GPU & accelerator enclosures
High-airflow multi-GPU AI compute trays for training and inference racks — confirm PCIe generation, power budget, and cooling path with thermal program.
4 representative models — browse below or search by model. Drawings, rail kits, and export MOQ confirmed on inquiry.
GPU AI compute enclosures
Multi-GPU rack trays for training and inference — high airflow paths, accelerator-dense layouts, and factory rail options.
F-G587-H12-A8
F-G587-H12-A8 — High-airflow multi-GPU rack enclosure for AI training and inference workloads. Dimensions (mm): 870（L）*436（W）*222（H）. Confirm form factor, rails, and export MOQ on inquiry.
- Drive bays: supports up to12 drives 2.5-inch/3.5-inchSATA/NVME/SASdrive
- Fan mounts: supports 8 ×8080hot-swaphigh-performance fan
- Drawings, rail kits, and export packaging on inquiry
- Primary material: Ma Steel (eco-grade, anti-fingerprint1.2mm SGCC)
- Surface finish: textured black front panel assembly, other panels in natural finish
- GPU bays: supports 8 ×4090/5090blower cardsGPU cards or AIcompute accelerators
- Memory: 32 ×DDR5memory slots,supports DDR5 4400/4800/5600MHzmemory
- PCIe support: 21 ×MCI05.0 ports, supports10G/25G/40G PCIe expansion cards
- Other interfaces: 1 ×management NIC, 1 ×OCP3.0 ports, 2 ×USB3.0 ports
- PSU support: 4 ×CRPSstandard PSU,hot-swap support,3+1 redundancy,optional 2000W/2700W/3200W
- Front panel LEDs: ID LED*1, Power SWwith LED*1, HDD LED*1 , LAN LED*2
- System support: Windows Server 、Redhat Enterprise Linux Server Suse Enterprise Linux Server 、Ubuntu Server 、Debian Server、 Citrix XenServer 、Vmware
- Applications: big data, cloud compute, HPC, AI, rendering, deep learning, edge computeapplications
- Supplier program: Toploong (Shenzhen)
F-G791-H12-A8
F-G791-H12-A8 — High-airflow multi-GPU rack enclosure for AI training and inference workloads. Dimensions (mm): 910（L）*436（W）*311（H）. Confirm form factor, rails, and export MOQ on inquiry.
- Drive bays: supports up to12 drives 2.5-inch/3.5-inchSATA/NVME/SASdrive, internal: 2 × M.2 NVMe
- Fan mounts: supports 12 ×hot-swaphigh-performance fanmodules
- Drawings, rail kits, and export packaging on inquiry
- Primary material: Ma Steel (eco-grade, anti-fingerprint1.2mm SGCC)
- Surface finish: textured black front panel assembly, other panels in natural finish
- GPU bays: supports 8 ×4090/5090fan cardsGPU cards or AIcompute accelerators
- Memory: 32 ×DDR5memory slots,supports DDR5 4400/4800/5600MHzmemory
- PCIe support: 21 ×MCI05.0 ports, supports10G/25G/40G PCIe expansion cards
- Other interfaces: 1 ×management NIC, 1 ×OCP3.0 ports, 2 ×USB3.0 ports
- PSU support: 8 × totalCRPSAPSU, motherboardsmax 4 × CRPS standard PSU, GPU modulesmax 4 × CRPS PSU, hot-swap support, supports N+1 redundant
- Front panel LEDs: ID LED*1, Power SWwith LED*1, HDD LED*1 , LAN LED*2
- System support: Windows Server 、Redhat Enterprise Linux Server Suse Enterprise Linux Server 、Ubuntu Server 、Debian Server、 Citrix XenServer 、Vmware
- Applications: big data, cloud compute, HPC, AI, rendering, deep learning, edge computeapplications
- Supplier program: Toploong (Shenzhen)
G6958-E12
G6958-E12 — High-airflow multi-GPU rack enclosure for AI training and inference workloads. Dimensions (mm): 950（L）*445（W）*265（H）. Confirm form factor, rails, and export MOQ on inquiry.
- Drive bays: front12*3.5-inchdrive bays, SATA or NVMEbackplane (compatible with2.5-inch)
- Motherboard support: custom motherboards
- Fan mounts: Includedfront4 ×8038hot-swapfan bays, rear4 ×8038ball-bearingfan and internal4 ×8038hot-plugfan(high static pressure/high airflow)
- Drawings, rail kits, and export packaging on inquiry
- Primary material: Ma Steel (eco-grade, anti-fingerprint1.2mm SGCC)
- Surface finish: textured black front panel assembly, other panels in natural finish
- GPU bays: supports 8 ×dual-widthblowerGPU cards/gaming GPUs
- I/O expansion: 4 ×full-height PCI/PCIEexpansion slots, vertical mount
- External I/O: rear2 * USB 3.0, 1*VGA, 1Gigabit Ethernet port, 1*tpyc-c
- System controls: POWER LED*1，HDD LED*1，RESET SW*1，POWER SW*1，NIC LED*2
- PSU support: supports standard CRPSPSUmodule drives , 3+1power redundancy mode
- Applications: big data, cloud compute, HPC, AI, rendering, deep learning, edge computeapplications
- Supplier program: Toploong (Shenzhen)
G4838-SE12-HKW
G4838-SE12-HKW — High-airflow multi-GPU rack enclosure for AI training and inference workloads. Dimensions (mm): 830（L）*437.6（W）*176（H）. Confirm form factor, rails, and export MOQ on inquiry.
- Drive bays: front12 ×3.5 " /2.5"drive bays, optional 2 × 2.5" Nvme or SATA SSD
- Motherboard support: standard 12"*13" and smaller motherboards, motherboardsrequires at least with 2 ×PCIE4.0 X16 ports
- Fan mounts: Includedinternal8 ×8056hot-swapfan bays, rear5 ×8038ball-bearingfan(high static pressure/high airflow)
- Drawings, rail kits, and export packaging on inquiry
- Primary material: Ma Steel (eco-grade, anti-fingerprint1.2mm SGCC)
- Surface finish: textured black front panel assembly, other panels in natural finish
- GPU bays: supports 8 ×dual-widthblowerGPU cards RTX30/40 series (length 270MM)
- I/O expansion: 7 ×fronthalf-height PCI/PCIEexpansion slots, vertical mount, 2 ×rearfull-height PCI/PCIEexpansion slots
- External I/O: front2 * USB 3.0
- System controls: POWER LED*1，HDD LED*1，RESET SW*1，POWER SW*1，NIC LED*2
- PSU support: supports standard CRPSPSUmodule drives , 4+1power redundancy mode
- PCIe lanes: 2Broadcom expansion chipsupports 10 ×PCIE4.0 x16lanes , 1 × PCIE4.0 x16pass-through lanes
- Front VGA: motherboardsIOfront
- Applications: AIinference, rendering, HPC, deep learningapplications
Chassis and cabinet catalog live — programs from partner factories (manufacturer noted in specs). Confirm U-height, GPU count, rails, cabinet size, and export MOQ on inquiry.
