Author page description
May 29
- This is the NVIDIA MGX PCIe Switch Board with ConnectX-8 for 8x PCIe GPU Servers
➀ NVIDIA unveiled its new MGX PCIe Switch Board with ConnectX-8, designed for 8x PCIe GPU servers, marking a significant architectural shift in AI server infrastructure;
➁ The board integrates four PCIe Gen6 ConnectX-8 NICs, enabling 800Gbps networking and direct PCIe Gen6 x16 connectivity between GPU pairs, bypassing CPU limitations;
➂ This innovation allows NVIDIA to deploy PCIe Gen6 and high-speed networking capabilities independently of next-gen CPU platforms, accelerating AI factory scalability.
May 28
- Auras Shows Next-Gen Intel Oak Stream Cold Plates at Computex 2025
➀ Auras showcased its next-gen Intel Oak Stream cold plates for upcoming Xeon CPUs at Computex 2025, designed to support high thermal loads;
➁ The cold plates measure 156.0×107.5×24.2mm, larger than AMD’s SP7 counterparts, and feature a revised four-post retention mechanism;
➂ These designs highlight the growing adoption of liquid cooling in data centers, particularly for accelerated servers and high-density computing environments.
- Auras AMD SP7 Liquid Cooling Cold Plate for 600W CPUs Shown at Computex 2025
➀ Auras showcased its AMD SP7 liquid cooling cold plate for 600W CPUs at Computex 2025, targeting next-gen high-performance server processors;
➁ The compact cold plate measures 120.1×100.6×22.3mm and supports dual-socket designs with six mounting points, outperforming Intel's upcoming Oak Stream counterpart in size;
➂ AMD's SP7 platform, though not yet detailed publicly, is expected to power next-generation servers with 16-channel memory and higher core counts for advanced computing workloads.
May 27
- This is the ASPEED AST2700 Next-Gen BMC
➀ ASPEED unveiled its next-gen AST2700 BMC at Computex 2025, featuring upgraded quad Arm Cortex-A35 and dual Cortex-M4 cores for enhanced server management capabilities;
➁ The BMC integrates new 1Gbps LTPI interface, PCIe Gen4, DDR5 support, and Caliptra-based silicon root of trust security;
➂ Shown earlier than expected on Pegatron's OCP DC-SCM 2.1 module, it aims to counter emerging BMC competitors like Axiado with higher performance and integration.
May 26
- Teamgroup TForce Triple M.2 Liquid Cooler Shown at Computex 2025
➀ Teamgroup unveiled its TForce Triple M.2 liquid cooler at Computex 2025, designed to cool three M.2 SSDs with an AIO solution featuring a single-fan radiator and liquid cooling blocks;
➁ The compact design addresses thermal management challenges for clustered M.2 drives, though tube routing may complicate installation on motherboards with adjacent PCIe slots;
➂ While innovative, the product raises practicality concerns due to added cost and complexity for components with relatively low thermal demands.
May 25
- Intel Xeon 6 Priority Cores as a Big NVIDIA GPU AI Server Feature
➀ Intel Xeon 6 processors with High-Priority Cores (P-cores) are positioned as a key feature in NVIDIA DGX B300 AI servers, using the 64-core Xeon 6776P CPU for enhanced workload distribution;
➁ Intel markets its Xeon 6700P's 2DPC memory speed advantage (5,200 MT/s) over AMD EPYC 9005 (4,000 MT/s), though AMD offers more memory channels and capacity;
➂ NVIDIA's preference for Intel CPUs in GPU servers stems from reduced competitive overlap in AI accelerators, benefiting from Intel's canceled Rialto Bridge and Falcon Shores projects.
May 23
- Qualcomm Discrete NPU Spotted at in Dell Pro Max Plus laptop at DTW
➀ Dell showcased a prototype Pro Max Plus laptop with Qualcomm's discrete Cloud AI 100 NPU at its 2025 Tech World event, replacing traditional dGPUs for AI tasks;
➁ The dual NPU configuration offers 64GB unified memory, enabling local execution of 109-billion-parameter models like Llama 4 Scout, outperforming NVIDIA GPUs in memory capacity and power efficiency;
➂ This solution integrates with Dell's Pro AI Studio software, targeting developers needing mobile AI acceleration, potentially challenging NVIDIA's CUDA ecosystem and AMD's integrated NPU solutions.
