Is NVIDIA's GB10 Superchip Ready to Challenge AMD's Ryzen AI Max+ in Traditional Computing? You might know NVIDIA's GB10 superchip, particularly the Dell Pro Max GB10, for its cutting-edge AI capabilities powered by its Blackwell GPU. However, a burning question for many has been: how does its CPU performance stack up in everyday Linux tasks against formidable competitors like AMD's Ryzen AI Max+ 395 "Strix Halo"? Well, Phoronix has dived deep into this, and the results are quite illuminating.
Unpacking the GB10's CPU Powerhouse: The NVIDIA GB10 isn't just about AI; it boasts a robust CPU configuration featuring 20 Arm cores. This impressive array is composed of ten high-performance Cortex-X925 cores and ten energy-efficient Cortex-A725 cores. Coupled with a generous 128GB of LPDDR5x memory, these AArch64 cores are designed to deliver substantial processing power for a wide range of applications, going beyond its renowned graphical prowess.
To truly gauge the GB10's CPU mettle, a series of Linux CPU benchmarks were meticulously conducted. The Dell Pro Max GB10 was put head-to-head against the Framework Desktop, which is equipped with the AMD Ryzen AI Max+ 395 Strix Halo SoC. This direct comparison aims to provide a clear picture of how these two architectures perform in more conventional computing scenarios.
The Power Consumption Conundrum: A fascinating aspect of this comparison is the approach to power consumption. While both AMD and Intel platforms readily offer CPU power data through Linux's PowerCap/RAPL interfaces, the NVIDIA GB10 presented a different challenge. It did not expose its CPU power metrics through these standard channels. Consequently, the evaluation of power efficiency had to rely on measuring the total AC system power consumption for both testbeds. A WattsUp Pro power monitor was instrumental in capturing this real-time data, allowing for a detailed performance-per-Watt analysis.
A Level Playing Field: To ensure a fair and accurate comparison, both the Dell Pro Max GB10 and the Framework Desktop were running the same operating system: Ubuntu 24.04.3 LTS. This choice is significant, as NVIDIA's own DGX OS is also based on this version. Furthermore, both systems utilized the Linux 6.14 kernel and the GCC 13.3 compiler, ensuring that software variables were kept consistent. A big thank you to Dell for providing the Pro Max GB10 for this in-depth testing, and to Framework Computer for supplying the Strix Halo Framework Desktop review unit!
But here's where it gets controversial... While the GB10's multi-core performance might be impressive, how does its single-core performance compare in everyday tasks that still rely heavily on individual core speed? And this is the part most people miss: could the GB10's focus on AI lead to compromises in its traditional CPU strengths, or is it a true all-rounder? What are your thoughts on NVIDIA venturing further into the CPU space? Do you believe its AI-centric design will ultimately benefit or hinder its general-purpose computing capabilities? Let us know in the comments below!
