Dedicated CPU VPS with cores that are yours alone.
First month $1 on D-4, D-8 and D-16, then the normal monthly price. One per customer.
On the D-series each vCPU is pinned 1:1 to one physical AMD EPYC Genoa core, backed by DDR5 4800 MT/s ECC memory and NVMe RAID10 storage. Nobody else runs on that core. So the per-core speed you measure at 03:00 is what you get at 15:00 too. The images, the API and the full root are the same as on every other plan, and Amsterdam or New York costs one price.
Transparent pricing
Plans and prices
| Plan | vCPU | Memory | NVMe | Traffic | Monthly | 1 year | |
|---|---|---|---|---|---|---|---|
| Standard | |||||||
| S-2Standard | 1vCPU | 2 GBMemory | 30 GBNVMe RAID10 | UnlimitedTraffic | $7.99/moFirst month $1 | $7.19/mo on 1 year | Start for $1 |
| S-4Standard | 2vCPU | 4 GBMemory | 60 GBNVMe RAID10 | UnlimitedTraffic | $13.99/moFirst month $1 | $12.59/mo on 1 year | Start for $1 |
| S-8Standard | 3vCPU | 8 GBMemory | 120 GBNVMe RAID10 | UnlimitedTraffic | $24.99/moFirst month $1 | $22.49/mo on 1 year | Start for $1 |
| S-16Standard | 4vCPU | 16 GBMemory | 240 GBNVMe RAID10 | UnlimitedTraffic | $44.99/moFirst month $1 | $40.49/mo on 1 year | Start for $1 |
| S-32Standard | 8vCPU | 32 GBMemory | 480 GBNVMe RAID10 | UnlimitedTraffic | $84.99/moFirst month $1 | $76.49/mo on 1 year | Start for $1 |
| S-64Standard | 16vCPU | 64 GBMemory | 720 GBNVMe RAID10 | UnlimitedTraffic | $159.99/moFirst month $1 | $143.99/mo on 1 year | Start for $1 |
| Dedicated CPU | |||||||
| D-4Dedicated | 1vCPU | 4 GBMemory | 45 GBNVMe RAID10 | UnlimitedTraffic | $37.00/moFirst month $1 | $31.45/mo on 1 year | Start for $1 |
| D-8Dedicated | 4vCPU | 8 GBMemory | 90 GBNVMe RAID10 | UnlimitedTraffic | $69.00/moFirst month $1 | $58.65/mo on 1 year | Start for $1 |
| D-16Dedicated | 4vCPU | 16 GBMemory | 160 GBNVMe RAID10 | UnlimitedTraffic | $118.00/moFirst month $1 | $100.30/mo on 1 year | Start for $1 |
| D-32Dedicated | 6vCPU | 32 GBMemory | 220 GBNVMe RAID10 | UnlimitedTraffic | $215.00/mo | $182.75/mo on 1 year | Deploy D-32 |
| D-64Dedicated | 10vCPU | 64 GBMemory | 400 GBNVMe RAID10 | UnlimitedTraffic | $377.00/mo | $320.45/mo on 1 year | Deploy D-64 |
| High Memory | |||||||
| H-64High Memory | 8vCPU | 64 GBMemory | 480 GBNVMe RAID10 | UnlimitedTraffic | $249.00/mo | $211.65/mo on 1 year | Deploy H-64 |
| H-128High Memory | 16vCPU | 128 GBMemory | 960 GBNVMe RAID10 | UnlimitedTraffic | $459.00/mo | $390.15/mo on 1 year | Deploy H-128 |
| H-192High Memory | 24vCPU | 192 GBMemory | 1.4 TBNVMe RAID10 | UnlimitedTraffic | $669.00/mo | $568.65/mo on 1 year | Deploy H-192 |
Who needs a pinned core, and who does not.
Most workloads burst. Idle, then a spike, then idle again, and shared vCPU handles that pattern just fine, we say so on our own Standard pages. Where a dedicated core earns its price is work that never stops, or work where a single slow moment costs you something. A game tick that stutters, say.
Game servers under mods
Minecraft, Valheim and similar servers run their world on one thread, so tick rate depends on that core never being interrupted. A handful of friends on vanilla is fine on Standard; a modded server wants D-4 or D-8. See Minecraft VPS.
