Intel quickly adapted to the AI boom with the launch of its Core Ultra processors late last year. Following the launch of its AI-powered CPUs, Intel also launched its AI Acceleration programs giving incentives to hardware vendors, and developers allowing the maximization of the NPUs inside the Core Ultra processors. That said, we have the Intel Core Ultra 5 125H representing the new Intel Core Ultra family up for review to see what’s changed inside and how its AI features help in daily computing.
Intel Core Ultra Architecture
Meteor Lake redefines Intel’s existing x86 architecture with an improved 3D performance architecture featuring P-cores, E-cores, and new low-power E-cores as well as new IO tiles for industry-leading connectivity and a built-in NPU AI engine for AI-accelerated workloads. The new Meteor Lake powered CPUs integrated graphics now comes with Intel ARC graphics. The new Ultra Core processors are now built under Intel 4 Logic process technology resulting in 2x area scaling and 20% better power efficiency over the previous Intel 7 process.
One of the main highlights of the new Core Ultra CPUs is the addition of a built-in NPU engine to work alongside the P-cores, E-cores, and ARC GPU. The integrated NPU not only allows for offline AI acceleration – meaning you don’t need to have a subscription or internet connection to access AI tools. But also lightens up the entire CPU’s workload by having a specialized module(aka the NPU) to handle AI-related tasks. To put it simply, think of the NPU as the CPU counterpart of the dedicated Ray Tracing Accelerators found on modern GPUs.
Core Ultra Processor Naming Scheme Explained
Intel introduced the Core Ultra brand last year to replace its signature Core i3/5/7/9 branding marking the brand’s largest client SoC architectural shift in 40 years with the introduction of AI. We have discussed the Intel Naming Scheme in a separate article but here’s a quick explainer for the purpose of this review.
- Core iX is now Core Ultra X with the X with the X digit being the brand level from lowest to highest: 5,7, and 9
- The first number of the processor number indicates the series of the Core Ultra processor – in this case, 1 being the first for the new Core Ultra brand
- The second and third numbers indicate the SKU number. The higher number indicates a much more powerful processor on the same brand level and series.
- The suffix indicate the brand segment with the common segments being
- K – Unlocked
- F – no iGPU
- H – Highest Performance, Mobile
- HK – Highest Performance, unlocked, mobile
- U – Power efficient
- Y – extremely power efficient
The new Core Ultra processors currently have 11 SKUs with the Intel Core Ultra 5 125H being the base model for the H-series. The Intel Core Ultra 5 125H has 4 P-cores, 8 E-cores, and 2 LP E-cores, for a total of 14-cores – 2 more cores than its Core i5 13400H predecessor. Despite having a higher core count, Xe cores, and an NPU engine, its base power is only at 28W from last gen’s 45W with a much higher turbo power at 64~115W from 95W. The Intel Core Ultra 5 125H has the same NPU engines as the rest of the product stack making it a great CPU to benchmark the IPC core improvement and AI capabilities of the new Core Ultra family.
AI Use Cases
Before we dive into the test, it’s important to understand how AI can affect daily computing and even your workflow and how a dedicated processor helps in terms of accessibility, cost, and privacy. There are a ton of AI use cases available right now but we want to focus on what we personally use on a daily basis – Generative AI and AI-driven editing.
Generative AI is the most common segment of AI and is probably what you’re familiar with – the prime examples of it are ChatGPT, Gemini, and Co-Pilot. The first two are free but with limited features or datasets available resulting in limited or outdated responses. You’d need a subscription if you want to access the updated library. Co-pilot on the other hand comes with the latest Intel Core Ultra-based laptops thanks to Intel’s partnership with Microsoft meaning you could access its latest version for free and better yet have offline access thanks to the integrated NPU engine.
AI-driven photo and video editing is something that I personally use thanks to my interest in KPOP. While these features are more commonly accessed in mobile phones with AI, Photo, and video editing apps have made their way to harnessing the NPUs and GPU of desktop computing. You have Adobe Photoshop’s generative fill feature and Stable Diffusion for photos. While the former requires both Internet access and a subscription, Stable Diffusion works offline and works on the NPU engine and GPU thanks to Intel’s OpenVINO interference engine.
The same story can be said with video editing as there as much apps that have these AI such as Capcut, Premiere, and Resolve features but are tied down to cloud-based subscriptions. That said, we have experimented with Wondershare’s Filmore 13 software as despite it requiring a subscription it still has offline functionality and can use the NPU engine thanks to the dedicated NPU driver.
