
Supermicro
Supermicro AS-4125GS-TNRT
4U dual-socket AMD EPYC GPU server taking up to eight double-width PCIe accelerators for AI training and inference.
| 供货状态 | 预订 |
|---|---|
| SKU | AXL-SRV-4125GS |
| 最低订购量 | 1 unit |
| 成色 | 全新 |
| 价格 | 按报价 |
概述
The AS-4125GS-TNRT is Supermicro 4U AMD EPYC GPU platform, holding up to eight double-width PCIe accelerators alongside two Socket SP5 processors. Compared with an SXM system it trades peak GPU-to-GPU bandwidth for flexibility: cards can be mixed or changed later, and the platform suits inference fleets, rendering, HPC and mid-sized training runs.
The PCIe topology matters on this chassis, so AXISLINK confirms whether your workload needs a direct-attach or switched layout before quoting, along with GPU model and count, memory, NVMe storage and networking. Builds are validated under sustained load rather than a quick power-on, the condition of the goods is stated on the quotation, and export documentation and freight from Hong Kong are arranged by our team.
规格参数
| GPU support | Up to 8x double-width PCIe 5.0 accelerators |
|---|---|
| Topology | Direct-attach and switched PCIe variants exist, confirmed per order |
| GPU model | Specified on the quotation, subject to the Supermicro support list |
| Processors | 2x AMD EPYC 9004 series, Socket SP5 |
|---|---|
| Newer silicon | EPYC 9005 series support confirmed per order and firmware level |
| Memory | 24 DDR5 DIMM slots, 12 memory channels per processor |
| Drive bays | Front hot-swap 2.5-inch bays, NVMe and SATA backplane options |
|---|---|
| Boot | M.2 NVMe boot devices available |
| Onboard LAN | Dual 10GBase-T |
|---|---|
| Add-in cards | PCIe 5.0 slots for InfiniBand or high-speed Ethernet adapters |
| Management | Onboard BMC with IPMI 2.0 and Redfish |
| Chassis | 4U rack |
|---|---|
| Cooling | Air cooled, front to rear airflow |
| Power | Redundant Titanium level power supplies sized for a full GPU load |
索取报价
Supermicro AS-4125GS-TNRT 的报价申请。我们通常在 24 小时内回复正式形式发票。
更多服务器产品
查看分类
Dell
Dell PowerEdge R660
1U dual-socket rack server on 4th/5th Gen Intel Xeon Scalable, built for dense virtualization, VDI and general-purpose compute.

Supermicro
Supermicro SYS-821GE-TNHR
8U GPU system built on the NVIDIA HGX H100 8-GPU baseboard with NVSwitch, for large model training and dense inference.

Dell
Dell PowerEdge XE9680 AI Server
Flagship 6U AI training server: 8x NVIDIA SXM GPUs on an HGX baseboard with NVSwitch, built for LLM training and dense inference.