Feed aggregator

• Garmin's Connect IQ store has seen an 'enormous influx' of new submissions
• These include apps and watch faces, and are causing a big backlog as developers struggle to get their submissions reviewed
• It's likely AI is to blame, as the sudden wave of submissions 'was never a problem in the past'

If you wear one of the best Garmin watches, you might have used the Garmin Connect IQ store to download an app like Spotify or Komoot, a new free or paid watch face (even a Pokémon-branded one), or a fun community-made complication like the Walk with Frodo app. However, downloading stuff from the Connect IQ store in future may involve sifting through thousands of low-quality, AI-generated submissions.

Watch face and app developers are complaining on Garmin's forums about delays in submission review: while the process typically takes 1-2 days, according to some devs, the process is now taking up to 13 days.

Garmin Connect IQ moderator/representative Brandon posted directly to one thread, stating: "This is not due to a bug — we have gotten an enormous influx of app submissions in the last month, with no signs of slowing down, and are working through approvals as quickly as we can.

"This was never a problem in the past, except perhaps delays because of a long holiday weekend (e.g. Christmas)."

(Image credit: Future)

The fact this glut of submissions is a sudden problem or new phenomenon points to one thing only — AI. Garmin Connect IQ has likely been flooded with vibe-coded or AI-generated products designed to make a quick buck by flooding the digital marketplace.

One Garmin user said in response to Brandon: "You need to limit the upload frequency for each developer; those poorly made AI-generated watch faces will consume everything."

Another said: "I have observed this trend worsening significantly as well. The Connect IQ Store is increasingly flooded with low-effort, repetitive submissions that hurt discoverability, degrade user trust, and crowd out well-crafted apps."

While this is bad for developers (who get charged $100 annually to sell apps on the Connect IQ store) it's also bad for Garmin users. We'll now have to wade through a sea of AI slop watch faces in order to get to faces and apps created the old-fashioned way — with human-led, thought-out complications developed and designed by real Garmin users.

It's just one more example of digital merchants being reduced to dealing with an avalanche of AI slop submissions, just like other app stores, and almost every other breed of online storefronts for digital products. We reported as far back as 2023 that AI-generated e-books were flooding Kindle Unlimited, making it difficult for users to find new books, and the problem has only gotten worse with time. The same will likely happen to the Connect IQ store.

I've reached out to Garmin for comment on the situation, and will update this article if I get any response. In the meantime, if you're a Connect IQ user and you've noticed the change, let us know in the comments below.

• Microsoft unveils Guardian Agents to monitor, constrain, and manage risks posed by autonomous AI agents
• New controls dynamically adjust agent permissions, behavior, and access based on changing risk levels
• Company warns “agentic insiders” require continuous AI-driven oversight beyond traditional security controls

AI fans, your vocabulary is about to become two phrases richer: Agentic Insider, and Guardian Agent. These are the two phrases Microsoft used while announcing a swathe of new tools and features to help businesses secure their operations running AI agents.

A new Microsoft blog, published by Microsoft's Corporate VP for Identity and Network Access, Nadim Abdo, argues that agents, while not inherently malicious, can make mistakes or be manipulated and compromised.

Since they need to be granted relatively extensive permissions in order to be able to work properly, it creates the sort of risk that can prove disastrous if not properly handled.

Agentic Overwatch

Abdo also added that AI agents doing their master's bidding cannot be trusted with their own security.

"Satya Nadella, Chairman and Chief Executive Officer of Microsoft, recently described the trust architecture this new era requires: we must separate the supply of intelligence from the authority over it," Abdo writes.

"We should not ask a model to be its own security authority. The controls that determine what it can access, what actions it can take, and whether it should continue operating must sit outside the model, where the model cannot bypass or tamper with them."

So, oversight is the keyword here. But AI agents move at machine speed, something no human can match, and therefore humans are not properly equipped for the job. Instead, separate AI agents need to be introduced to keep their juniors in check. Earlier, Microsoft introduced Project Perception and the Microsoft cyber stack, to help organizations defend with agents.

Project Perception is an AI-powered cybersecurity system designed to coordinate specialized AI agents that analyze security data, identify vulnerabilities, and investigate different threats. The system uses three types of agents: red agents that probe for weaknesses, blue agents that investigate potential threats, and green agents that implement corrections and fixes.

Now, Microsoft says it is extending the project to secure “the agentic insider”. They’re calling it - Guardian Agent.

