AMD Dev Cloud Storage UI Limitations | Generated by AI

Home 2026.06

Question: Why can’t we see Block Storage / Volume options in the AMD Dev Cloud portal UI? Do we need to use the API instead?

Answer:

AMD Dev Cloud is powered by DigitalOcean infrastructure (you can see “Powered by DigitalOcean” at the bottom of their portal). However, AMD has a customized/restricted portal — they only expose a subset of DigitalOcean’s features in their UI. Block Storage Volumes is one of the features not surfaced in the AMD Dev Cloud portal UI.

Your options:

Option 1: Use the DigitalOcean API directly

AMD Dev Cloud gives you API access. Go to Settings → API → Tokens and generate a token. Then use the standard DigitalOcean API:

# Create a volume
curl -X POST https://api.digitalocean.com/v2/volumes \
  -H "Authorization: Bearer $AMD_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "size_gigabytes": 500,
    "name": "ml-data-vol",
    "region": "atl1",
    "filesystem_type": "ext4"
  }'

# List volumes
curl -X GET https://api.digitalocean.com/v2/volumes \
  -H "Authorization: Bearer $AMD_API_TOKEN"

# Attach volume to a droplet
curl -X POST https://api.digitalocean.com/v2/volumes/actions \
  -H "Authorization: Bearer $AMD_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "attach",
    "droplet_id": YOUR_DROPLET_ID,
    "region": "atl1"
  }'

Option 2: Use doctl CLI

# Install
brew install doctl  # or snap install doctl on Ubuntu

# Auth with your AMD token
doctl auth init --access-token $AMD_API_TOKEN

# Create volume in atl1
doctl compute volume create ml-data-vol \
  --size 500GiB \
  --region atl1 \
  --fs-type ext4

# Attach to droplet
doctl compute volume-action attach VOLUME_ID --droplet-id DROPLET_ID

# After attach, on the droplet:
# lsblk  → find the new disk (e.g. /dev/sda)
# mkfs.ext4 /dev/sda  (if new)
# mount /dev/sda /mnt/data

Likely limitation

Not all DigitalOcean regions support block storage, and ATL1 may have limited feature parity since it’s AMD’s dedicated datacenter. Check via API first:

curl -X GET "https://api.digitalocean.com/v2/regions" \
  -H "Authorization: Bearer $AMD_API_TOKEN" \
  | python3 -c "
import json, sys
regions = json.load(sys.stdin)['regions']
atl = [r for r in regions if 'atl' in r['slug']]
print(json.dumps(atl, indent=2))
"

Look at features in the ATL1 region response — if private_networking is there but storage is not, block storage isn’t available in ATL1 at all.

Practical alternative for ML workloads

If block storage isn’t available in ATL1, your best bet for large dataset storage is:

# Use Spaces (S3-compatible object storage) — always available via API
# Upload datasets
aws s3 cp ./dataset s3://your-space/ \
  --endpoint-url https://atl1.digitaloceanspaces.com

# Or rsync from your local machine to the droplet directly
rsync -avz --progress ./data/ root@<droplet-ip>:/mnt/data/

Check the AMD Dev Cloud support page or open a support ticket — they may enable volumes on request for MI300X droplets given the storage-heavy nature of ML workloads.

Back Donate