5.3 GHz boost, not 5.3 GHz once.
The hosts are cooled to hold the boost clock under sustained load, not to touch it for four seconds in a benchmark and settle back to 4.2.
Ryzen 7000 hosts, clocked as high as the silicon goes. For the work that finishes when one core finishes, and not a moment before.
Size your server
HF 8
3 tiers
| Plan | vCPU | RAM | NVMe | Transfer | Port | Ready in | Price | Order |
|---|---|---|---|---|---|---|---|---|
| HF 4 Game servers, matching engines, bots | 4 @ 5.3 GHz | 16 GB | 320 GB | Unlimited | 10 Gbps | ~60 s | $29$23/mo | Deploy HF 4 |
| HF 8 Trading tooling, real-time APIs | 8 @ 5.3 GHz | 32 GB | 640 GB | Unlimited | 10 Gbps | ~60 s | $59$47/mo | Deploy HF 8 |
| HF 12 Simulation, compilation, single-thread heavy loads | 12 @ 5.3 GHz | 48 GB | 1 TB | Unlimited | 10 Gbps | ~60 s | $99$79/mo | Deploy HF 12 |
Every tier: Ryzen 7000 host · 1 IPv4 · routed IPv6 /64 · always-on DDoS filtering · 10 Gbps port · full API. Need a specific stepping? Ask — we build them.
Why clock still matters
Adding cores to a single-threaded job changes nothing. Game ticks, matching engines, interpreters, most compilers on most files — they finish when one core finishes, and the only lever is how fast that core runs.
5.3 GHz boost, not 5.3 GHz once.
The hosts are cooled to hold the boost clock under sustained load, not to touch it for four seconds in a benchmark and settle back to 4.2.
Ryzen 7000, no older stepping.
One generation of silicon across the fleet. You will not land on a five-year-old Xeon because that was the machine with room on it.
Fewer cores, on purpose.
You are paying for clock, so the tiers stay narrow and cheap. If your work does scale across cores, the standard line gives you more of them for less.
Ten-gigabit port from the floor.
A fast core waiting on a slow link is a slow server. The entry tier already gets the 10 Gbps port.
No upsells
No protected-IP tier, no per-snapshot charge, no support plan to buy.
Full root, KVM
Real hardware virtualisation. Your own kernel, Docker, nested VMs, sysctl. Not a container wearing a server costume.
IPv4 + a routed /64
One IPv4 on every plan, more on request, and a whole /64 of IPv6 so you never subnet a /128.
DDoS filtering
Layer 3/4 scrubbing upstream of the port, on by default, at no charge and with nothing to enable.
Two snapshot slots
Take one before you break something. Restore in under a minute, as often as you like.
Custom ISO and console
Mount any image, including your own hardened build, and watch it boot over the serial console.
API and SSH keys
Provision, resize, snapshot and destroy from your own code. Keys installed at order time, password login off.
Before you order
Anything else, support answers in minutes — and says no plainly when the answer is no.
Open a ticketIt is the boost clock the host holds under sustained single-thread load, measured on our own machines rather than copied from a spec sheet. Ask support for a trial and read it yourself with lscpu and a loop.
Shared, under the same published 4:1 cap as the standard line. If you need both the clock and the predictability, ask — we rack pinned high-frequency builds to order.
Sometimes. A single heavy query on one connection, yes. A hundred concurrent connections, no — that is core count and memory, so the reserved-core line will serve you better for the same money.
Because the money went into the clock. If you need both, the top tier here carries a terabyte, and support will quote a larger volume on any row.
The rest of the range
Same jurisdiction, same network, same refusal to ask who you are. They differ on the one thing named in each title.
Ready when you are
No account needed to see every price, no identity needed to order. $29 a month gets you a server nobody can talk us into switching off.