May 22
- Maxsun Intel Arc Pro B60 Dual GPU 48GB at Computex 2025
➀ Maxsun and Intel showcased a dual GPU card at Computex 2025, combining two 24GB Intel Arc Pro B60 GPUs for 48GB total memory;
➁ The blower-style cooler design enables deployment in workstations/servers, supporting PCIe x8 bifurcation for efficient resource allocation;
➂ Targeting AI and graphics workloads, the solution offers cost advantages over NVIDIA's RTX 6000 Ada (48GB) at a projected lower price point.
May 21
- AMD Reveals Ryzen Threadripper 9000 Family: Big Zen 5 Chips for HEDT and Workstations
➀ AMD announced the Ryzen Threadripper 9000 family at Computex 2025, featuring Zen 5 architecture with up to 96 cores, 384MB L3 cache, and support for DDR5-6400 memory and PCIe 5.0 lanes;
➁ The Threadripper Pro 9000 series targets professional workstations with 12-96 cores and 128 PCIe 5 lanes, while the non-Pro models (24-64 cores, 48 PCIe 5 lanes) cater to HEDT users, offering backward compatibility with existing sTR5 motherboards;
➂ Zen 5 delivers ~16% IPC gains and up to 5.4GHz boost clocks, enabling ~20% performance improvement over previous gen, with availability starting July 2025 for retail and OEM systems.
- AMD Computex 2025 Live Coverage
➀ AMD's Computex 2025 keynote focuses on client and workstation products, led by Senior VP Jack Huynh;
➁ Anticipated announcements include Zen 5-based Threadripper CPUs, mainstream RDNA 4 Radeon GPUs (response to NVIDIA's RTX 5060), and promotion of Ryzen AI MAX laptops;
➂ Emphasis on gaming, workstation hardware, and competition in the mainstream GPU market.
- Innodisk 5QS-P is a 128TB PCIe Gen5 NVMe SSD in E3.L
➀ Innodisk showcased the 5QS-P, a 128TB PCIe Gen5 NVMe SSD in the E3.L form factor at Computex 2025, targeting high-density storage applications;
➁ The SSD offers sequential read/write speeds of up to 12GB/s and 5GB/s, leverages QLC NAND technology, and supports up to 40W power for improved performance;
➂ The E3.L form factor enables higher storage density and chassis flexibility compared to traditional 2.5" drives, marking a shift in data center storage design.
May 20
- Intel Arc Pro B50 and B60 For Lower Cost Pro GPUs and 18A Pather Lake Shown at Computex 2025
➀ Intel unveiled its Arc Pro B50 and B60 GPUs at Computex 2025, targeting the workstation and AI market with 16GB (B50) and 24GB (B60) memory capacities, emphasizing cost efficiency and AI inference capabilities;
➁ The B50 features a 70W bus-powered design, while the B60 supports up to 200W, with a focus on Linux-based virtualization (SR-IOV) and scalable configurations like dual-GPU systems for AI workloads;
➂ Intel also showcased its Panther Lake 18A processors, aiming to challenge TSMC's process leadership, with GPUs expected to launch in Q3 2025 to address the underserved low-end AI inference and server market.
- Sanmina Buys ZT Systems Manufacturing from AMD
➀ Sanmina acquired ZT Systems' manufacturing division from AMD for $2.55B upfront and up to $3B total, aiming to expand U.S. production capacity;
➁ The sale follows AMD's 2024 acquisition of ZT Systems for data center infrastructure capabilities, with AMD divesting manufacturing to avoid conflicts with customers;
➂ Transition allows AMD to focus on rack-scale system design while Sanmina gains hyperscale manufacturing expertise tied to clients like Microsoft Azure.
May 19
- Qualcomm Teases New Data Center CPUs and AI Inference Accelerators in the Works
➀ Qualcomm announced at Computex 2025 its plans to enter the server market with new data center CPUs and AI inference accelerators, focusing on energy efficiency as a key differentiator;
➁ The upcoming hardware includes a server-grade CPU compatible with NVIDIA's NVLink Fusion, indicating collaboration in heterogeneous computing architectures;
➂ This move expands Qualcomm's reach beyond client devices, leveraging its Nuvia-acquired expertise to challenge Intel and AMD in the lucrative data center sector.
