Altos BrainSphereTM R680 F7
Built for AI Training and HPC
Ultra-High Performance 4U GPU Server
Powerful Architecture Reinvented for Compute, Storage, and Management
Dual Processor Architecture GPU Expansion Large-Capacity Memory
Supports dual 6th Gen Intel® Xeon® 6700/6500 Up to 8 PCIe GPUs, including NVIDIA H200 NVL, Up to 32 DDR5 DIMM slots, delivering up to 8000
Series, up to 350W TDP per CPU RTX PROTM 6000 Blackwell, L40S MT/s. Ideal for AI and HPC workloads
High-Speed Storage Enterprise-Grade Power & Cooling Remote Management & Security
Up to 12× 2.5-inch Gen5 NVMe / SATA / SAS-4 3+1 redundant 3000W titanium PSU, BMC with IPMI / HTML5 KVM,
drives. Hot-swappable design fanless design with high-efficiency cooling plus TPM 2.0 & UEFI Secure Boot
Evolved High-Performance AI GPU Server
Next-Generation Server Core for AI & HPC Extreme Compute Power for AI and HPC
Altos BrainSphereTM R680 F7 is built to handle demanding AI training, HPC, Equipped with Intel® Xeon® 6700/6500 series CPUs, Altos BrainSphereTM
and large-scale data workloads. Powered by dual 6th Gen Intel® Xeon® R680 F7 delivers strong per-core performance and AI acceleration. It
processors, DDR5 memory, and a 4U multi-GPU design, it delivers high- supports up to 8x dual-slot GPUs and is capable of hosting 8x NVIDIA H200
density performance and reliability. Ideal for cloud AI platforms and data NVL with 4- way NVLinkTM bridges—ideal for training large AI models,
centers, the R680 F7 offers the scalability and flexibility needed for real- simulations, and generative workloads. With 32 DDR5 RDIMM/MRDIMM
time analytics, deep learning, and large model training. slots supporting up to 8000 MT/s, it ensures high-throughput memory
access for latency-sensitive applications like deep learning and analytics.
Optimized Thermal, Power, and Manageability Flexible Architecture with Scalable Investment Efficiency
The 4U chassis maintains thermal stability under full load, with streamlined Altos BrainSphereTM R680 F7 offers a more flexible PCIe-based
airflow and tool-less fan modules. Redundant 3+1 Titanium 3000W PSUs architecture compared to HGX platforms, providing greater configuration
support efficient, continuous operation in dense deployments. It also scalability to balance performance and cost. This enables phased GPU
features the ASPEED® AST2600 BMC for remote management via IPMI, deployment, supports a broader range of accelerators, and adapts more
HTML5 KVM, and SNMP v3, along with built-in security features such as easily to evolving AI workloads. With the integration of Altos aiWorks
TPM 2.0 and UEFI Secure Boot. solution, the R680 F7 further enhances resource allocation, simplifies AI
workflow orchestration, and maximizes return on infrastructure investment.
Contact your Altos Partner now and accelerate your smart computing journey.
sales@[Link]
Altos BrainSphereTM R680 F7 Specification
Dual Intel® Xeon® 6 Processors, TDP up to 350W
CPU - Intel® Xeon® 6700-Series Processors
- Intel® Xeon® 6500-Series Processors
Up to 8 PCIe GPUs: NVIDIA H200 NVL (with a 4-way NVLink bridge), RTX PROTM 6000 Blackwell Server Edition*,
GPU
L40S
32 DIMM slots, 8-Channel DDR5 RDIMM/MRDIMM
Memory RDIMM Up to 6400 MT/s (1DPC), 5200 MT/s (2DPC)
MRDIMM Up to 8000 MT/s**
2 x 10GbE (Support NCSI function)
LAN
1 x 1 GbE Management port
TPM 1 x TPM2.0 module
Up to12 x 2.5"(Hot-swapable) Gen5 NVMe/SATA/SAS-4 (SATA and SAS drives require a storage controller card)
Drive Bay
M.2 support (optional M.2 PCIe card required)
I/O Ports 2 x USB 3.2 Gen1, 1 x VGA, 2 x 10GbE RJ45 ports, 1 x 1 GbE Management port
- 8 x FHFL x16 (Gen5 x16) for GPUs
PCIe Expansion Slots - 4 x FHHL x16 (Gen5 x16)
- 3 x LP x16 (Gen5 x16)
Power Supply 3+1 3000W 80 Plus Titanium redundant power supplies
Cooling Air cool
Operating: 10°C to 35°C, 8% to 80% (non-condensing)
Operating Properties
Non-operating: -40°C to 60°C, 20% to 95% (non-condensing)
ASPEED® AST2600 BMC
TPM 2.0 (FIPS, CC-TCG certification)
Security and Manageability
UEFI Secure
Boot SNMP v3 Support
4U
448 (W) x 176 (H) x 880 (H) mm
Form factor
448 (W)x 176(H) x 1000.5(H)mm (includes the dedicated fan module corresponding to the NVIDIA H200 NVL or RTX PROTM 6000
Blackwell server configuration)
OS Microsoft® Windows Server, Red Hat Enterprise Linux Server, Ubuntu Server LTS
MC / Safety Compliance FCC, CE
* Supports up to 8 × NVIDIA H200 NVL (2×4 NVLink) or 8 × NVIDIA RTX PROTM 6000; operation requires 25°C ambient temperature.
** MRDIMMs are only supported with Intel® Xeon® 6 Processors with P-cores and in a 1DPC configuration.
About Altos Computing Inc.
Altos Computing Inc. (abbr. Altos) is established in 2017, is a subsidiary of Acer Group. Leveraging in-house R&D and collaborations with ODM, IHV, and ISV partners, Altos delivers streamlined and cost-effective
integrated solutions. Its product portfolio covers AI servers, AI workstations, and AI resource management platforms.
In an era of rapid innovation, Altos provides leading solutions across diverse domains, including High Performance Computing (HPC), Virtual Desktop Infrastructure (VDI), Cloud Infrastructure, and Software-
Defined Storage (SDS). Our customers include government agencies, academic institutions, cloud service providers, data center operators, and enterprises.
In a continuing effort to improve the quality of our products, information in this brochure is subject to change without notice.
Images appearing are only representations of some of the configurations available for this model. Availability may vary depending on region. Altos disclaims any liability for errors and omissions in product
descriptions. © 2025. All rights reserved.
Altos and the Altos logo are registered trademarks of Altos Computing, Inc. Other trademarks, registered trademarks and/or service marks, indicated or otherwise the properties of their respective owners.