Flavors¶
A flavor defines the size of your instance: how many vCPUs, how much RAM and how large the root disk is. You pick one when launching an instance.
Flavor names use a simple scheme:
- The prefix indicates the pool:
a1is our main pool,i1is an older pool with T4 GPUs. standardfor balanced compute,highmemfor memory-heavy workloads,hpcfor compute-heavy workloads,gpu-<model>for machines with NVIDIA GPUs.- The trailing number is the number of vCPUs (or, for GPUs, the number of GPUs).
- A
.500gbsuffix means the flavor has a 500 GB root disk instead of the usual 100 GB.
Note
This is a curated summary. When launching an instance, Horizon shows the currently available flavors. Flavors
ending in slurm.* and the amphora flavor are internal and not intended for your use.
Standard¶
Balanced CPU flavor, roughly 4 GB of RAM per vCPU (a1 pool).
| Flavor name | vCPUs | RAM | Root disk |
|---|---|---|---|
a1.standard.1 |
1 | 4 GB | 100 GB |
a1.standard.4 |
4 | 16 GB | 100 GB |
a1.standard.8 |
8 | 32 GB | 100 GB |
a1.standard.16 |
16 | 64 GB | 100 GB |
a1.standard.32 |
32 | 128 GB | 100 GB |
a1.standard.64 |
64 | 256 GB | 100 GB |
High memory¶
For memory-heavy workloads, roughly 8 GB of RAM per vCPU.
| Flavor name | vCPUs | RAM | Root disk |
|---|---|---|---|
a1.highmem.2 |
2 | 16 GB | 100 GB |
a1.highmem.4 |
4 | 32 GB | 100 GB |
a1.highmem.8 |
8 | 64 GB | 100 GB |
a1.highmem.16 |
16 | 128 GB | 100 GB |
a1.highmem.32 |
32 | 256 GB | 100 GB |
a1.highmem.64 |
64 | 512 GB | 100 GB |
HPC¶
Compute-heavy flavors with a high memory ratio, roughly 16 GB of RAM per vCPU.
| Flavor name | vCPUs | RAM | Root disk |
|---|---|---|---|
a1.hpc.2 |
2 | 32 GB | 100 GB |
a1.hpc.4 |
4 | 64 GB | 100 GB |
a1.hpc.8 |
8 | 128 GB | 100 GB |
a1.hpc.16 |
16 | 256 GB | 100 GB |
a1.hpc.30 |
30 | 480 GB | 100 GB |
a1.hpc.60 |
60 | 960 GB | 100 GB |
a1.hpc.120 |
120 | 1.9 TB | 100 GB |
GPU¶
GPU flavors attach one or more NVIDIA GPUs to the instance. The number in the name is the number of GPUs.
| Flavor name | vCPUs | RAM | Root disk | GPUs |
|---|---|---|---|---|
a1.gpu-a40.1 |
10 | 120 GB | 200 GB | 1x A40 |
a1.gpu-a40.2 |
20 | 240 GB | 200 GB | 2x A40 |
a1.gpu-a40.4 |
40 | 480 GB | 200 GB | 4x A40 |
a1.gpu-a40.4.500gb |
40 | 480 GB | 500 GB | 4x A40 |
a1.gpu-a40.8 |
80 | 960 GB | 100 GB | 8x A40 |
a1.gpu-v100.1 |
2 | 4 GB | 10 GB | 1x V100 |
a1.gpu-v100.4 |
28 | 720 GB | 100 GB | 4x V100 |
a1.gpu-v100.8 |
124 | 1.4 TB | 100 GB | 8x V100 |
a1.gpu-v100.16 |
108 | 1.4 TB | 100 GB | 16x V100 |
a1.gpu-h200.4 |
124 | 1.4 TB | 200 GB | 4x H200 |
a1.gpu-rtx6000.8 |
124 | 2.2 TB | 200 GB | 8x RTX 6000 |
i1.gpu-t4.1 |
2 | 80 GB | 100 GB | 1x T4 |
i1.gpu-t4.4 |
8 | 320 GB | 100 GB | 4x T4 |
i1.gpu-t4.9 |
32 | 720 GB | 100 GB | 9x T4 |
i1.gpu-t4.18 |
64 | 1.4 TB | 100 GB | 18x T4 |
Warning
All of our GPUs are currently in use, so GPU flavors are generally not available. If you need persistent GPU capacity, please contact us about hosting your own GPUs; we can then provision a GPU flavor for you.