ASUS brings a compact AI supercomputer that can deliver up to 1000 AI TOPs of FP4 precision through NVIDIA's high-performance GB10.
ASUS Reveals Ascent GX10 Mini Supercomputer, a Powerful System with 128 GB of Unified Memory, Ready to Train AI Models
While everyone's eyes are on NVIDIA's GB300 platform, its GB10 is fairly powerful and far more affordable than the former. This led ASUS to build a mini supercomputer called "Ascent GX10" that uses the GB10 and can still train and perform quite intensive operations. While major hardware manufacturers showcased their leading hardware at the GTC event, ASUS also came up with the Ascent GX10 AI Supercomputer, featuring the GB10 Grace Blackwell Superchip and a 20-core Arm CPU.
With such a powerful CPU-GPU combo, the Ascent GX10 supercomputer is ready to deliver up to 1,000 AI TOPs (Tera operations per Second) of processor power that can handle large AI workloads. In contrast, the GB300 platform can deliver 20 times higher AI TOPs (20,000) but is aimed at very intensive AI workloads. The Ascent GX10 focuses more on the budget segment, offering developers an affordable solution to train their AI models.
The Ascent GX10 is shipped with 128 GB of coherent unified system memory and can help in the development, experimentation, and inference of AI models with up to 200 billion parameters. Since the system is based on the NVIDIA GB10 platform, it features the fifth-gen Tensor cores and FP4 support as well. The FP4 support helps the hardware to offer faster computation and NVIDIA's Blackwell GPUs excel at it, especially when it comes to lower memory utilization.
ASUS Ascent GX10 is aimed at offering an economical solution for aI development and, as per KuoWei Chao, the General Manager of ASUS IoT and NUC Business Group,
AI is transforming every industry, and the ASUS Ascent GX10 is designed to bring this transformative power to every developer’s fingertips,
By integrating the NVIDIA Grace Blackwell Superchip, we are providing a powerful yet compact tool that enables developers, data scientists, and AI researchers to innovate and push the boundaries of AI right from their desks.
Ascent GX10 also supports high-speed connectivity through the integration of NVIDIA NVLink-C2C, which offers a cohesive CPU-GPU memory model with the ability to offer 5x the bandwidth of PCI-E 5.0. Moreover, developers can interconnect two GX10 supercomputers via the NVIDIA ConnectX-7 network interface cards and the system will be ready to handle models with up to twice the parameters (roughly 405b) like the Meta Llama 3.1.
There is no word on the pricing at the moment, but ASUS will start taking pre-orders in the second quarter of 2025.
About the author: Sarfraz Khan is a hardware reporter with a focus on PC components and the builder community. With years of experience writing about PC hardware and laptops, his work has been featured on several reputable technology publications. Sarfraz's hands-on experience is demonstrated through his first-person accounts of using and comparing different hardware configurations, providing practical and relatable insights for everyday users. His technical analysis is respected by peers in the enthusiast community and has been cited by specialized hardware sites such as Germany's Igor's Lab.
Follow Wccftech on Google to get more of our news coverage in your feeds.