Databases with a latency budget
An API on Postgres or MySQL with a p99 target lives or dies by its worst moment. The average is irrelevant. Pinned cores take the neighbour effect out of that tail, which is where the complaints come from. On a dedicated instance our published pgbench median is 7,364 TPS.
CI runners, encoding, transcoding
GitHub Actions and GitLab runners, ffmpeg jobs, audio processing: all of them push every core to 100%. On D-8 or D-16 a job takes the same time every run. Not roughly. The same.
Headless browsers at scale
Playwright and Chrome workers want CPU that doesn't wobble, plus about 2 GB per concurrent browser. A couple of browsers fit on D-4. For a farm, D-16 or D-32.
D-4 to D-64: sizing, prices and how the tiers differ.
Monthly prices, unlimited traffic, one IPv4 and a /64 IPv6 included, and no setup fee. Amsterdam and New York cost the same.
| Plan | vCPU / RAM / NVMe | Monthly | Fits |
|---|---|---|---|
| D-4 | 4 GB and 45 GB NVMe on 1 pinned core | €34.00 ($37.00) | One modded game server, one trading bot, or a small Postgres |
| D-8 | 8 GB and 90 GB NVMe on 4 pinned cores | €64.00 ($69.00) | A CI runner, media encoding, or a busy game server |
| D-16 | 4 pinned cores with 16 GB and 160 GB NVMe | €109.00 ($118.00) | Production database, several browser workers |
| D-32 | 32 GB and 220 GB NVMe across 6 pinned cores | €199.00 ($215.00) | App and database on the same box, video pipelines |
| D-64 | 10 pinned cores, 64 GB of RAM, 400 GB NVMe | €349.00 ($377.00) | Browser farms, big builds, analytics that has a deadline |
The €1 first month covers D-4, D-8 and D-16. D-32 and D-64 bill at the normal price from day one, no exception. Pay annually and 15% comes off, pay for two years and it is 20%. Extra NVMe on the D-series costs €16.00 per 100 GB a month, an extra pinned vCPU is €16.00 a month, and an extra IPv4 is €5.00 a month plus a $10 setup. Need more RAM or vCPU later? We resize in place, one short reboot, and your data and IP stay put.
Two things we tell people before they order. D-8 and D-16 run on the same four cores, so moving between them buys you memory and disk and no extra CPU at all. And if what you actually need is 64 GB of RAM rather than ten pinned cores, look at Standard S-64. More vCPU, more disk, lower monthly price. Just not pinned.
Dedicated, Standard or High Memory
The iron is the same. All three tiers run on AMD EPYC with ECC memory and NVMe RAID10, on nodes where we deliberately keep some headroom unsold. Between the tiers only one thing changes, and that is how the CPU is allocated. Nothing else moves.
| Tier | CPU allocation | Memory | Best for |
|---|---|---|---|
| Standard (S-2 to S-64) | EPYC Milan, shared vCPU at a 1:4 ratio | DDR4 ECC | Dev boxes, web, small production, anything bursty |
| Dedicated (D-4 to D-64) | EPYC Genoa, each vCPU pinned 1:1 to a physical core | DDR5 4800 MT/s ECC | Continuous CPU, latency-sensitive work |
| High Memory (H-64 to H-192) | EPYC Milan, shared vCPU, with 8 GB RAM per vCPU | DDR4 ECC | In-memory databases and caches |
What does pinned actually mean? The hypervisor maps your vCPU to one physical core and puts nothing else on it. Ever. On a Standard plan a vCPU is a share of a core that up to four guests can draw on, which works out well for workloads that idle most of the day, and it is the reason S-4 (2 vCPU, 4 GB, 60 GB NVMe) costs €11.99 a month ($13.99). What you give up is consistency. When the neighbours get busy, single-thread speed can dip a little. On the D-series it can't, because there are no neighbours on that core. So don't read the tiers as a ladder. Dedicated is for work that runs flat out or carries a latency target. Standard is for everything else.
Setting it up: nothing special, on purpose.
