Server Catalog

1U rack servers pack maximum compute density into a single rack unit (44.45 mm height), making them the default choice for hyperscale deployments, web-serving clusters, and latency-sensitive edge workloads where rack space is at a premium. A modern 1U platform typically supports dual Intel Xeon Scalable (Sapphire Rapids / Emerald Rapids) or dual AMD EPYC (Genoa / Bergamo) processors with up to 128 cores total, 4 TB of DDR5 ECC registered memory across 32 DIMM slots, and two to four NVMe U.2 or E1.S drive bays for fast local storage. Low-profile PCIe Gen5 expansion slots accommodate single-slot GPUs like the NVIDIA L4 or high-speed NICs such as the Mellanox ConnectX-7 for 100/200 GbE connectivity. The compact form factor does impose thermal constraints -- most 1U servers are limited to processors with a TDP of 270 W or below and cannot house full-height, full-length PCIe cards. Dual redundant power supplies (typically 800 W to 1,200 W 80 Plus Titanium) are standard. RawCompute supplies 1U platforms from Dell PowerEdge R660, Supermicro SYS-1029U, and HPE ProLiant DL360 Gen11 families, configured and tested to order.

2U rack servers offer the most popular balance between compute power, storage capacity, and expansion flexibility. With double the internal volume of a 1U chassis, a 2U platform can house up to 24 front-loading 2.5-inch NVMe/SAS/SATA drive bays (or 12 x 3.5-inch bays for high- capacity HDDs), full-height full-length PCIe cards, and higher-TDP processors up to 350 W. Typical configurations support dual AMD EPYC 9004 series or dual Intel Xeon Scalable 5th Gen processors, up to 6 TB of DDR5 memory, and six to eight PCIe Gen5 expansion slots. The extra airflow volume allows for quieter, more efficient cooling, making 2U servers suitable for environments without raised-floor cooling. For storage-forward workloads, 2U NVMe-optimised models like the Dell PowerEdge R760 or Supermicro SSG-2029P can deliver over 300 TB of raw NVMe flash in a single chassis. For compute-forward needs, models like the HPE ProLiant DL380 Gen11 support two double-wide GPUs alongside a generous CPU and memory footprint. RawCompute stocks a wide range of 2U servers and can pre-configure them with your choice of OS, firmware, RAID controller, and network cards before shipping anywhere in India.

Not every workload fits a standard SKU. RawCompute's custom-configured server service lets you specify every component -- from the chassis and motherboard platform down to the exact DIMM population, NVMe drive model, GPU accelerator, network card, and cable lengths. We start with a consultation to understand your workload profile (compute-bound, memory-bound, storage-bound, or GPU-bound), then recommend a bill of materials optimised for performance, power efficiency, and budget. Common custom requests include asymmetric DIMM configurations for memory-bandwidth- sensitive HPC codes, specific PCIe slot layouts for multi-NIC setups (e.g., dual Mellanox ConnectX-7 plus a Broadcom HBA), non-standard boot configurations (NVMe RAID-1 mirrors for OS, separate data pool), and direct liquid cooling (DLC) loops for high-TDP CPU and GPU deployments in Indian data centres where ambient temperatures can exceed 35 degrees Celsius. Every custom build goes through a multi-stage validation process: component-level verification, BIOS/UEFI firmware update to the latest stable release, 72-hour burn-in stress test under full load (using Prime95, FurMark, or DCGM for GPU systems), and a final thermal and power audit. We ship the server rack-ready with your chosen operating system pre-installed, iDRAC/IPMI/BMC configured with your network settings, and rail kits included.

GPU servers are purpose-engineered platforms designed to host between two and eight high-power GPU accelerators with the thermal headroom, PCIe bandwidth, and power delivery they demand. A 4U GPU server like the Supermicro SYS-421GE-TNRT or Dell PowerEdge XE9680 can house eight NVIDIA H100 SXM5 GPUs on an HGX baseboard with NVSwitch interconnect, dual high-core-count CPUs (AMD EPYC 9004 or Intel Xeon w9), up to 4 TB of DDR5 system memory, and 3,000 W+ redundant power supplies. The chassis design prioritises GPU-to-GPU communication bandwidth -- an HGX- based system provides 900 GB/s NVLink per GPU, while PCIe-attached multi-GPU systems use PCIe Gen5 switches to deliver 128 GB/s per slot. For inference-focused workloads, 2U servers can pack four to eight NVIDIA L4 or L40S cards with lower power requirements. Key considerations when selecting a GPU server include per-GPU TDP (up to 700 W for H100 SXM5), total system power draw (often 6-10 kW), cooling strategy (air vs. direct liquid cooling), and the GPU interconnect topology. RawCompute is an authorised partner for Supermicro, Dell, and ASUS GPU server platforms and provides end-to-end services including GPU installation, stress testing with DCGM diagnostics, thermal validation, and InfiniBand fabric integration for multi-node clusters. We also help customers plan power and cooling requirements with their colocation provider before deployment.

High-density storage servers are designed for workloads where raw capacity, sequential throughput, and cost-per-terabyte matter more than compute performance. These platforms typically offer 12 to 24 front-loading 3.5-inch LFF drive bays in a 2U chassis (e.g., Supermicro SSG-6029P-E1CR24H) or up to 60 bays in a 4U top-loading JBOD configuration (e.g., Supermicro SC946SE2C). Populated with 22 TB or 24 TB enterprise SATA HDDs from Seagate Exos or WD Ultrastar, a single 4U 60-bay server can deliver over 1.4 PB of raw storage. For performance-tier storage, the same bays can be fitted with NVMe U.2 SSDs behind a PCIe switch for all-flash configurations. Hardware RAID controllers (Broadcom MegaRAID 9600 series) or HBA pass-through cards (LSI 9500-series) connect the drives to the host CPU, while software-defined storage stacks like Ceph, MinIO, or TrueNAS SCALE handle data protection and distribution at the software layer. High-density storage servers are the backbone of on-premises object storage, backup repositories (Veeam, Commvault), media asset management libraries, and CCTV / surveillance recording systems. RawCompute supplies preconfigured storage nodes with drives installed, burned in, and tested, along with network cards for 10/25 GbE or 100 GbE backend connectivity to your compute cluster.