Nvidia's forthcoming N1X system-on-chip (SoC) has recently surfaced in benchmark databases, revealing intriguing yet perplexing performance metrics. Despite boasting a substantial 6,144 CUDA cores—equivalent to the desktop RTX 5070—the N1X's integrated GPU (iGPU) delivered performance levels comparable to the older RTX 2060 in FurMark tests. This disparity has sparked discussions about the chip's current state and its potential capabilities upon release.
Benchmark Insights
In a recent FurMark 1.39 Donut test, the N1X's iGPU achieved a score of 4,286 points, translating to 71 frames per second (fps) at 720p resolution. This performance is notably lower than expected, especially considering the chip's impressive core count. For context, the RTX 5060, with fewer CUDA cores, outperforms the N1X in similar tests. The N1X was identified as "Nvidia JMJWOA" in the benchmark, running on Windows 11 with graphics driver version 590.22.
Further analysis indicates that during the test, the GPU utilization peaked at only 63%, with temperatures reaching 59°C. Such low utilization suggests potential throttling mechanisms or unoptimized drivers, which are common in early development stages. Additionally, the chip's power consumption is estimated at 120W, significantly lower than the RTX 5070's 250W, which could also contribute to the reduced performance.
Technical Specifications and Comparisons
The N1X SoC features a 20-core ARM-based CPU, divided into two 10-core clusters, and an iGPU with 48 Streaming Multiprocessors (SMs), totaling 6,144 CUDA cores. This configuration mirrors that of Nvidia's GB10 Superchip, utilized in DGX Spark AI mini-PCs, suggesting that the N1X may be a consumer-oriented adaptation of this architecture. (tomshardware.com)
Despite the high core count, the N1X's current performance aligns more closely with the RTX 2050, as evidenced by an OpenCL score of 46,361 in Geekbench benchmarks. This outcome is likely due to the early engineering sample's modest clock speed of 1.05 GHz and reliance on shared LPDDR5X memory, which can limit bandwidth compared to dedicated GDDR memory found in discrete GPUs. (tomshardware.com)
Development Status and Future Prospects
It's crucial to recognize that the N1X is still under development, and the observed performance metrics may not reflect the final product's capabilities. Early benchmarks often suffer from unoptimized drivers and firmware, leading to subpar results. The N1X's appearance in Windows 11-based tests indicates ongoing efforts to optimize the software stack across platforms ahead of the anticipated 2026 launch.
Moreover, the N1X's design, which integrates a Blackwell GPU with an ARM-based CPU, aims to balance AI performance, gaming, and efficiency. This hybrid approach positions the N1X as a potential competitor to existing x86 processors from AMD and Intel, as well as ARM-based solutions from Apple. However, achieving performance parity will depend on further optimizations and hardware refinements. (tomshardware.com)
Conclusion
The Nvidia N1X SoC presents a promising yet currently underwhelming performance profile. While its high CUDA core count suggests significant potential, early benchmarks reveal performance levels that do not yet meet expectations. As development progresses and optimizations are implemented, it's anticipated that the N1X will better realize its capabilities, potentially offering a compelling option in the ARM-based processor market.
Source: Club386 Nvidia's N1X SoC hammered by RTX 5060 in FurMark despite having more CUDA cores | Club386