Asus Leads the Charge with the Ascend GX10, but Pricing Remains a Mystery

Nvidia’s highly anticipated first desktop PC chip, the GB10 Grace Blackwell platform, is set to land this month, with Asus taking the lead with its Ascend GX10 Grace Blackwell desktop platform. However, the precise availability date and pricing of the system remain unclear.

According to an Asus invitation spotted by VideoCardz, the company plans to re-introduce its Ascend GX10 mini-PC on July 22. The event, which will feature a keynote speaker from Nvidia, promises to showcase the power of the compact Ascend GX10 system in empowering AI development. The system is touted as more than just a mini-PC, with Nvidia positioning it as a workstation and edge AI solution that delivers data center-level performance in a compact form-factor.

At the heart of the GB10 platform lies the Superchip system-in-package (SiP), which integrates a Grace CPU with 10 high-performance Arm Cortex-X925 cores running up to 3.90 GHz and 10 power-efficient Cortex-A725 cores. This CPU architecture is designed to provide a perfect balance between performance and power efficiency, making it an ideal choice for AI workloads. The SiP also features a Blackwell GPU capable of delivering 1 PetaFLOPS of FP4 compute throughput for AI workloads, which is a significant increase over traditional CPU-GPU setups.

One of the key advantages of the GB10 platform is its unified memory architecture, which allows for seamless communication between the CPU and GPU. This enables the system to perform complex AI tasks, such as large language model training, with ease. Additionally, the GB10 platform boasts a 256-bit memory interface supporting 128GB of unified LPDDR5X memory, reaching bandwidths up to 273 GB/s. This level of memory bandwidth is comparable to that of Apple’s M4 Pro, making it an ideal choice for demanding workloads.

Nvidia emphasizes the GB10 platform’s strong single-thread performance, which is critical for AI workstation use cases. The company positions the GB10 platform as an ideal solution for building and running large language models, generative AI applications, and other demanding workloads on desktop-class systems.

However, leaked performance figures from Geekbench may raise some concerns. According to the leaked scores, the GB10 is close to Qualcomm’s Snapdragon X Elite and near Apple’s M3 processor in general-purpose compute. Despite this, high single-thread performance remains an important factor for AI workstation use cases.

Asus’ Ascend GX10 mini-PC is expected to closely resemble Nvidia’s own DGX Spark small form-factor system, which is priced at $3,000. Other companies, including Dell, HP, and Lenovo, are also prepping their own versions of the DGX Spark, but pricing and positioning remain unclear.