a2-megagpu-16g
Compute Engine · A2 family
The a2-megagpu-16g is a gpu & accelerated Compute Engine machine type with 96 vCPUs and 1,360 GiB of memory, 16 × NVIDIA A100 40GB, running on Intel Cascade Lake with up to 100 gbps networking, starting at $55.74 per hour on demand.
- vCPUs
- 96
- memory
- 1,360 GiB
- gpu
- 16× A100 40GB
- on-demand
- $55.74
- spot vm
- $30.18
- 1-yr committed use
- $35.12
dedicated
14.17 GiB per vCPU
640 GiB total
$40,690/mo · us-central1
−46% vs on-demand
−37% vs on-demand
a2-megagpu-16g price by region
3 regions · hourly list prices in USD · On-demand · Spot VM · 1-yr Committed use · 3-yr Committed use
| Iowaus-central1cheapestreferenceNorth America | $55.74 | $30.18 | $35.12 | $19.51 |
| Netherlandseurope-west4Europe | $56.63 | $26.28 | $35.67 | $19.82 |
| Singaporeasia-southeast1Asia | $60.48 | $36.20 | $37.75 | $21.26 |
Specifications
compute
- vCPUs
- 96
- Memory
- 1,360 GiB
- Memory per vCPU
- 14.17 GiB
- Processor
- Intel Cascade Lake
- CPU vendor
- Intel
- Architecture
- x86-64
- Clock speed
- 2.2 GHz (turbo 2.9 GHz)
- Burstable CPU
- No
storage & network
- Local storage
- Persistent disk (local SSD attachable)
- Network
- Up to 100 Gbps
- Peak bandwidth
- 100 Gbps
accelerators
- Model
- NVIDIA A100 40GB
- Count
- 16
- Memory (total)
- 640 GiB
- Architecture
- Ampere
- Vendor
- NVIDIA
google cloud details
- API name
a2-megagpu-16g- Family
- A2
- Category
- GPU & accelerated
- Generation
- Current
- Regions
- 3 (Asia, Europe, North America)
- Cheapest region
- Iowa · $55.74/hr
- CoreMark score
- 1,258,852
- Max persistent disks
- 128
a2-megagpu-16g equivalents on other clouds
closest match by vCPU, memory, CPU architecture, accelerator and class · Linux on-demand in each provider's reference region
| Provider | Instance | vCPU | Memory | Processor | On-demand | vs a2-megagpu-16g |
|---|---|---|---|---|---|---|
| Google Cloud | a2-megagpu-16g | 96 | 1,360 GiB | 16 × NVIDIA A100 40GB | $55.74 | – |
| AWS | p4d.24xlarge1152 GiB (1.2× less) · 8× GPU | 96 | 1,152 GiB | 8 × NVIDIA A100 40GB | $21.96 | -61% |
| Azure | ND96asr A100 v4900 GiB (1.5× less) · 8× GPU | 96 | 900 GiB | 8 × NVIDIA A100 40GB | $27.20 | -51% |
A2 family sizes
| Size | vCPU | Memory | Accelerator | Local storage | On-demand | Monthly | Spot | 1-yr |
|---|---|---|---|---|---|---|---|---|
| a2-highgpu-1g | 12 | 85 GiB | 1 × NVIDIA A100 40GB | – | $3.673 | $2,682 | $1.989 | $2.314 |
| a2-ultragpu-1g | 12 | 170 GiB | 1 × NVIDIA A100 80GB | 375 GB NVME | $5.069 | $3,700 | $2.977 | – |
| a2-highgpu-2g | 24 | 170 GiB | 2 × NVIDIA A100 40GB | – | $7.347 | $5,363 | $3.978 | $4.628 |
| a2-ultragpu-2g | 24 | 340 GiB | 2 × NVIDIA A100 80GB | 750 GB NVME (2 disks) | $10.14 | $7,400 | $5.954 | – |
| a2-highgpu-4g | 48 | 340 GiB | 4 × NVIDIA A100 40GB | – | $14.69 | $10,726 | $7.957 | $9.257 |
| a2-ultragpu-4g | 48 | 680 GiB | 4 × NVIDIA A100 80GB | 1.5 TB NVME (4 disks) | $20.28 | $14,801 | $11.91 | – |
| a2-highgpu-8g | 96 | 680 GiB | 8 × NVIDIA A100 40GB | – | $29.39 | $21,453 | $15.91 | $18.51 |
| a2-ultragpu-8g | 96 | 1,360 GiB | 8 × NVIDIA A100 80GB | 3 TB NVME (8 disks) | $40.55 | $29,602 | $23.81 | – |
| a2-megagpu-16g | 96 | 1,360 GiB | 16 × NVIDIA A100 40GB | – | $55.74 | $40,690 | $30.18 | $35.12 |
a2-megagpu-16g FAQ
How much does a2-megagpu-16g cost?
a2-megagpu-16g costs $55.74 per hour on demand with Linux in Iowa (us-central1), about $40,690 per month. Across 3 regions the Linux on-demand price ranges from $55.74 in Iowa to $60.48 in Singapore. Spot VM capacity costs $30.18 per hour (46% less), but can be reclaimed. A 1-yr Committed use commitment brings it to $35.12 per hour. With Windows it costs $60.16 per hour.
How many vCPUs and how much memory does a2-megagpu-16g have?
a2-megagpu-16g has 96 vCPUs and 1,360 GiB of memory (14.17 GiB per vCPU), running on Intel Cascade Lake. It is a gpu & accelerated machine type in the A2 family.
What GPU does a2-megagpu-16g have?
a2-megagpu-16g has 16 × NVIDIA A100 40GB with 640 GiB of accelerator memory in total.
What are the equivalents of a2-megagpu-16g on other clouds?
The closest matches by vCPU, memory, CPU architecture, accelerator and instance class are on AWS, p4d.24xlarge (96 vCPU, 1,152 GiB, $21.96/hour in US East (N. Virginia)); on Azure, ND96asr A100 v4 (96 vCPU, 900 GiB, $27.20/hour in East US).