Single-tenant dedicated server hosting with AMD EPYC and Intel Xeon processors, NVMe storage arrays, and 10 Gbps uplinks. No noisy neighbors, no virtualization tax — just hardware that's yours alone.
Monthly billing, no long-term contracts. Free DDoS protection up to 20 Gbps. Free IPMI access on every dedicated server.
A dedicated server is a physical computer rented entirely by one customer. Unlike a VPS, where the server's resources are split between multiple tenants through virtualization, a dedicated server gives you 100% of the CPU, 100% of the RAM, 100% of the disk, and 100% of the network interface. There are no other tenants on the box. There is no hypervisor between you and the hardware.
Dedicated server hosting is the right choice when you need predictable performance, regulatory isolation, hardware-level access (custom kernels, GPU passthrough, raw block devices), or simply more raw horsepower than virtualization can efficiently deliver. It's the closest you can get to running your own server in your own data center, without actually running your own data center.
Cloud VPS and bare metal solve different problems. Cloud VPS gives you flexibility — instant provisioning, easy scaling, automatic backups, and the ability to spin up and tear down servers programmatically. Bare metal hosting gives you predictability — guaranteed performance, hardware-level isolation, and no surprise throttling. Most modern infrastructure stacks combine both: bare metal for stateful, performance-critical components (databases, compute-heavy backends, GPU workloads), and cloud VPS for stateless, elastic components (web servers, queue workers, cron jobs).
Our customers commonly run hybrid architectures across both products. A typical pattern: a Premium dedicated server hosting the primary database, with a fleet of VPS instances running the application tier. Both products share the same network, the same control panel, and the same support team — so there's no operational complexity penalty for mixing them.
Most workloads do not need a dedicated server. A well-tuned VPS handles WordPress sites with millions of visitors, application backends serving thousands of requests per second, and game servers with dozens of concurrent players. VPS hosting is cheaper, faster to provision, and easier to scale.
Our customers commonly run hybrid architectures across both products. A typical pattern: a Premium dedicated server hosting the primary database, with a fleet of VPS instances running the application tier. Both products share the same network, the same control panel, and the same support team — so there's no operational complexity penalty for mixing them.
We deploy modern enterprise hardware exclusively. No refurbished consumer CPUs, no desktop motherboards rebranded as servers, no second-hand components from auction lots. Every dedicated server we ship is built on Supermicro, Dell PowerEdge, or HPE ProLiant chassis, with components chosen for sustained-load reliability rather than peak burst performance.
We pick CPU generations carefully. AMD EPYC has dominated server price-to-performance for the last three generations, and we've shifted most new deployments to EPYC accordingly. Intel still wins on single-thread performance for some workloads (notably gaming and certain databases), which is why we keep the Xeon E-2388G in our lineup. We don't deploy older Skylake or Cascade Lake systems — they're cheap, but the IPC difference matters for production work.
Every dedicated server ships with DDR4 ECC RAM in matched, paired DIMMs. ECC catches single-bit memory errors automatically and reports double-bit errors before they corrupt your data — essential for any production workload, and standard on every plan we sell. We do not sell non-ECC configurations because we don't believe non-ECC has a place in production hosting.
Storage is exclusively NVMe. We deploy enterprise U.2 drives from Samsung PM9A3, Micron 7400 PRO, and Kioxia CD8 series. These are not consumer drives — they ship with power-loss protection capacitors, sustained write endurance measured in petabytes, and 5-year manufacturer warranties. The performance difference between consumer and enterprise NVMe is invisible at low load and enormous at sustained high load; we choose enterprise drives because hosting workloads are sustained high load by definition.
Starter plan ships with a 1 Gbps NIC; Pro and Premium ship with dual 10 Gbps NICs in active-backup configuration. Premium plans can upgrade to 25 Gbps on request. All NICs are Intel or Mellanox hardware with full kernel driver support — we don't ship Realtek or other consumer-grade NICs even when the motherboard happens to include one.
All servers ship in 1U or 2U rack-mount chassis with redundant power supplies (1+1 hot-swap) on Pro and Premium plans. Drive bays are hot-swap on every plan. IPMI is on a dedicated management port physically separate from your data network — even if you misconfigure the data NICs and lock yourself out, the IPMI interface remains accessible.
Our dedicated server hosting network is multi-homed across Cogent, Lumen, NTT, Telia, GTT, and Hurricane Electric, with peering at major internet exchanges in every region. We carry IPv4 and IPv6 natively. Every server gets one primary IPv4 (with reverse DNS configurable from the panel) and a routed /64 IPv6 subnet. Additional IPv4 addresses are $2/mo each with light justification.
Bandwidth allocations:
Different workloads benefit from different hardware. Here's where bare metal pulls ahead of virtualized infrastructure:
Millions of monthly visitors. One big server beats five small ones operationally.
30-50% lower transaction latency than VPS. Direct NVMe access, no hypervisor scheduler.
64-128 player CS, large Minecraft networks. Xeon E-2388G is purpose-built for this.
NVIDIA L40S, H100, AMD MI300 GPUs available in Frankfurt, Dallas, Tokyo on request.