For this quick test, we have a 4k 16:9 video of KPOP girl group NMIXX Haewon focus cam which I then converted to a 4k 9:16 Haewon Fancam using Filmora’s AI motion tracking While we can’t measure the speed of the NPU, we can still see how well it performs AI tracking and the convenience of having only to press three buttons for it to work. This saves us the time of scrubbing the footage to set a keyframe and reframing it every few seconds which translates to a faster workflow, especially for professional video editors.
Compare the two videos here:
- Original 16:9 4K Video – https://youtu.be/sVLHQT5WGrI
- AI Motion Tacked 9:16 Video – https://www.youtube.com/watch?v=uVKX_deZV5A
Benchmark Setup and Test Methodology
Gadget Pilipinas’ testing philosophy is to provide detail-oriented results as accurately as possible so that our readers can replicate our tests given that these conditions are met. Different benchmarking apps and sequences are used depending on the tested component or device.
Our 2024 Laptop benchmark methodology usually tests different profiles to see how the brand’s cooling affects the CPU’s performance. However, since we are focusing solely on the CPU, we tested each one with the highest profile setting for maximum performance and TDP. We’ve opted out of benchmarks and tests requiring GPU computing. We’ve added the Intel Core Ultra 9 as a reference point on how the NPU engine performs and how the core counts affect it considering the spec difference with the Core Ultra 5 125H – note that this is not a direct 1:1 comparison as these tests are selected to focus more on the NPU and its scalability with more CPU cores.
For more information read our Benchmark Methodology article.
Intel Core Ultra 5 125H Synthetic Benchmarks
Cinebench 2024
Geekbench 6
PCMark10
OCCT Enterprise
Intel Core Ultra 5 125H AI Benchmarks
AI Computer Vision Windows ML
AI Computer Vision Windows – Intel OpenVINO
Temperatures and Power Draw
Throughout our testing, the Intel Core Ultra 5 125H only managed to pull a 64.127W power draw yet it was able to go neck and neck with its bigger brother showing the efficiency of the built-in NPU engine. That said, we did reach a peak 103°C temperature due to the intense workload of some synthetic benchmarks but it was still well under the TJMax temp of the CPU.
Performance Summary and Conclusion
The Intel Core Ultra 5 125H and the new Intel Core Ultra family processors usher in a new era of computing with their extremely efficient performance hybrid core architecture design which is further enhanced by the built-in NPU engine. Albeit less powerful than what we’re introduced with when AI launched thanks to server/cloud computing, Intel is able to bring AI functionalities to your device at virtually no cost compared to AI subscription-based services.
One more important factor that’s unfortunately glossed over is privacy. By having a built-in NPU engine on your device and thus having offline functionality, you won’t have to submit your data and digital footprint online to access applications. Using these features offline means you get more control over your data for better privacy but that also comes at a cost of speed and accuracy.
Seasoned AI-dependent users may find an NPU or GPU’s AI image generation to be far more less accurate compared to cloud-based AI. There’s also the speed factor with NPU being the slowest among the three which varies depending on the AI workload as it can scale from a few milliseconds to minutes or hours as shown in our benchmarks. Given the same NPU engine core count across all Intel Core Ultra models, power users may have to wait for the next generation of CPUs or pay more for either a GPU or a subscription for faster AI performance.
But for the average user or first-time AI user, these nitpicks won’t matter as much because getting to experience the AI toolkit is what’s important. Intel’s decision to have the same NPU engine count with its entry-level CPU to its flagship Core Ultra 9 185H is a great move as there will be more users who will get to experience these AI features which will then set the stage for the next generation of processors and introduce the importance of AI-integration across all types of workloads from basic AI acceleration down to complex generative AI.
Core Ultra CPUs and NPUs are relatively new with less than a year at the hands of the consumers. Intel’s push towards NPU integration and AI acceleration programs have literally accelerated the usability of these new features making them something that’s actually useful instead of a price grab gimmick. So, if you’re eager to test out these new AI features without having to pay absurd prices or having limited functionality, getting an Intel Core Ultra-based device is your most reasonable choice – given that you don’t mind the limitations mentioned earlier in favor of an offline, private, and all-access AI feature-set.
Where to buy Intel Core Ultra devices
Acer, Asus, HP, Lenovo, and MSI are the brands that currently offer Intel Core Ultra laptops. If you’re looking to buy Intel Core Ultra devices, be sure to select those with Ultra badge and Intel Evo versions with Intel Core Ultra at authorized dealers.
Grant is a Financial Management graduate from UST. His passion for gadgets and tech crossed him over in the industry where he could apply his knowledge as an enthusiast and in-depth analytic skills as a Finance Major. His passion allows him to earn at the same time help Gadget Pilipinas' readers in making smart, value-based decisions and purchases with his reviews and guides.
