Exxact TensorEX 6U MI350X OAM Server - 2x AMD EPYC 9005/9004-Series processor - TS4-149591758
The TensorEX TS4-149591758 is a 6U rack mountable server supporting 2x AMD EPYC 9005/9004-Series processors, 24x DDR5 memory slots, and 8x AMD Instinct MI350X OAM GPU modules with GPU-GPU Infinity Links at 1075 GB/s.
5th Generation AMD EPYC 9005™ series CPU
Up to 192 Core, 382 Threads of Unprecedented Performance and Efficiency
Data centers requiring the best speed, security, and scalability gravitate to AMD EPYC™. The 5th Generation AMD EPYC™ delivers leadership memory bandwidth, capacity, and next-generation I/O with up to 160 PCIe 5.0 lanes and CXL 2.0 for system memory expansion. Train dense AI models, execute highly complex simulations, and power your zero-downtime data center with confidence.
- Highly density EPYC processors with up to 192 Zen 5c Cores (EPYC 9965) or 128 traditional Zen 5 cores (EPYC 9755).
- Extraordinarily performant EPYC processors boosting up to 5.00GHz on 64 cores (EPYC 9575F).
- Expansive stack of processors for targeting your unique workload
- Consolidate your data center and reduce your carbon footprint with core density and efficiency.
AMD Instinct MI350X OAM GPU
The AMD Instinct MI350X Platform integrates 8 connected MI350X GPU OAM modules via next‑generation AMD Infinity Fabric™ links, delivering up to 2.3 TB of HBM3E memory for ultra‑low‑latency, large‑scale AI workloads. Delivered on a dedicated system board, AMD Instinct MI350X leverages the CDNA™ 4 architecture and high‑bandwidth GPU‑to‑GPU interconnect exceeding 1 TB/s to enable advanced HPC and AI computing. Purpose‑built for modern data centers, MI350X accelerators are designed to train, deploy, and serve next‑generation artificial intelligence models while accelerating the most demanding high‑performance computing, simulation, and analytics workloads.
MI350X OAM GPU
| Model | HBM Memory | HBM Bandwidth (TB/s) | Peak Engine Clock (MHz) | Scale-up Interconnect (Infinity Fabric) | Host I/O | FP16 Matrix Peak (PFLOPS, w/ sparsity) | BF16 Matrix Peak (PFLOPS, w/ sparsity) | OCP-FP8 Matrix Peak (PFLOPS, w/ sparsity) | MXFP6 Matrix Peak (PFLOPS, w/ sparsity) | MXFP4 Matrix Peak (PFLOPS, w/ sparsity) | Explore |
|---|---|---|---|---|---|---|---|---|---|---|---|
| MI350X OAM | 288 GB HBM3E | 8 | 2200 | 7 links; 153.6 GB/s bidirectional per link (mesh) | PCIe Gen5 x16 | 4.6192 | 4.6192 | 9.2274 | 9.2274 | 9.2275 | --- |
About
The TensorEX TS4-149591758 is a 6U rack mountable server supporting 2x AMD EPYC 9005/9004-Series processors, 24x DDR5 memory slots, and 8x AMD Instinct MI350X OAM GPU modules with GPU-GPU Infinity Links at 1075 GB/s.
5th Generation AMD EPYC 9005™ series CPU
Up to 192 Core, 382 Threads of Unprecedented Performance and Efficiency
Data centers requiring the best speed, security, and scalability gravitate to AMD EPYC™. The 5th Generation AMD EPYC™ delivers leadership memory bandwidth, capacity, and next-generation I/O with up to 160 PCIe 5.0 lanes and CXL 2.0 for system memory expansion. Train dense AI models, execute highly complex simulations, and power your zero-downtime data center with confidence.
- Highly density EPYC processors with up to 192 Zen 5c Cores (EPYC 9965) or 128 traditional Zen 5 cores (EPYC 9755).
- Extraordinarily performant EPYC processors boosting up to 5.00GHz on 64 cores (EPYC 9575F).
- Expansive stack of processors for targeting your unique workload
- Consolidate your data center and reduce your carbon footprint with core density and efficiency.
AMD Instinct MI350X OAM GPU
The AMD Instinct MI350X Platform integrates 8 connected MI350X GPU OAM modules via next‑generation AMD Infinity Fabric™ links, delivering up to 2.3 TB of HBM3E memory for ultra‑low‑latency, large‑scale AI workloads. Delivered on a dedicated system board, AMD Instinct MI350X leverages the CDNA™ 4 architecture and high‑bandwidth GPU‑to‑GPU interconnect exceeding 1 TB/s to enable advanced HPC and AI computing. Purpose‑built for modern data centers, MI350X accelerators are designed to train, deploy, and serve next‑generation artificial intelligence models while accelerating the most demanding high‑performance computing, simulation, and analytics workloads.
MI350X OAM GPU
| Model | HBM Memory | HBM Bandwidth (TB/s) | Peak Engine Clock (MHz) | Scale-up Interconnect (Infinity Fabric) | Host I/O | FP16 Matrix Peak (PFLOPS, w/ sparsity) | BF16 Matrix Peak (PFLOPS, w/ sparsity) | OCP-FP8 Matrix Peak (PFLOPS, w/ sparsity) | MXFP6 Matrix Peak (PFLOPS, w/ sparsity) | MXFP4 Matrix Peak (PFLOPS, w/ sparsity) | Explore |
|---|---|---|---|---|---|---|---|---|---|---|---|
| MI350X OAM | 288 GB HBM3E | 8 | 2200 | 7 links; 153.6 GB/s bidirectional per link (mesh) | PCIe Gen5 x16 | 4.6192 | 4.6192 | 9.2274 | 9.2274 | 9.2275 | --- |