Any hardware, any port.
13 of them are on the shelf.
The 13 stock builds are what is racked and ready. Past that we assemble the machine you describe: Xeon or EPYC, GPU if the workload wants one, and a port up to 100 Gbps with nothing counting it.
- cpu
- Intel Xeon or AMD EPYC
- gpu
- available on a built-to-order machine
- port
- 1, 10, 25, 40, 100 Gbps, always unmetered
- on the shelf
- stock builds stop at 10 Gbps; 25, 40, 100 are assembled to order
- hardware
- Supermicro, Dell, Gigabyte; all of the hardware is ours, with a stock of spare parts
- disks
- any type and size the job needs: SSD, NVMe or HDD
- raid
- on two disks with Linux the default is a software mirror; with more disks, the RAID level you need, software or hardware
- addressing
- 2 × IPv4 + native IPv6 · extra IPv4 $5/mo
- access
- root and IPMI
- operating system
- any Linux distribution, mount your own ISO over IPMI, Proxmox or VMware ESXi
- windows
- installed on request · licence yours
- locations
- AMS (NL) · MEP (NL) · IAD (US East) · PHX (US West)
- delivery
- a few hours to 7 business days, tested in full before handover
Prices and the 13 stock configurations live on the pricing page; anything assembled to order is quoted with the build.
Off the shelf, or built to order.
A stock build is already racked, so it is the quick end of the handover range. There are 13 of them, the ports stop at 10 Gbps, and the price is on the price list.
A build to order is assembled to your description: Xeon or EPYC, a GPU if the workload wants one, your memory and disks, and a port up to 100 Gbps. It is quoted per build, and handover sits at the longer end, up to 7 business days.
The range is a range for one reason: whether the configuration you need is already on the shelf.
How to pick without a sales call.
- 01 ─ the job
Start from what the machine does.
Every workload below has its own page: what that job does to a server, what to look for in the machine, and where we draw the line. Read the one closest to yours before you read a spec.
- 02 ─ the port
Size the port from the peak.
With no counter on the port, the average stops mattering. What matters is the busiest hour you expect. Ports are 1, 10, 25, 40, 100 Gbps, and a wider port is a bigger pipe, not a bigger bill per gigabyte.
- 03 ─ the place
Pick the site nearest your users.
Each location has its own builds in stock and its own prices, so check the site nearest your users first. Two audiences on two continents usually means two machines.
- 04 ─ the doubt
If you are unsure, send the spec.
Describe the workload and where the users are. The answer is a machine, a port size and a price, not a call about your roadmap.
What people run on them.
What each workload is sized around. The servers that fit it today are marked on the price list.
| workload | sized around | what to look for |
|---|---|---|
| unmetered servers → | port | 1 to 100 Gbps, nothing counted |
| CDN origins → | port | 10 Gbps or wider |
| streaming servers → | cpu | cores for the ladder |
| game servers → | cpu | clock speed over core count |
| storage servers → | disks | large arrays, sized to the data |
| download servers → | port | sized to the release-day peak |
| database & compute → | disks | NVMe under the database |
| virtualisation hosts → | memory | first to run out |
| blockchain node servers → | disks | NVMe, IOPS first |