Geline Salac

Article (news)

The HP Zbook Ultra G1a

Melvin Ryan  Fetalvero

ABS-CBN News

<div><span class="fr-img-caption fr-fic fr-dib" style="width:960px;display:block" width="960"><span class="fr-img-wrap"><img src="https://od2-image-api.abs-cbn.com/prod/editorImage/1772236484444hp zbook g1a ultra.jpg" style="width:960px" class="fr-fic fr-dib" width="960"/><span class="fr-inner" style="text-align:left">The HP Zbook Ultra G1a packs workstation power in a thin and light laptop. Arthur Fuentes, ABS-CBN News</span></span></span></div><p>MANILA -- I am a software engineer by trade, working for ABS-CBN News Digital Media. I am always on the lookout for how we can leverage technology in our various workflows. </p><p>This month, I got my hands on a ZBook Ultra 14 G1a. I had the opportunity to do some experiments to see how a laptop designed for AI can be used in various workflows in the newsroom.</p><p>The Hardware: A Workstation in Ultrabook Clothing</p><div><strong>Key Specifications</strong></div><ul style="margin-top:0;margin-bottom:0;padding-inline-start:48px"><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Processor: AMD Ryzen AI Max+ PRO 395 (16-core, 64MB L3 cache)</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Memory: 64GB LPDDR5x-8533 MT/s (soldered)</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Storage: 1TB or 2TB PCIe Gen4 NVMe TLC M.2 SSD</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Graphics: Integrated AMD Radeon™ 8050S Graphics</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Display: Options include 14&quot; WUXGA (1920x1200) IPS 400-nits or 14&quot; 2.8K (2880x1800) OLED 120Hz</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Weight: Starting at 1.57 kg (3.46 lb)</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Battery: 140W USB Type-C adapter</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Networking: Wi-Fi 7 (MediaTek MT7925) + Bluetooth 5.4</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Ports: 2x Thunderbolt 4 (40Gbps), 1x USB Type-A (10Gbps), 1x HDMI 2.1, 3.5mm combo jack</p></li><li dir="ltr" style="list-style-type:disc;color:#000000;background-color:transparent;text-decoration:none;vertical-align:baseline;white-space:pre"><p>Security: HP Wolf Pro Security, IR Camera, Fingerprint reader, TPM 2.0 </p></li></ul><p>This laptop features the AMD Ryzen™ AI MAX PRO (Strix Halo) architecture. One highlight is the 64 GB of shared memory between the CPU and GPU, which is a significant advantage. </p><p>The combination of the OLED Display, Integrated AMD Radeon™ 8050S Graphics, and 1TB storage should be able to run Adobe Photoshop, Adobe Premiere and After Effects, enabling Video Editors to run their video production workflows without issues. </p><p>Multiple applications should run smoothly while running concurrently because of the large memory. The storage should enable them to work with a lot of raw materials and export several versions of their video. The HDMI and Thunderbolt port enables expansion of Display and storage to maximize the core compute capabilities of this laptop.</p><p>On paper, the same Graphics, Display and Compute features also enable this laptop for AAA Gaming.</p><p>But where it can really shine is in running Local AI Models.This unified pool enables the loading of much larger AI models, such as 32-billion or 70-billion quantized models, which are inaccessible to laptops limited to standard 16GB or 32GB memory configurations.</p><div><strong>Benchmarking &quot;Raw&quot; vs. &quot;Real&quot;</strong></div><p>Synthetic Tests: I ran Geekbench tests</p><div><img src="https://od2-image-api.abs-cbn.com/prod/editorImage/1772097007289hp z1 geekbench.png" style="width:960px" class="fr-fic fr-dib" width="960"/></div><p>To put these new numbers into perspective, here is how this performance compares to other top-tier AI-capable machines:</p><table style="width:100%"><tbody><tr><td style="width:25%;background-color:rgb(204, 204, 204)"><div style="text-align:center"><strong>Device Category</strong></div></td><td style="width:25%;background-color:rgb(204, 204, 204)"><div style="text-align:center"><strong>Processor</strong></div></td><td style="width:25%;background-color:rgb(204, 204, 204)"><div style="text-align:center"><strong>Single Core</strong></div></td><td style="width:25%;background-color:rgb(204, 204, 204)"><div style="text-align:center"><strong>Multi Core</strong></div></td></tr><tr><td style="width:25.0000%">This device</td><td style="width:25.0000%">Ryzen™ AI MAX PRO (Strix Halo)</td><td style="width:25.0000%">2,106</td><td style="width:25.0000%">14,149</td></tr><tr><td style="width:25.0000%">AI Workstation</td><td style="width:25.0000%">Ryzen™ AI MAX+ PRO 395 (16-core)</td><td style="width:25.0000%">~2,825</td><td style="width:25.0000%">~17,562</td></tr><tr><td style="width:25.0000%">High-End Laptop</td><td style="width:25.0000%">Apple M4 Pro (14-core)</td><td style="width:25.0000%">~3,400+</td><td style="width:25.0000%">~22,000+</td></tr><tr><td style="width:25.0000%">AI Laptop</td><td style="width:25.0000%">Intel Core Ultra 9 185H</td><td style="width:25.0000%">~2,200</td><td style="width:25.0000%">~12,000</td></tr><tr><td style="width:25.0000%">Desktop CPU</td><td style="width:25.0000%">Ryzen 9 7900X (12-core Desktop)</td><td style="width:25.0000%">~2,800</td><td style="width:25.0000%">~17,000</td></tr><tr><td style="width:25.0000%">Mini PC</td><td style="width:25.0000%">Ryzen 9 7940HS (Performance Mode)</td><td style="width:25.0000%">~2,600</td><td style="width:25.0000%">~12,500</td></tr></tbody></table><div><br/></div><p><strong>The &quot;Workstation&quot; Threshold:</strong> A multi-core score above 14,000 is a critical benchmark for software engineers. It signifies that the device can handle intensive multithreaded tasks—like compiling large codebases or local LLM inference—at speeds formerly reserved for much larger desktop towers.</p><p><strong>Multi-Core Dominance:</strong> Your multi-core result now beats flagship AI laptops like the Intel Core Ultra 9 185H and entry-level Apple M4 configurations.</p><p>While single-core performance is solid, the multi-core scaling is where the ZBook Ultra G1a truly shines for engineering workflows. It offers a unique balance of thin-and-light portability with the raw parallel processing power of a high-end desktop.