HP OmniBook Ultra 16 初步評測：RTX Spark 開局表現不俗

Matt Elliott/CNET

Pros & Cons

• Strong multi-core application and local AI performance
• Thin and elegant design
• Solid build quality
• Beautiful 16-inch 3K OLED touch display

• Middling single-core performance
• Underwhelming battery life
• Gets hot and loud under load
• Not cheap

The HP OmniBook Ultra 16 helps introduce Nvidia’s RTX Spark platform of high-powered (and high-priced) laptops for creative pros and AI developers. Gamers, too, will find something to like with the OmniBook Ultra 16 and other RTX Spark laptops from Asus, Dell, Lenovo, Microsoft and MSI, but these are not gaming laptops first and foremost. Using Nvidia’s new RTX Spark N1X system-on-chip that integrates an Arm-based processor and Blackwell RTX graphics and can access up to 128GB of unified memory, these laptops are built primarily to power agentic AI and content creation workloads — and to bring would-be MacBook Pro buyers to the Windows side of the aisle.

The OmniBook Ultra 16 packs this power inside a thin-and-elegant design that also introduces a new thermal architecture and display hinge design. But even the most valuable company in the world can change the rules of thermal dynamics: the OmniBook Ultra gets hot and loud when it really gets going, just as Apple’s similarly thin-and-powerful MacBook Pro does with the second-most valuable company in the world’s M-series silicon.

You’ll need to be financially solvent yourself to afford an RTX Spark laptop. The OmniBook Ultra 16 starts at $3,200 for a configuration with 32GB of RAM and a 1TB SSD. With double the RAM and storage, the price jumps to $5,000. And if you want to max out the RAM to 128GB, the price rises to $6,200. 

More from CNET

HP sent me the baseline configuration, which is geared more toward creative endeavors than running local AI agents that can make use of the added RAM. I’m still testing the laptop, but here are some early thoughts on the OmniBook Ultra’s design and preliminary benchmark results.

HP OmniBook Ultra 16 specifications

There are two versions of Nvidia’s RTX Spark N1X CPU. The lower-end processor has an 18-core Grace CPU and Blackwell GPU with 5,120 CUDA cores that can support 32GB or 64GB of RAM. The high-end variant has a 20-core Grace CPU and a 6,144-core GPU with support up to 128GB of RAM. HP offers only the higher of the two RTX Spark CPUs for the OmniBook Ultra 16. (HP’s other RTX Spark laptop at launch, the OmniBook X 14, starts off with the lower-end N1X chip for $3,000 with 32GB of RAM and a 1TB SSD.)

Like Apple’s M-series and Qualcomm’s Snapdragon X-series processors, Nvidia’s RTX Spark N1X is an Arm-based chip as opposed to those from Intel and AMD that use x86 architecture. As with laptops with Snapdragon X processors, you may encounter some Windows-on-Arm compatibility issues with RTX Spark laptops. Many Windows applications have native ARM versions, but others won’t run or will need to run through Microsoft’s Prism emulator, which can negatively impact performance.

The amount of memory you choose for an RTX Spark laptop is important for two reasons. The first is the unified memory architecture that lets the CPU and GPU directly share a coherent pool of RAM. This means the total allotment of RAM is used more effectively than an Intel chip, where the memory is split, with portions dedicated to the CPU and GPU, resulting in low-bandwidth bottlenecks between the processor and graphics processors and the memory. RTX Spark has a fast memory bandwidth of up to 300GB per second. The second reason for making a careful choice of memory with an RTX Spark laptop is that the RAM is soldered to the motherboard and is not expandable or upgradable after purchase.

You can’t replace or add to the laptop’s memory, but you do have some say in how it’s best allocated. In the Nvidia app, you can use a slider to choose how much RAM you want to dedicate to the GPU. On my OmniBook Ultra 16 test system, it was set at 12GB (of the 32GB total), and you can go up to 16GB. (With 64GB of RAM, you can dedicate up to 48GB to the GPU, and with 128GB, you can dedicate up to 96GB.)

Tower hinge and trunk design

To keep the new RTX Spark N1X performing at its peak, HP introduced a new cooling system. On most laptops, the display opens and sits behind the back edge. With the OmniBook Ultra 16, the display rests on top of the back edge. It looks a bit like the thermal shelf found on some gaming laptops, but in miniature. This design keeps the back edge clear to maximize airflow out of the system. Internally, two cooling fans and three heat pipes help dissipate heat.

