A rackmount server is a computer designed to be mounted in a standardized metal frame called a rack. Unlike tower servers, rackmount servers are horizontal and slide into the rack on rails or ears, saving floor space and making it easy to manage many servers in one location — a data center, server room, or colocation facility.

Rackmount servers offer dense deployment (many servers per rack), easier cable management, shared power distribution, better airflow and cooling, and simplified physical access. They are the standard choice for data centers, colocation facilities, and any environment with more than a few servers.

A rack unit (abbreviated U or RU) is the standard unit of measure for rack-mounted equipment height. 1U = 1.75 inches (44.45 mm). A standard full rack is 42U tall (about 6 feet). The U measurement is defined by the EIA-310 standard, so equipment from any manufacturer fits any standard rack.

1U (1.75"): The most compact form factor — high density, typically 1-2 CPUs, 2-4 drive bays, and 1 PCIe slot. Ideal for web servers, load balancers, and edge nodes. 2U (3.5"): More room for drives (up to 8-12 bays), more PCIe slots, and better cooling. The most popular general-purpose form factor. 3U (5.25"): Less common; used for specialized storage or workstation-class rackmount systems. 4U (7") and above: Maximum expandability — 12-24+ drive bays, multiple PCIe slots, room for full-length GPUs and dual high-power CPUs. Ideal for HPC, AI/ML, and large storage servers.

A standard rack is 42U tall, so theoretically 42 × 1U servers, or 21 × 2U servers. In practice, racks also hold patch panels, power distribution units (PDUs), KVM switches, and cable management units, so the usable count is lower. A typical rack might hold 20-30 1U servers with proper power and cooling.

Rack depth is the front-to-back measurement of a server, typically 26-30 inches for standard servers. Shallow-depth servers (12-20 inches) are designed for telecom closets, wall-mount racks, and spaces where full-depth equipment won't fit. eRacks offers shallow-depth rackmount servers for these applications.

Rack ears (or mounting flanges) are the metal brackets on the front of a rackmount server that bolt directly to the rack rails. Rail kits are sliding assemblies that let a server slide in and out of the rack like a drawer, making it much easier to service. Most eRacks servers include mounting hardware; sliding rail kits are available as an option.

Linux is the dominant OS for servers worldwide: it's stable, secure, highly configurable, and free of licensing costs. It supports virtually every server workload, has excellent driver support for modern server hardware, and benefits from a vast open-source ecosystem. Linux servers have a lower total cost of ownership and are the standard choice for cloud, HPC, data centers, and enterprise IT.

eRacks can pre-install virtually any Linux distribution. Popular choices include Ubuntu Server, Debian, Rocky Linux, AlmaLinux, CentOS Stream, Fedora Server, openSUSE, and Arch Linux. We also support specialized distros for HPC (such as Rocks Cluster Distribution) and storage (such as TrueNAS SCALE). Contact us if you have a specific distribution or version in mind.

Yes. Every eRacks system is assembled, configured, burned-in for 72 hours under load, and Linux pre-installed before shipping. You receive a fully working, tested system ready to deploy.

An open source rackmount server is built with commodity, non-proprietary hardware and runs open source software — Linux or BSD. Unlike vendor-locked servers from large OEMs, open source rackmount servers can be serviced with standard parts, upgraded freely, and run any open source OS or application stack without licensing fees or proprietary management software. eRacks has specialized in open source rackmount servers since 1999.

Yes, the hardware is fully compatible with Windows Server. However, eRacks specializes in Linux and open source systems, so we ship with Linux pre-installed. You can install Windows Server yourself, or contact us to discuss mixed-OS configurations.

It depends on your workload. 1U is best for lightweight tasks (web, DNS, load balancing) where density matters most. 2U is the most popular general-purpose form factor — good for application servers, databases, and moderate storage. 4U+ is for GPU computing, HPC, large storage arrays, and workloads needing lots of PCIe expansion or drive capacity. Contact eRacks for a free configuration recommendation.

1U servers are ideal for web servers, DNS/DHCP servers, load balancers, monitoring nodes, edge/CDN nodes, and network appliances — any application where you need many servers in a small space. Their trade-off is limited drive capacity (2-4 bays) and fewer PCIe expansion slots compared to 2U or 4U.

2U servers are the most popular form factor for general-purpose rackmount computing. They're well suited for application servers, database servers, NAS/file servers, mail servers, and virtualization hosts. The extra 1.75 inches of height allows up to 12 drive bays and 3-4 PCIe slots vs a 1U.