Step 1: Order and choose an image
Images are the same set every other plan gets. Ubuntu 22.04 to 26.04, Debian 11 to 13, AlmaLinux, Rocky Linux, CentOS Stream, Oracle Linux, Fedora, openSUSE, or upload a custom ISO if none of those suit you. Cloud-init runs on first boot. And the server is online 30-60 seconds after payment. Not minutes.
Step 2: First login and updates
Step 3: Confirm what you got
Check the CPU model and the core count first. Then throw real load at it and watch steal time. It should sit at zero.
lscpu gives you the core count and the model name, and a look under /sys/devices/system/cpu tells you whether the guest sees a frequency governor at all. On our dedicated plans it does not, the host owns that.
Step 4: Take a baseline you can repeat
Run one single-thread test and one all-core test now, while nothing is broken and nobody is shouting, and write the numbers down. Something feels slow three months later? Run both again. On pinned cores the two sets should line up.
Single-thread first, then all cores, then the disk. Save the three numbers with the date. When something feels slow later, run the same three lines and compare.
On a D-8 or bigger you can also keep one process on its own core and everything else off it.
Compare with the 5-run medians on our benchmarks page, taken on a dedicated instance.
Step 5: Resize through the API when the job grows
One REST API across every tier, so your tooling carries over untouched. Add vCPU or RAM, reboot once, and the data and the IP are exactly where you left them. The clearest case we see is game server operators. They start on D-4. Then the mod list grows, and grows a bit more, and up they go to D-8.
Pitfalls we see, and how to avoid them.
Buying pinned cores for a RAM problem
If the app swaps or the database cache misses, more cores change nothing. Check free -m first; S-16 or S-32 often fixes it for less.
Expecting a single core to be a server
On D-4 you get one pinned core. One. A game world runs happily on it, and a bot does too. But stack a database, a web app and a build job on that single core at the same time and they'll spend the whole day elbowing each other for it, which is the ticket we get from people who sized on price. Anything parallel wants D-8 or up.
Ignoring disk on CI boxes
Runner caches, Docker layers and artifacts fill 90 GB fast. Prune with docker system prune -af on a schedule, or add NVMe at €16.00 per 100 GB.
Measuring once
Skip the baseline on day one and you have no proof later. What you have is a feeling, and nobody, not you and not us, can act on a feeling. So keep the sysbench numbers next to the deploy script.
Pick the region closest to your users.
Amsterdam to London measures 9.0 ms on our side, Amsterdam to Frankfurt 8.5 ms. Out of New York, Boston sits at 5-8 ms and Toronto at 10-15 ms.
Questions about Dedicated CPU VPS.
What does pinned 1:1 actually mean?
Every vCPU on your VPS maps to one physical EPYC Genoa core, and the hypervisor schedules no other guest on it. The whole core is yours, all day. So single-thread speed doesn't move with whatever the neighbours are doing, because on that core there are none. Standard plans share cores at a 1:4 ratio instead.
Is the Dedicated tier faster than Standard?
Per core, yes. More to the point, it stays faster hour after hour, and that steadiness is what you are paying for. Genoa cores, DDR5 4800 MT/s ECC memory, and nothing else scheduled on those cores. Is it worth it for a bursty workload that idles most of the day? Honestly, the gap is small there. For continuous work, or anything with a latency budget, that same gap is the entire reason the D-series exists.
Does the €1 first month apply to Dedicated plans?
On D-4, D-8 and D-16, yes. Once per customer, paid by card or iDEAL. D-32 and D-64 are the exception and bill at the normal monthly price from the first month. Either way, every first order sits under the 30-day performance guarantee.
Can I move from a Standard plan to Dedicated later?
Yes. There are two routes. Order the D-plan and migrate your stuff across yourself, or open a ticket and ask support what the options are for your account. Once you're on the D-series the story gets simpler, because from there RAM and vCPU resize in place with one short reboot.
Do I get the same images and features as other plans?
Yes. The same operating system images, custom ISO upload, cloud-init, VNC console, nested virtualisation and REST API, all of it. What differs is one layer down, in how the CPU and memory are allocated. Nothing else.
Deploy a measured VPS
From $6.79/mo on a 2-year term. KVM on AMD EPYC, NVMe RAID10, ECC memory. No setup fee, cancel monthly.
30-day performance guarantee. Not satisfied? Full refund, no admin fee. How it works