Film renderingB200
Customer
PlaceholderCloud Servers
Placeholder quote. A customer explains, in two plain sentences, what changed once their film rendering workload moved to Nexpod.
Customer nameRole, Company
Read case studyUse case · Compute-heavy tasks
Batch jobs, done by dawn.
Rendering, simulation, genomics, transcoding. Queue the work, fan it out over dedicated 8-GPU servers, cap the spend in euros, and let the pool shut itself down when the last task lands.
4× B200 · €160.00 /hrcap in EURpool stops when done
jobs / render-batch-114
running
Frames
1,812
of 2,400 · 75%
Spend so far
€428.80
cap €2,000.00
rendered on a GPU queued
4× B200 · 32 GPUsETA 52 min
Teams running batch work on Nexpod
Placeholder- Customer
- CUSTOMER
- customer
- Customer Co
- customer.ai
- CUSTOMER
Customer stories
Batch work, finished on time.
Reserved capacity
Long run ahead? Lock in the rate.
Commit for 12 months and every GPU-hour costs €1.00 less, on every family. Reserved servers are held for the dates you book.
H100 · per GPU-hour
€2.00reserved
on demand €3.00
−€1.00 every hour
on demand · €3.00reserved · €2.00
month 1≈ €5,840 less per server-month12
01Throughput
Keep all eight GPUs at work.
Shard the job per device and watch utilisation per GPU in the console. If one card idles, you see it long before the invoice does.
- Per-GPU metrics
- Utilisation, memory and temperature, streamed every second.
- Large host memory
- Selected servers ship with up to 8192GB of system RAM.
GPU utilisation · b200-fra-01avg 95%
- GPU 096%
- GPU 194%
- GPU 297%
- GPU 391%
- GPU 495%
- GPU 593%
- GPU 698%
- GPU 792%
02Fan-out
Spread the queue across nodes.
Queue thousands of tasks and let workers pull them. Add nodes mid-run when the deadline moves, remove them when the queue drains.
- Work queues
- A task that fails on one node is retried on another.
- Elastic pools
- Grow or shrink a pool without restarting the job.
pool sim-sweep-22 · 4× H100
queue · sim-sweep-223,024 / 6,000
h100-ams-01912 done
h100-ams-02897 done
h100-ams-03904 done
h100-ams-04Added
failed tasks retry on another node4 nodes · 32 GPUs
03Spend
A budget the job cannot exceed.
Set a cap in euros per project. Nexpod alerts at the thresholds you choose and stops the servers at the cap, so an overnight run stays an overnight cost.
- Hard caps
- Pause or stop servers when a project reaches its limit.
- Threshold alerts
- Email and webhook alerts at the percentages you set.
Spend in € · project render-q4
€160.00/hr
cap €2,000
50% alert
80% alert
Developer tools
Drive batch jobs from code.
Create a pool, submit tasks and tear it all down from one script. Every status response carries the spend so far, in euros.
import nexpod client = nexpod.Client() pool = client.pools.create( family="B200", nodes=4, region="eu-fra-1", image="ghcr.io/your-org/render:2.4", budget_eur=2000,) pool.submit([{"frames": f"{i}-{i + 9}"} for i in range(0, 2400, 10)])pool.wait() # nodes stop when the queue is emptyPools API
Create a pool, submit tasks, read progress and spend.
$ nexctl pools status render-114Budgets
Caps and alerts per project, set and read in euros.
$ nexctl budgets set --cap-eur 2000Scheduled starts
Book a pool to start at a set time and stop at a set time.
$ nexctl pools create --start 22:00Start with compute-heavy tasks
Queue it tonight. Read results at breakfast.
Start a pool of 8-GPU servers, set a budget in euros and submit the work. The pool shuts down when the queue runs dry.
On-demand ratesEUR · excl. VAT
- B300· 288 GB HBM3e€6.00/GPU·hr
- B200· 180 GB HBM3e€5.00/GPU·hr
- H200· 141 GB HBM3e€4.00/GPU·hr
- H100· 80 GB HBM3€3.00/GPU·hr