CrestVPS

FRA2 · DE · 50.11, 8.68

VPS hosting in Frankfurt, Germany

Our FRA2 point of presence occupies a facility in Frankfurt rated ISO 27001 / Tier IV. Every node uplinks at 2 x 25 Gbps into 14.2 Tbps of transit, filtered inline by 14 Tbps of scrubbing capacity.

FlagshipEuropeGPU

from€5.90/mo

CodeFRA2
FacilityISO 27001 / Tier IV
Uplink2 x 25 Gbps
Capacity14.2 Tbps
DDoS14 Tbps

Why Frankfurt

Frankfurt is where European routing actually converges. CrestVPS runs its own cage here, with GPU instances and the widest hardware choice in the fleet, which makes FRA2 the default whenever a workload needs both continental reach and accelerator capacity in the same rack.

How Frankfurt is connected

DE-CIX Frankfurt is the largest exchange on the continent by peak traffic, and we peer there alongside ECIX and KleyReX. That combination reaches most European eyeball networks without buying transit for them. Terrestrial fibre runs east to Warsaw and Vienna, south towards Milan, and north-west to Amsterdam and London, so onward latency stays short in every direction.

What FRA2 is good at

  • 01EU-facing production workloads
  • 02GPU training and inference jobs
  • 03Low-latency reach across the continent
Hosting under Germany lawGerman courts have long treated hosts as conduits rather than publishers, with duties triggered by specific notice instead of general monitoring. Blanket telecoms data retention has been suspended since 2017, and the CJEU later held the German rules incompatible with EU law; national statute adds detail on top of GDPR.

Measured latency

Frankfurt

Amsterdam6 ms
Paris11 ms
Warsaw18 ms
New York82 ms
Dubai108 ms

Round-trip time from Frankfurt to the listed hub, median over the last 30 days.

Peering and transit

DE-CIX FRAECIXKleyReX

Addressing

IPv4/20 owned
IPv6/29 allocated

Jurisdiction

EU / German TMG + BDSG. Strong host-provider liability shield.

Hardware available here

  • AMD EPYC 9754
  • AMD Ryzen 9 9950X
  • Intel Xeon 6544Y

We own and operate the hardware in this facility.

Pricing

Plans available in Frankfurt

Prices in Frankfurt are indexed at 100% of the base list, reflecting local transit and power cost.

Server familiesCPUvCPURAMMonthlyAnnual · −30%
PULSE
The fast entry point
AMD Ryzen 9 7950X28432 GB€5.90€4.13Deploy
TORQUE
Dense, dedicated, cheap per core
AMD EPYC 9754 Bergamo43216128 GB€24.90€17.43Deploy
NITRO
Highest clock we can buy
AMD Ryzen 9 9950X41616128 GB€32.90€23.03Deploy
FORGE
Memory first
Intel Xeon 6544Y42464512 GB€74€51.80Deploy
VAULT
Terabytes that stay cheap
AMD EPYC 9354412864 GB€21.90€15.33Deploy
FUSION
Whole GPUs, passed through
EPYC 9654 + NVIDIA accelerator82432256 GB€189€132.30Deploy

Fleet discounts stack on top: 3% from three servers, 6% from five, 10% from ten, applied to the whole invoice.

FAQ

Questions about Frankfurt

EU / German TMG + BDSG. Strong host-provider liability shield. We comply with valid legal process issued by that jurisdiction and nothing beyond it, and we publish the counts twice a year in our transparency report.

Yes. VPN and proxy endpoints are one of the most common workloads in every CrestVPS location, Frankfurt included, and nothing in the acceptable use policy restricts them. Tor middle relays and bridges are allowed everywhere; exit nodes are limited to Amsterdam, Reykjavik and Bucharest.

Yes, and in eight other assets across dozens of networks, settled through OxaPay. Card payment is not offered anywhere, which is what keeps signup down to an email address.

GPU availability is listed in the specification block at the top of this page. Where FRA2 carries them, the FUSION family passes an entire physical card through over PCIe Gen5 rather than slicing it, so the device you see is the device you have.

No. Registration in every location, Frankfurt included, is an email address and a password. We do not ask for identity documents, a phone number or a card anywhere in our footprint.

It depends what you are serving. FRA2 peers at DE-CIX FRA, ECIX, KleyReX, which sets how directly it reaches the networks your users sit on. The latency table above is measured, not modelled: read the row for the hub closest to your audience and decide from that rather than from the map.