- NVIDIA Announces NVLink Fusion: Bringing NVLink to Third-Party CPUs and Accelerators
➀ NVIDIA announced NVLink Fusion at Computex 2025, enabling third-party CPUs and accelerators to integrate with its NVLink ecosystem through semi-custom designs;
➁ The technology includes two approaches: NVLink C2C for connecting custom CPUs to NVIDIA GPUs, and a proprietary NVLink 5 chiplet for third-party accelerators to join NVLink networks;
➂ While offering flexibility, all nodes must include NVIDIA hardware, reflecting a strategic balance between openness and maintaining control over its ecosystem.
- NVIDIA Computex 2025 Keynote Live Coverage
➀ NVIDIA CEO Jensen Huang headlines the Computex 2025 keynote, focusing on AI advancements and data center innovations;
➁ The event emphasizes software updates and partnerships with Taiwanese manufacturers, following major hardware announcements at GTC 2025;
➂ The 90-minute keynote highlights NVIDIA's AI ecosystem and live-streaming accessibility via YouTube.
May 18
- Phison Pascari X200E 6.4TB PCIe Gen5 NVMe SSD Review
➀ The Phison Pascari X200E 6.4TB is a high-performance PCIe Gen5 NVMe SSD designed for enterprise storage, offering U.2 form factor compatibility and enhanced thermal management with integrated heatsink fins;
➀ It features capacities up to 25.6TB and endurance ratings up to 3 DWPD, with the 6.4TB variant prioritizing performance for demanding server workloads;
➂ The drive's 25W power consumption highlights the growing importance of power efficiency in modern server and storage system designs.
May 17
- Kioxia CM9 PCIe Gen5 NVMe SSDs Launched in 2.5in and E3.S
➀ Kioxia launched its new CM9 PCIe Gen5 NVMe SSDs in 2.5-inch and E3.S form factors, leveraging 8th-gen BiCS flash and CBA technology for improved performance and density;
➁ The drives offer capacities up to 61.44TB (2.5-inch) and 30.72TB (E3.S), with significant speed boosts of up to 95% sequential write and 65% random write performance over previous models;
➂ Designed for high-performance data center applications, these SSDs comply with NVMe 2.0 and OCP standards, emphasizing power efficiency and compatibility with modern server ecosystems.
May 16
- Inside the Dell Factory that Builds AI Factories
➀ The article provides an exclusive look inside Dell's 700,000 sq ft factory in Franklin, Massachusetts, which assembles large-scale AI infrastructure systems like the NVIDIA GB200 NVL72 racks;
➁ It details the factory's adaptation to taller 52U racks, component reuse/recycling practices, and custom configurations for high-value AI clusters;
➂ The facility handles full assembly, testing, and packaging of multi-million-dollar AI racks optimized for diverse customer data center requirements.
May 15
- How to See if ECC is Working in Windows Quickly
➀ The article provides a method to verify ECC memory functionality in Windows using CPU-Z by checking the "correction" field in its system report;
➁ Older methods like PowerShell commands may fail, but CPU-Z's detailed DMI report reliably shows ECC status via physical memory array details and bit-width data;
➂ This tip is especially useful for AMD Ryzen/Intel Core users with ECC DDR5 SODIMMs to confirm error correction activation without BIOS checks.
May 14
- AMD EPYC 4005 Grado is Great and Intel is Exposed
➀ AMD's EPYC 4005 'Grado' Zen 5 processors dominate the entry server market with up to 16 cores at 65W TDP, doubling Intel Xeon's core count;
➁ Features include DDR5-5600 support, AVX-512 capabilities, and a 4585PX SKU with 3D V-Cache for cache-intensive workloads;
➂ AMD's pricing undercuts Intel, offering higher core counts at lower costs, pressuring Intel's Xeon E-2400/Xeon 6300P lineup.
May 13
- ASRock Industrial 4X4 BOX-AI350 AMD Ryzen AI 350 Review
➀ The ASRock Industrial 4X4 BOX-AI350 is a compact system powered by AMD's Ryzen AI 7 350, designed for workstation or light server use;
➁ It features a 4-inch square footprint with front/rear ports including USB4, HDMI 2.1, and dual Ethernet (1GbE/2.5GbE), but lacks port labeling;
➂ The review highlights its consistent chassis design, internal expandability, and potential for low-power AI/edge computing applications.