4U servers offer maximum expandability. They're ideal for GPU computing (AI/ML training, rendering, scientific simulation), HPC clusters, large-capacity storage servers (24-60+ drive bays), and any workload requiring multiple high-power components. eRacks' BEHEMOTH systems are 4U+ and are configured for maximum compute and storage capacity.

AI/ML workloads are GPU-bound. Look for a 4U chassis that supports multiple full-length, high-wattage GPU cards (NVIDIA H100, A100, RTX series, or AMD Instinct). You'll also want fast NVMe storage (for data loading), large RAM (512GB+), and a high-core-count CPU. eRacks can build and configure GPU-optimized servers for AI/ML workloads — contact us for a quote.

A shallow-depth server is a rackmount system designed to fit in spaces where full-depth (26-30") equipment won't fit — telecom closets, office server cabinets, wall-mount racks, and retail point-of-sale environments. Typical shallow-depth servers are 12-20 inches deep. eRacks offers shallow-depth rackmount configurations.

Since 1999 — over 25 years. eRacks is one of the longest-running Linux rackmount server builders in the United States, with customers in research, government, film/VFX, finance, and enterprise IT worldwide.

eRacks builds custom, pre-configured Linux servers with free lifetime telephone support, no vendor lock-in (standard commodity components), Linux pre-installed and burn-in tested, and direct access to engineers who know your system. We don't sell Windows licenses or proprietary management software — we're pure open source.

Yes — every eRacks system is built to order. Use our online configurator to choose CPU, RAM, storage, OS distribution, and add-ons. You can also contact us for configurations beyond what's shown online, including fully custom chassis designs and special component requests.

eRacks servers include a 1-year warranty on parts and labor, plus free lifetime telephone support. Extended warranties are available. Contact us for details.

Yes. Our 4U+ chassis support full-length, dual-slot GPU cards including NVIDIA (H100, A100, RTX 4090, RTX 6000) and AMD Instinct/Radeon Pro series. We can configure systems with single or multiple GPUs. Contact us for current GPU options and availability.

Yes. eRacks has built clusters for university research labs, national laboratories, and enterprise HPC. We configure complete clusters with high-speed interconnects (InfiniBand, 10/25/100 GbE), job schedulers (SLURM, PBS), and parallel filesystems (Lustre, BeeGFS). Contact us for a cluster quote.

Yes. Most eRacks servers include an IPMI (Intelligent Platform Management Interface) module, which allows remote power control, serial-over-LAN console access, and hardware monitoring — even when the OS is down or the system won't boot. We use open-standard IPMI rather than proprietary management tools.

Most eRacks systems are built to order (BTO). Typical configurations with standard components ship within 15 business days of order confirmation. Priority treatment is given to orders with expedited shipping. Contact us for lead time on specific configurations or large orders.

eRacks ships fully insured via UPS, FedEx, and other major carriers. For international shipments and large rack orders, we can arrange freight forwarding and container shipping. Contact us for international shipping details.

eRacks performs a 72-hour burn-in procedure on every system, load-testing the CPU, memory, and storage under sustained stress. This identifies marginal components and ensures maximum reliability before the server leaves our facility.

Yes. Enter special requests in the Notes field of the product configurator, or email us at info@eracks.com. We frequently build custom configurations for customers with unique requirements — unusual chassis, specialized storage, specific OS configurations, or pre-installed software stacks.

Yes. eRacks provides free lifetime telephone support on all systems. Call us at +1 (408) 455-0010 or email info@eracks.com.

It depends entirely on the workload. A web server or DNS server may need only 16-32 GB. A database server typically needs 64-256 GB. A virtualization host should have as much RAM as possible — 256 GB to 2 TB is common. GPU/AI servers often pair 512 GB+ of RAM with multiple GPUs. Contact eRacks for a recommendation based on your workload.

eRacks recommends hardware RAID for production systems. Software RAID (mdraid, ZFS) is fine for development or home use, but hardware RAID cards provide battery-backed write caching, dedicated I/O processing, and are independent of the OS — meaning a OS crash won't corrupt your RAID set. ZFS is a notable exception and is excellent for storage servers.

Yes. eRacks uses non-proprietary, industry-standard components, so your system is upgradeable at reasonable prices using standard parts. You are not locked into buying upgrades from eRacks — any compatible component from any supplier will work.

