Wiwynn Showcases Next-Generation AI Infrastructure at Computex 2025
Wiwynn , an innovative cloud IT infrastructure provider for
data centers, showcases its next-generation AI servers, developed in
collaboration with Wistron, and advanced direct liquid cooling (DLC) solutions
at Computex 2025 (Booth G0017, TaiNEX1).
"At
Computex, we're showcasing our commitment to innovation and partnership through
our latest accelerated solutions and cooling innovations," said William
Lin, President and CEO at Wiwynn. "We are thrilled to collaborate with
industry leaders to push the boundaries of AI computing and thermal efficiency.
These partnerships have resulted in a comprehensive suite of cutting-edge
products that address the surging demands across modern and legacy data centers
in the AI era."
Next-Generation
AI Systems
- NVIDIA GB300 NVL72: Wiwynn and Wistron are among the first
to be offering NVIDIA GB300 NVL72 systems. This
fully liquid-cooled, rack-level AI system is built on 72 NVIDIA Blackwell Ultra GPUs and
features NVIDIA ConnectX-8 800Gb/s SuperNICs. It delivers a 50X increase in
AI factory output for reasoning model inference compared to NVIDIA Hopper platforms, leading the
next wave of accelerated computing.
- NVIDIA HGX B300: The cutting-edge 10U system built
on NVIDIA Blackwell Ultra, NVIDIA HGX B300 systems,
delivers up to 7X more AI compute than Hopper platforms. With 2.3TB of
HBM3e memory and integration with NVIDIA ConnectX-8 800Gb/s
SuperNICs, it unlocks breakthrough performance for the most complex AI
workloads.
- AMD Instinct™ MI350 Series GPUs and platforms: The brand-new AI server,
featuring the latest AMD Instinct™ MI350 series GPU built on AMD CDNA 4
architecture, presents an impressive 35x improvement in inference
performance compared to Instinct MI300X. The system adopts AMD Pollara 400
Card which delivers a fully programmable 400 Gbps RDMA Ethernet Network
Interface Card.
Advanced
cooling solutions
- Cold Plate Advancements: Showcasing an optimized double-sided cold
plate, designed for future high-power chips. In addition, a 3.5Kw
Electrochemical Additive Manufacturing (ECAM) based cold plates, developed
in partnership with Fabric8Labs is also featured. Both
designs are engineered to lift thermal capability and enhance direct
liquid cooling efficiency, addressing the growing demand for high-power
computing.
- 200 kW AALC Sidecar: Developed with Shinwa Controls Co.,Ltd., the
solution leverages the advanced thermal technologies from semiconductor
industry to further boost AALC (Air-Assisted Liquid Cooling) capability
for legacy data centers.
- In-rack CDU: In collaboration with nVent, the in-rack CDU integrated with the
NVIDIA GB300 NVL72 rack is showcased, delivering high-performance AI
computing and efficient cooling capabilities for data centers.
Cutting-edge
Networking Solutions
- NVIDIA Spectrum-X: The companies are showcasing NVIDIA
Spectrum-X Ethernet for AI networking, building upon
NVIDIA Spectrum-4 MAC technology. This platform integrates SONiC and NVIDIA Cumulus to deliver advanced
Ethernet connectivity for multi-tenant, hyperscale AI clouds, enabling
flexible datacenter deployment scenarios.
- Broadcom Tomahawk 5: The switch features Near-Packaged Optical
(NPO) technology with Broadcom Tomahawk 5 Switch MAC, combined with copper
and optical solutions. It demonstrates significant integration
capabilities on optical engines with either embedded laser & ELSFP.
- 102T Broadcom Switch: The revolutionary datacenter switch adopted
the latest Broadcom Switch MAC, supporting 1.6Tbps x 64 ports in a 4RU
air-cooled system designated for modern datacenter.
Leave A Comment