</p><div><strong>Local LLM Performance</strong></div><p><strong>HP AI Companion</strong> - Out of the Box this laptop comes with HP AI Companion with default models loaded. You can download LLM models in hugging face and load them in this HP AI Companion App. The user interface is very different from LM Studio, Ollama and Comfy UI.</p><div><img src="https://od2-image-api.abs-cbn.com/prod/editorImage/1772097642829LLM.png" style="width:960px" class="fr-fic fr-dib" width="960"/></div><p>Simple tests, such as queries using the HP AI companion, using the default models setup, we observed around a 10-20 tokens per second average. The default model has limited capabilities. So, doing tests on open source models from Hugging face is needed to do further review on this app. It’s a good software to do local AI tests out of the box.</p><p><strong>LM Studio</strong> - After setting up ROCM and allocating 32GB of RAM to the GPU, I was able to do several tests in LM Studio. On Average after doing some tweaking on the Bios and the Operating system. I was getting on average 30-50 Tokens Per Second on various local Models and Prompts for use cases in the newsroom in this workstation.</p><p>I was able to install <strong>Ollama</strong> and do shallow testing using the same model in LM Studio and was getting almost the same average of Tokens Per Second output on the test prompts.</p><p>The workstation is also compatible with <strong>Comfy UI,</strong> which will allow you to create videos and photos using AI models. Unfortunately, I did not have enough time to do these tests as the models for generative AI take a long time to download.</p><p>With 32 GB allocated to the GPU, they were loaded in full in the GPU. And while the Tokens Per Second is not as fast as the paid cloud versions or NVDIA top tier GPUs. </p><div><strong>Multimodal &amp; Generative Deep Dive</strong></div><p>Using <strong>Qwen/Qwen3-coder-30b</strong> Parameter model, I am getting approximately 30-40 Tokens Per Second on prompts related to SEO, GEO, HTML, CSS, and Web page optimizations. </p><p>I did tests using the <strong>Qwen/Qwen3-v1-30b </strong>Parameter model, which is a model that can process images. I did prompt tests to get recommendations for captions, descriptions and keywords on the photos and was getting 30-40 Tokens Per Second. </p><p>I ran <strong>OpenAI/GPT-OSS-20b</strong> Parameter model, which is a reasoning model. I did prompts to get recommendations for copies, title, summary and outline of articles and got  45-50 Tokens Per Second. These models are around 12-19 GB in size. </p><p>Considering the form factor of this laptop, this is not bad at all. It is possible to increase the memory allocated to the GPU up to 48GB and load bigger models than the one I tested.  Using these open source models, the output of the prompts was quite usable.</p><div><span class="fr-img-caption fr-fic fr-dib" style="width:960px;display:block" width="960"><span class="fr-img-wrap"><img src="https://od2-image-api.abs-cbn.com/prod/editorImage/1772236549346zbook g1a side 2.jpg" style="width:960px" class="fr-fic fr-dib" width="960"/><span class="fr-inner" style="text-align:left">The HP Zbook Ultra G1a packs workstation power in a thin and light laptop. Arthur Fuentes, ABS-CBN News</span></span></span></div><div><strong>The Frontier: Subagents and OpenClaw</strong></div><p><strong>Claude Code</strong> - This is my daily driver and preferred AI product because of the models from anthropic and its support for subagents, skills, and MCP. An issue I always encounter using this is token limitations and throttling because of the plan I am subscribed to. Using Ollama or LM Studio, I can actually offload some of the minor tasks to Local LLM models hosted in this HP Zbook so I can continue my development and AI experiments, or use Claude Code and the local model while I am offline for use cases that do not need the paid and performant models.</p><p><strong>Open Claw -</strong> Everyone is doing tests using Open Claw these days. The memory capacity of this laptop enables you to install Open Claw via docker and run Ollama/Notebook LM locally. This will enable you to have some security measures in Open Claw. And configure your subagents to use Local hosted models for menial tasks and use the paid Online models for the complicated scenarios. This should maximize your workflow. </p><div><strong>Portability and Daily Use</strong></div><p>The HP Zbook Ultra G1a is light considering the computing resources that it packs. It’s as light as a 14 inch Macbook Air from Apple and is lighter than most gaming laptops. During the tests, I did not notice any heat issues with the actual laptop, but I noticed the power brick can get hot specially when I am running AI processes. </p><p>The expansion ports are sufficient to support extended display and storage. The Thunderbolt port also opens the possibility of expanding AI capacity by using a GPU. Using the battery standard compute operations can yield 4.5 - 7 hours of power. I did not do any tests using high heavy AI load using the battery to avoid throttling of the performance of the compute resources. The materials used look premium, and the OLED screen really makes the screen an eye candy.</p><div><strong>Verdict: Who is this for?</strong></div><p><strong>Video editors - </strong>I can see video editors who are also using AI-generated content(Graphics, Videos) and are exporting videos in various formats. The combination of the GPU, Memory, Storage, Screen and expansion ports should enable them to run various adobe products and AI tools concurrently. </p><p><strong>Software developers -</strong> Because of the CPU and Unified Memory, Software Developers have plenty of memory and space to run virtual machines, docker containers and leverage on a combination of cloud and local AI models to assist them in their software development. This is ideal specially for those doing R&amp;D on AI models running pytorch and tensor flow processes.</p><p><strong>Data Scientists -</strong> They can train their models and run their python code and AI subagents using this workstation even if they have limited connectivity. This is ideal specially if you are processing private data like clinical data.</p>