FFmpeg scales with cores and memory bandwidth. Premium dual-EPYC: ~30× a VPS.
HIPAA, PCI-DSS, regulated finance. Single-tenant by design. BAAs available.
Most dedicated servers ship within 2 hours of payment. Custom configurations (specific drive layouts, non-standard network setups, GPU additions) may take up to 24 hours. We email you when the server is ready, including SSH/RDP credentials and IPMI access details.
Default OS images include:
Custom ISO uploads are supported on every plan. Reinstallation through the panel takes 5-15 minutes depending on the OS.
Every plan ships unmanaged by default — you're responsible for OS-level administration, security patching, and application configuration. We don't lock down the server with proprietary tools; you have full root, full IPMI, and full freedom to manage it as you wish.
Managed services available as add-ons:
Every dedicated server includes free always-on Layer 3 and Layer 4 DDoS mitigation up to 20 Gbps. Scrubbing is performed at our network edge before traffic reaches your server. Common attack vectors — SYN floods, UDP amplification, ICMP floods, DNS reflection — are handled automatically without false positives.
For applications needing Layer 7 protection (HTTP floods, slowloris, application abuse), we offer DDoS Pro as an add-on at $50/mo. DDoS Pro adds a managed Web Application Firewall, intelligent rate limiting, and bot-challenge pages with sub-100ms decision time.
On the host side, dedicated customers get:
Every dedicated server hosting plan is backed by a 99.99% uptime SLA covering network, power, and our control plane. Hardware failures are remediated under a 4-hour replacement target — most failures are detected by our monitoring before customers notice. We keep spare parts (SSDs, DIMMs, NICs, PSUs) on-site at every facility for rapid swap.
Support is delivered by the same engineers who run the data centers. Median first response is 7 minutes. Coverage is 24×7×365. Premium customers get a named technical account manager and a direct phone number for escalation.
Most servers ship within 2 hours of payment confirmation. Custom configurations may take up to 24 hours. We email you the moment the server is ready with SSH credentials and IPMI access details.
Yes, on request. We stock NVIDIA L40S, H100, and AMD MI300 cards in select regions. Email sales@olivevps.com for current availability and pricing.
Yes. Every dedicated plan supports custom ISO uploads. We have customers running everything from Talos Linux to AIX to ESXi.
Yes. Managed services range from $25/mo for automated patching to $150/mo for full management. Email support for a custom managed plan.
Dedicated plans carry a 24-hour satisfaction window — if the server doesn't perform as advertised, we refund 100% within 24 hours of provisioning.
Yes, on a 25 Gbps port, subject to fair-use limits. We've never enforced fair use against a customer running a legitimate workload.
Yes. Both are available as default OS templates. ESXi requires a valid VMware license (vSphere Essentials or higher) which you bring yourself. Proxmox VE is fully open-source and ships ready-to-go. Many of our customers run their own private virtualization on top of our dedicated hardware to host their own internal VMs.
Yes. Standard remote hands tasks (reboot, recable, swap a drive) are included free for hardware-related failures. Custom requests (install a new component, run a specific diagnostic, photograph the server) are billed at $75/hour with a 15-minute minimum. Most requests complete within 2 hours of submission.
Yes — we offer colocation in our Frankfurt, Dallas, and Tokyo facilities at $89 per U per month with 1 Gbps included and 5A 120V power. Custom configurations (10 Gbps networking, redundant power, larger U commitments) are available on request.
Profitable proof-of-work mining typically requires GPUs at consumer electricity rates, neither of which we offer. We do see customers running cryptocurrency full nodes (Bitcoin, Ethereum, Monero), staking nodes, and validator infrastructure successfully on our dedicated servers — these workloads are storage-and-network bound rather than compute-bound, which our hardware handles excellently.
Our monitoring detects most hardware failures (failed drives, failing DIMMs, NIC errors) before customers notice. We open an incident ticket automatically, swap the affected component within our 4-hour SLA, and send you a post-incident report. Drive failures specifically are nearly invisible because we deploy in software RAID — your server keeps running while we hot-swap the failed drive.
Yes. We offer free private VLANs between any servers within the same region, with up to 10 Gbps of internal bandwidth. Cross-region private networking is available via WireGuard mesh or via dedicated layer-2 circuits (priced per region pair on request).
The Pro plan ($229/mo) ships with a single AMD EPYC 7402P (24 cores), 128 GB DDR4 ECC, 2× 2 TB NVMe in RAID 1, and a 10 Gbps uplink — sufficient for most production database, app server, and game server workloads. The Premium plan ($499/mo) doubles the EPYC count to 7763 (128 cores total), 512 GB RAM, 8× 3.84 TB NVMe drives, and a 25 Gbps uplink — designed for very large databases, ML inference clusters, video transcoding farms, or single-server deployments handling enormous traffic.
All hardware is covered under our standard SLA — failures are repaired or replaced within 4 hours of detection at no charge. We stock spare drives, DIMMs, NICs, and PSUs at every facility. For business-critical workloads, the Premium plan includes a hot-spare server kept on standby in the same rack, ready to swap in within 30 minutes if needed.