NVIDIA Told me How to Benchmark RTX Spark
Linus Tech Tips
111,153 views • 16 hours ago Save 8 min 5 min read
Video Summary
NVIDIA's new RTX Spark laptop chip, designed to rival Apple Silicon, faces scrutiny over dated components and performance that falls short of claims. Despite combining ARM CPUs with NVIDIA graphics and ample RAM, early benchmarks reveal it struggles in gaming and productivity tasks, often underperforming against competitors like Qualcomm and Apple.
While the chip shows promise in AI compute with features like CUDA support, its real-world effectiveness is questioned, especially given the high price point of devices like the Microsoft Surface Laptop Ultra ($6,000) and Asus ProArt P16 ($7,000). The review concludes that the RTX Spark is difficult to recommend, with its niche appeal limited to users with specific, extreme needs for on-the-go AI capabilities.
Short Highlights
- Dated Components: The N1X chip utilizes older ARM core designs (Cortex-X925 and A725) rather than newer available options.
- Gaming Performance: While capable of playable frame rates in some titles like Cyberpunk 2077, it underperformed in others like Total War Pharaoh and Grid Legends.
- Productivity Lag: In benchmarks like Cinebench and 7-Zip, the RTX Spark lagged behind competitors, including the latest Snapdragon and Apple Silicon.
- AI Compute Mixed Results: While showing strength in prompt processing with CUDA, text generation performance was average, and overall AI capabilities were not groundbreaking.
- Battery Life Concerns: Despite larger batteries, Spark laptops offered comparable or less battery life than some competitors.
- Thermal Issues: The Surface Laptop Ultra hovered around 95 degrees Celsius, though it did not throttle; reliable thermal data was difficult to obtain.
- High Price Point: The reviewed devices, the Microsoft Surface Laptop Ultra and Asus ProArt P16, are priced at $6,000 and $7,000 respectively, making them difficult to justify.
Key Details
NVIDIA's RTX Spark Launch and Benchmarking Restrictions [0:00]
- NVIDIA's RTX Spark laptop chip faced numerous delays before its launch, with the company initially barring independent benchmarks.
- NVIDIA cited concerns about misleading benchmark results, especially for local AI performance, which has evolved significantly since the chip's unveiling.
- The chip represents a significant power play by NVIDIA, entering AMD and Intel's traditional laptop CPU territory.
"In fairness to NVIDIA, they did make some good points. Benchmarks that spit out a number but don't represent real workloads can be very misleading, and local AI performance, both from a speed and accuracy standpoint, has changed so much since they unveiled this product in June that the value proposition is very different today than it was just a few short months ago."
Core Specifications and Design [0:38]
- The RTX Spark chip (N1X) combines an ARM CPU with NVIDIA graphics, supported by up to 128GB of LPDDR5X memory.
- It is designed to compete directly with Apple Silicon.
- The processors appear to be based on older ARM core designs, Cortex-X925 and Cortex-A725, revealed over two years ago, rather than newer C-Series designs.
"What's really interesting here is that while the SoC is fresh off the line, the processors seem to be DOA. Not dead, but dated on arrival, since the design is kind of a mash-up of the Cortex-X925 and the Cortex-A725 that ARM revealed over two years ago."
Gaming Performance Gauntlet [1:30]
- Testing included games like Cyberpunk 2077, Grid Legends, Shadow of the Tomb Raider, and Total War Pharaoh, running via Microsoft Prism on the ARM-based lineup.
- Discrete graphics generally outperform integrated solutions, but the Spark chip's onboard graphics showed promise.
- In Cyberpunk 2077, the Spark achieved playable frame rates at 1200p medium settings, but struggled in Total War Pharaoh, averaging below 60 FPS at ultra settings.
"Discrete graphics still run the best, but if you're going to have onboard graphics, boy, do you ever want one of these big GPU chips like you find in the Spark laptops, because they did manage to eke out a win over both Apple and Qualcomm's integrated offerings."
Productivity and AI Compute Tests [3:20]
- Native Windows ARM builds were used, avoiding performance losses from translation.
- Cinebench results showed the Spark lagging in multi-core and single-thread performance compared to Snapdragon and Apple Silicon.
- In AI benchmarks like Llama Bench, the Spark excelled in prompt processing due to CUDA but fell to the middle of the pack in text generation.
"NVIDIA's AI laptops are performing in line with the hardware design that they were given, it's just that that hardware is a little old."
Battery Life, Thermals, and Final Verdict [7:00]
- Battery endurance tests showed the Surface RTX Spark lasting about 9 hours, similar to a MacBook Pro 14, while the Asus ProArt P16 lasted 12 hours, falling short of other devices with smaller batteries.
- Thermal testing on the Surface showed fan speeds hovering around 95 degrees Celsius without throttling, but reliable data was limited due to lack of full support from hardware monitoring software.
- The high price points (nearly $6,000 for the Surface, $7,000 for the ProArt) and performance limitations make the RTX Spark devices difficult to recommend for most users.
"All in all, these new N1X devices are... tough to recommend. Yes, they're fully featured, and they benefit from extensive optimization work by Microsoft, Qualcomm, and NVIDIA to make Windows on ARM a thing."