Aside from the unique display hinges that sit on top of instead of behind the laptop, the OmniBook Ultra 16 looks very much like the OmniBook Ultra 14, which happens to be my favorite laptop of the year. The all-aluminum chassis has a rigid feel and cuts a thin profile. It’s every bit as premium in look and feel as the MacBook Pro.

Other highlights of the design include its huge 3K OLED touch display providing stunning visuals, and the huge touchpad, which offers a lively haptic click response. I also liked typing on the keyboard — the keys have a plush yet solid feel — but data scientists and the like might miss the number pad that’s often included on a 16-inch laptop but not here.

RTX Spark performance

OK, now for the good stuff: How does the RTX Spark-powered HP OmniBook Ultra 16 stack up to the competition? We are in the process of testing two other RTX Spark laptops in addition to the HP, the Microsoft Surface Laptop Ultra and Lenovo Yoga 9N 2-in-1 16 Gen 11. Both are 16-inch models with the higher-end RTX Spark N1X with 20 CPU cores. This trio finished in a pack on most of the benchmarks we’ve run them through so far. (Scroll down to see the benchmark charts below.)

On the Geekbench 6 and Cinebench 2024 tests, the RTX Spark systems finished behind the latest 16-inch MacBook Pro models with Apple’s M5 Pro and M5 Max processors on both the multi- and single-core versions of the tests.  

On Geekbench 6, the RTX Spark laptops had similar multicore performance as a pair of laptops with Qualcomm Snapdragon X2 Elite chips — the Asus Zenbook A16 and OmniBook Ultra 14 — but their single-core performance was lower than that of the two Snapdragon X2 laptops. On Cinebench 2024, the RTX Spark laptops finished just ahead of these two Snapdragon X2 laptops, but again fell off with single-core performance. 

The RTX Spark laptops were able to flex their muscles a bit more on two 3DMark tests with native Arm versions, Steel Nomad and Solar Bay, and finished with scores below the M5 Max-based MacBook Pro but ahead of the M5 Pro system and the other Windows laptops assembled here. And the RTX Spark laptops topped both MacBook Pros on Procyon’s Stable Diffusion XL diffusion benchmark that measures AI inference performance of the GPU, and did so by fairly significant margins.

I ran the OmniBook Ultra 16 through a couple of gaming benchmarks and found that the RTX Spark N1X’s GPU was able to achieve playable frame rates at 1080p. The OmniBook Ultra 16 averaged 96 frames per second on Guardians of the Galaxy (High at 1080p) and 99 fps on Shadow of the Tomb Raider (Highest at 1080p). These are lower frame rates than what you can get with a cheaper RTX 5050 or 5060 gaming laptop, but higher than those of Intel’s integrated Arc B390 GPU.

HP trumpets all-day battery life for the OmniBook Ultra 16, but I didn’t see anything approaching that here. The laptop lasted a little more than 11 hours on our battery test, which was about an hour more than the Microsoft Surface Laptop Ultra and two hours less than the Lenovo Yoga 9n 16 lasted. Meanwhile, the two 16-inch MacBook Pros each lasted more than 20 hours on the same test. 

The HP OmniBook Ultra 16’s strengths are with multicore application performance and local AI inferencing, based on the tests we’ve run so far. Its single-core performance was merely average, and its battery life was underwhelming. Check back soon for my full review of the HP OmniBook Ultra 16 with more performance results and a closer look at the design. And keep your eyes out for my review of Lenovo’s first RTX Spark laptop, the Yoga 9N 16 Gen 11.

How we test computers

The review process for laptops, desktops, tablets and other computer-like devices consists of two parts: performance testing under controlled conditions in the CNET Labs and extensive hands-on use by our expert reviewers. This includes evaluating a device’s aesthetics, ergonomics and features. A final review verdict is a combination of both objective and subjective judgments. The list of benchmarking software we use changes over time as the devices we test evolve. The most important core tests we’re currently running on every compatible computer include Primate Labs Geekbench 6, Cinebench R24, PCMark 10 and 3DMark Fire Strike Ultra. A more detailed description of each benchmark and how we use it can be found on our How We Test Computers page. 

System configurations