IPMI stands for Intelligent Platform Management Interface. It's an out-of-band management standard that lets you monitor and control a server independently of its OS via a dedicated network port. With IPMI you can: remotely power on/off/reset the server, access a serial console (even during boot or OS failure), monitor temperatures, fan speeds, and voltages, and receive hardware alerts.

Power consumption varies widely by configuration. A lightly loaded 1U server might use 50-150W at idle and 200-400W under load. A dual-CPU 2U server typically uses 100-300W at idle and 400-800W under full load. GPU servers (4U with 2-4 GPUs) can draw 2000-4000W. eRacks can provide power estimates for specific configurations to help with rack power planning.

Video editing demands fast storage and lots of RAM. For 4K/8K post-production, look for a system with NVMe SSDs (for fast scratch and media drives), 64-256 GB of RAM, and a modern multi-core CPU. A GPU accelerates encoding/decoding in apps like DaVinci Resolve and Premiere Pro. eRacks general-purpose and workstation systems can be configured with NVMe storage, high-core-count CPUs, and NVIDIA RTX cards. Contact us with your editing software and resolution requirements for a tailored recommendation.

Standard rackmount servers are loud — their high-RPM fans make them unsuitable for quiet studio environments. eRacks' STUDIO line (eRacks/STUDIO3, STUDIO5, STUDIO7) is purpose-built for audio and video production, with noise-optimized cooling, low-noise fans, and desktop-style enclosures that can sit near your workstation. The eRacks/SILENCE and eRacks/QUIET series are also engineered for near-silent operation. All ship with Linux pre-installed. Prices start at $1,200.

AI/ML training is GPU-bound. You need a 4U+ system with multiple high-end GPU cards (NVIDIA H100, A100, or RTX 6000 Ada), fast NVMe storage for data loading, large RAM (256 GB+), and a high-core-count CPU. eRacks' BEHEMOTH systems are our largest, most powerful platforms and are ideal for serious AI/ML workloads. We can also build multi-GPU systems in smaller chassis. Contact us for a GPU cluster quote.

eRacks' INTELLINATOR and OPTERNATOR lines are specifically designed for HPC and scientific computing. They feature high-core-count Intel Xeon and AMD EPYC processors, large ECC RAM (up to hundreds of GB), fast interconnects, and Linux pre-installed with your choice of distribution. Systems start at $3,895. We also build complete HPC clusters with InfiniBand interconnects, SLURM job scheduling, and parallel filesystems.

eRacks' NAS rackmount server is a purpose-built network attached storage system starting at $1,895. For larger raw storage capacity, our DAS (Direct Attached Storage / JBOD) line ranges from 12 to 90 drive bays: the DAS12 starts at $4,495 and the DAS90 holds up to 90 drives. All run Linux with open source storage software (ZFS, Samba, NFS). Contact us for configuration options.

eRacks' FIREWALL system is our entry-level rackmount Linux firewall, starting at $1,495. For high-availability and fully redundant configurations, our TWINGUARD series (TWINGUARD, TWINGUARD.PRO, TWINGUARD.ENT) provides dual redundant systems with live failover using OpenBSD CARP and PFSync — so if one unit fails, the other takes over with no downtime. Prices start at $2,985. All run OpenBSD or Linux with open source firewall software.

eRacks offers dedicated NVR (Network Video Recorder), DVR (Digital Video Recorder), and HVR (Hybrid Video Recorder) rackmount systems. These are purpose-built for surveillance camera recording, with Linux pre-installed and open source surveillance software. Contact us for current pricing and camera capacity options.

For VMware ESXi, Proxmox, or KVM virtualization hosts, prioritize maximum RAM (the more VMs you run, the more RAM you need), a high-core-count CPU, and fast local storage. eRacks' ENTERPRISE and dual-CPU systems are popular choices for virtualization. We can configure systems with 256 GB to 2 TB of RAM for dense VM environments. Contact us with your VM count and workload for a recommendation.

3D rendering is CPU- or GPU-bound depending on your renderer. For CPU rendering (Blender Cycles, V-Ray, Arnold CPU), maximize core count — our INTELLINATOR series with 96-112 cores is ideal for render nodes. For GPU rendering (Blender OPTIX, Redshift, Octane), configure with multiple high-end NVIDIA RTX cards. eRacks can build identical render nodes in quantity for a render farm. Contact us for volume pricing.