<p>HP packed workstation power in a thin-and-light laptop. And it can surprisingly run LLM</p>

HP packed workstation power in a thin-and-light laptop. And it can surprisingly run LLM

REVIEW: How the workstation class HP ZBook Ultra 14 G1a laptop reclaims local AI workflows

News

Entertainment

Lifestyle

Sports

Metro.Style

More

Sudoku

Word of the Day

Lotto

Horoscope

Technology

REVIEW: How the workstation class HP ZBook Ultra 14 G1a laptop reclaims local AI workflows

Melvin Ryan Fetalvero,

ABS-CBN News

Published March 02, 2026 09:00 AM PHT

|

Updated March 02, 2026 11:32 AM PHT

Read More:

HP Zbook Ultra G1a

workstation laptop

LLM

run AI locally

AI

ABS-CBN is the leading media and entertainment company in the Philippines, offering quality content across TV, radio, digital, and film. Committed to public service and promoting Filipino values, ABS-CBN continues to inspire and connect audiences worldwide.

Corporate

Investors

Governance

International Distribution

Corporate

Governance

Investors

International Distribution

NPC Seal of Registration

Privacy Policy

AI Policy

Terms of Service

Advertise with Us

© 2026 ABS-CBN Corporation. All Rights Reserved.

NPC Seal of Registration

Privacy Policy

Terms of Service

AI Policy

Advertise with Us

© 2026 ABS-CBN Corporation. All Rights Reserved.

NPC Seal of Registration

Privacy Policy

Terms of Service

AI Policy

Advertise with Us

© 2026 ABS-CBN Corporation. All Rights Reserved.