344176$2.300$0.690$1.449$1.035
c3d-standard-60C3D60240$2.880$0.864$1.814$1.296
h3-standard-88H388352$3.726$1.118$2.347$1.676
z3-standard-88Z388704$4.854$1.456$3.058$2.184
m3-ultramem-32M332960$5.851$1.755$3.686$2.633
z3-standard-176Z31761408$9.709$2.913$6.117$4.369
g2-standard-96G296384$10.416$3.125$6.562$4.687
m3-ultramem-64M3641920$11.702$3.511$7.372$5.266
m3-ultramem-128M31283904$23.790$7.137$14.988$10.706
a3-highgpu-8gA32081872$29.387$8.816$18.514$13.224

Price-Performance Rankings

Performance per dollar spent (higher = better value). Calculated as benchmark result divided by on-demand hourly cost.

Geekbench 6 Single-Core

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1C3c3-standard-81,850$0.4184,426
2E2e2-standard-81,180$0.2684,403
3T2Dt2d-standard-81,350$0.3084,383
4C3Dc3d-standard-81,680$0.3844,375
5N2Dn2d-standard-81,410$0.3384,172
6N2n2-standard-81,520$0.3883,918

Nginx Requests/sec

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1C3Dc3d-standard-876,000$0.384197,917
2C3c3-standard-882,000$0.418196,172
3T2Dt2d-standard-858,000$0.308188,312
4N2n2-standard-868,000$0.388175,258
5E2e2-standard-842,000$0.268156,716

PostgreSQL TPS

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1C3Dc3d-standard-86,800$0.38417,708
2C3c3-standard-87,400$0.41817,703
3N2n2-standard-85,200$0.38813,402
4E2e2-standard-82,800$0.26810,448
5Z3z3-standard-889,200$4.8541,895

FFmpeg H.265 Encoding

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1C3c3-standard-834$0.41881.3
2C3Dc3d-standard-830$0.38478.1
3N2n2-standard-822$0.38856.7
4E2e2-standard-812$0.26844.8

Llama 2 7B Tokens/sec

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1G2g2-standard-842$0.91845.8
2A3a3-highgpu-8g185$29.3876.3

LINPACK GFLOPS

#VM SeriesMachine TypePerformanceCost/hrPerf/$ (higher = better)
1H3h3-standard-882,150$3.726577
2C3Dc3d-standard-60920$2.880319.4
3C3c3-standard-44680$2.300295.7
4N2n2-standard-32380$1.553244.7

Spot vs On-Demand vs Committed

Spot instances offer the biggest savings (60-70%) but can be preempted. Committed-use discounts (CUDs) provide predictable savings for steady-state workloads.

On-Demand

Pay-as-you-go with no commitment. Best for variable workloads, development, and testing. Most expensive per hour but maximum flexibility.

Spot

60-70% discount but instances can be preempted with 30-second notice. Best for fault-tolerant batch processing, CI/CD, and ML training with checkpointing.

Committed Use (CUD)

1-year (37% off) or 3-year (55% off) commitments. Best for predictable production workloads. Applies automatically to matching usage.