For general web hosting and colocation, our 1U general-purpose VALUE ($795), SERVE ($995), and PREMIUM ($1,295) systems offer excellent density and value. They run Linux with Apache, Nginx, or any web stack you choose. For higher-traffic sites needing more CPU and storage, step up to the ENTERPRISE line. All include free lifetime telephone support.

eRacks' MINER6 and MINER8 are purpose-built cryptocurrency mining rigs designed to hold 6 or 8 GPU cards respectively in a rackmount form factor. They run Linux with your choice of mining software. Contact us for current pricing and GPU availability — the crypto market moves fast and component availability changes frequently.

Database servers need fast storage (NVMe SSDs), lots of RAM (for buffer pools and caching), and reliable ECC memory. For PostgreSQL, MySQL, or MongoDB, our ENTERPRISE line and dual-CPU configurations are popular. For very large databases (multi-TB), pair a compute server with our DAS storage expansion shelves. Contact us with your database engine and dataset size for a recommendation.

Edge and IoT deployments often need compact, power-efficient hardware that fits in non-data-center locations. eRacks' small form factor and shallow-depth systems are ideal — they fit in telecom closets, retail back-of-house racks, and branch office cabinets. All run Linux and can be pre-provisioned with your edge software stack.

Game servers vary widely in requirements. Most benefit from high single-core clock speed (Minecraft, Valheim, etc. are often single-threaded), fast NVMe storage for world data, and enough RAM for your player count. Our PREMIUM and ENTERPRISE systems are popular choices. For large multi-server game hosting operations, we can build and configure dedicated game server clusters. Contact us with your game engine and player count.

Yes. eRacks can build servers appropriate for HIPAA-compliant healthcare environments, including encrypted storage, auditing tools, and secure Linux configurations. We also offer network architecture consulting to help design compliant infrastructure. Note that HIPAA compliance is an organizational and procedural requirement as well as a technical one — contact us to discuss your specific requirements.

The eRacks/BEHEMOTH line is our most powerful offering. The BEHEMOTH-S and BEHEMOTH-E are massive multi-processor, high-density systems designed for the most demanding HPC, AI, and enterprise workloads. Contact us for current specifications and pricing.

A rackmount server is optimized for continuous 24/7 operation, remote management, and multi-user workloads. A rackmount workstation is designed for interactive use — it typically has a more powerful GPU, faster single-core CPU performance, more RAM, and quieter cooling, but may not run 24/7 unattended. eRacks offers both configurations.

An open-source rackmount server uses commodity, non-proprietary hardware components and runs open-source software (Linux or BSD). There are no vendor lock-in components, proprietary management software, or OS licensing fees. Any qualified technician can service it with standard parts. eRacks has built open-source rackmount servers since 1999.

Yes. eRacks can ship bare-metal systems with no OS installed, or with a minimal PXE boot environment. However, most customers take advantage of our Linux pre-installation and testing service — we install, configure, and burn-in test your OS before shipping, saving you setup time.

ECC (Error-Correcting Code) RAM detects and corrects single-bit memory errors on the fly, preventing data corruption and system crashes. All eRacks servers use ECC RAM. Non-ECC RAM (found in consumer desktops) is not suitable for production servers — a single memory error can corrupt data or crash the OS.

A dual-CPU (two-socket) server has two physical processor sockets, effectively doubling core count and memory capacity vs a single-CPU system. You need a dual-CPU server when your workload demands more than ~24 cores or more than ~512 GB RAM in a single server. They're common for large database servers, virtualization hosts, and HPC nodes.

Both are excellent choices. AMD EPYC generally offers more cores per dollar, more memory channels, and higher memory bandwidth — great for HPC, databases, and virtualization. Intel Xeon offers strong single-core performance, excellent ecosystem support, and some unique features (like Optane PMem). eRacks builds servers with both. Contact us and we'll recommend the best platform for your specific workload.

Standard rackmount servers use forced-air cooling with high-RPM fans — effective but loud. For quiet environments, eRacks' STUDIO and QUIET lines use noise-optimized cooling. For extremely high-density or high-TDP workloads (such as GPU clusters), eRacks also offers liquid-cooled configurations. Liquid cooling dramatically reduces noise and can handle much higher heat loads than air cooling.

A well-maintained rackmount server can run reliably for 5-10 years. eRacks systems use quality components and non-proprietary parts, so you can replace individual components (drives, RAM, PSU) rather than the whole server. Our free lifetime telephone support means we're available to help diagnose and resolve issues for the life of your system.