May 12
- This is the NVIDIA MGX 44RU Manifold for Liquid Cooling
➀ NVIDIA introduced the MGX 44RU Manifold for liquid cooling in its GB200/GB300 NVL72 AI racks, enhancing OCP ORv3 racks with standardized power and cooling;
➁ Eight vendors (Pinda, LiteOn, Delta, Foxconn, etc.) showcased customizable manifolds at GTC 2025, highlighting industry collaboration;
➂ The design supports blind-mating connections for compute trays, NVLink switches, liquid tubing, and busbars, streamlining high-density AI deployments.
May 11
- Samsung T7 2TB USB Type-C SSD Review
① The Samsung T7 2TB external SSD offers a compact, lower-cost alternative to the rugged T7 Shield model with a 10Gbps USB Type-C interface;
② Measuring 57x85x8mm and available in multiple colors, it includes both USB Type-A and Type-C cables, though labeling visibility is limited;
③ While not the fastest in its class, it targets users prioritizing affordability and Samsung-branded reliability for portable storage needs.
May 10
- Supermicro H13SAE-MF AMD EPYC 4000 Motherboard Review
➀ The Supermicro H13SAE-MF is an mATX motherboard designed for AMD EPYC 4000 series processors, supporting up to 192GB DDR5 memory and PCIe Gen5 slots;
➁ It features front-to-back airflow optimization, dual M.2 Gen5 slots, and robust I/O including 10Gbps/20Gbps USB ports and triple gigabit NICs;
➂ Positioned for both server and workstation use, it balances high-performance connectivity with enterprise-grade reliability.
May 9
- Gigabyte R113-C10 Review a 1U AMD EPYC 4000 Series Server
➀ The review examines the Gigabyte R113-C10, a cost-effective 1U server chassis designed for AMD EPYC 4000 series CPUs, targeting dedicated hosting and entry-level server applications;
➁ It features two hot-swap 2.5" SATA drive bays, a 400W PSU, basic rear I/O with dual 1GbE ports, and a PCIe Gen5 expansion slot, but lacks NVMe support or advanced networking;
➂ Compared to Lenovo's ST45 V3, it offers shorter depth (393mm) and lower pricing, making it a budget-friendly option despite limited scalability.
May 8
- AmpereOne M Finally Brings the Big 12-Channel Socket to its Arm CPUs
➀ Ampere launches AmpereOne M Arm server CPUs with 12-channel DDR5 memory, enhancing memory bandwidth for enterprise servers;
➁ The new CPUs feature up to 192 cores and DDR5-5600 support, but lag behind AMD EPYC Turin and Intel Xeon 6900P in memory speed;
➂ Released amid Softbank's acquisition of Ampere, the platform lacks immediate customer/partner announcements, raising questions about market adoption.
May 6
- NVIDIA RTX Pro 6000 Blackwell Family for Workstations and Servers
➀ NVIDIA announced the RTX Pro 6000 Blackwell family at GTC 2025, featuring three variants (Server Edition, Workstation Edition, and Max-Q Workstation Edition) tailored for AI and graphics workloads;
➁ The GPUs offer 96GB GDDR7 ECC memory, with power configurations ranging from 400W to 600W, and are positioned as cost-effective alternatives to H100/H200/B200 data center GPUs;
➂ The Server Edition supports 8x GPU configurations with up to 768GB memory, targeting high-density AI and rendering tasks in 4U/5U systems.
May 5
- Kioxia CD8P-R 30.72TB PCIe Gen5 NVMe SSD Review
➀ The Kioxia CD8P-R 30.72TB is a high-capacity PCIe Gen5 NVMe SSD designed for applications requiring both speed and storage density, surpassing traditional HDD capacities;
➁ It features a standard 2.5" form factor, delivers up to 1 DWPD endurance, and balances performance with capacity, though slightly slower than smaller 15.36TB models;
➂ The drive targets data center use cases where large-scale storage and PCIe Gen5 bandwidth are critical, aligning with observed low DWPD utilization trends in enterprise environments.
May 4
- ASUS RS720-E12-RS8G 2U Intel Xeon 6 Server Review
➀ The ASUS RS720-E12-RS8G is a 2U server supporting dual Intel Xeon 6 CPUs and up to three GPUs, including NVIDIA H100 NVL models;
➁ It features eight front NVMe drive bays, ten expansion slots (eight PCIe and two OCP NIC 3.0), and flexible storage/accelerator configurations;
➂ The design emphasizes thermal efficiency with tool-less components, modular risers, and ASUS-standard cooling solutions.