“Guardian Agents help security teams operate continuously, at machine speed and scale, while sensors, context, and actuators across the cyber stack help understand changing agent risk and continuously adapt protection in response,” Abdo explained. “Project Perception and Microsoft Agent 365 share signals, context, and a unified registry giving IT and security teams a consistent view of agents and risk, with the controls each needs.”

Drill sergeant

Abdo says Guardian Agents build on a security principle where non-deterministic intelligence needs to be surrounded with deterministic controls, independent evidence, and containment boundaries.

These controls are divided into seven aspects - identity, permissions, data, endpoints, network, cloud, and runtime - and these have four responsibilities: to know which agent exists, who owns them and why, and what they’re doing; to contain what each agents can access and do before it acts; to protect the data, tools, runtime, resources, and systems the agent uses; and to adapt protection continuously as behavior, context, and risk change.

In other words, the AI needs to be the drill sergeant for the drones, adapting to changes in real-time to make sure they’re doing good work and not going rogue.

Microsoft also said it is expanding discovery to 35 additional types of local agents, including Claude Code, OpenAI Codex, and GitHub Copilot. It also announced extending Microsoft Entra Agent ID with identity types for additional agent patterns, including both local and previously unknown agents discovered through security sensors.

With “Dynamic Permissions” and “Dynamic Agent Risk”, Microsoft wants to give businesses tighter control over what AI agents can access and do, while keeping tabs on their behavior and risk levels. Dynamic Permissions limit agents to the access they need for a specific task, while Dynamic Agent Risk checks for potential threats.

Dynamic Conditional Access can then respond to rising risks by restricting permissions, blocking certain operations, or asking for human approval. In more extreme scenarios, an agent can be stopped and isolated.

You don't need a bulky desktop tower to get a powerful Windows 11 Pro device. The Getorli GT106 mini PC is currently priced at $450 (was $500) at Amazon, and combines an eight-core AMD Ryzen 7 processor, 16GB of DDR5 RAM, and a 1TB SSD in a compact design. For the price and specs, our managing editor has dubbed this "the best mini PC under $500."

• Best for: Home offices, multitasking, multi-monitor workstations, media centers, and anyone who wants a compact system for everyday computing and lighter gaming.

• What we like: Alongside the Ryzen 7 chip and fast DDR5 memory, the mini PC also has integrated AMD Radeon 780M graphics and can drive up to four 4K monitors at a time via HDMI 2.1, DisplayPort 1.4, and two USB4 connections.

Today's best mini PC deal

GT106 Mini PC: was $499.99 now $449.99
This compact desktop features an eight-core Ryzen 7 H255 processor, Radeon 780M graphics, 16GB DDR5 memory, 1TB PCIe SSD, dual USB4 ports, quad 4K display support, Wi-Fi 6, and Ethernet.View Deal

Why we recommend it

The Getorli GT106 is powered by an AMD Ryzen 7 H255 processor with eight cores and 16 threads, running at up to 4.9GHz. There's also 16GB of DDR5 RAM, which comes installed as a single module, leaving the second slot available for an upgrade. For storage there's a 1TB PCIe 4.0 SSD, and a second M.2 slot, with support for up to 4TB capacity.

Alongside Wi-Fi 6 and Bluetooth 5.3, you get dual 2.5Gb Ethernet ports and two 40Gbps USB4 ports for connecting fast external storage and other peripherals.

It comes with Windows 11 Pro installed, and Getorli includes a VESA mounting bracket so you can tuck the device behind a compatible monitor.

For more of our favorites, we've tested the best mini PCs you can buy.

Price context & historical value

The Getorli GT106 is $50 below its usual $499.99 price. That's a 10% discount, bringing the mini PC below the $450 mark. I've seen it retail for as low as $404.99 in the past, however. But so long as it flies easily under the $500 mark, it's a great buy.

Should you buy it?

✅ Buy the Getorli GT106 if...

You want a small Windows 11 Pro mini PC with a Ryzen 7 processor, 16GB DDR5, and 1TB SSD. It's a useful option if you need several monitors or want to connect fast external drives via USB4.

❌ Skip the Getorli GT106 if...

You need a more powerful device. It ships with a single 16GB DDR5 module, which means it initially operates in single-channel memory mode, which can reduce integrated graphics performance.

• Shop all mini PCs and Amazon

• Rockstar Games has been working on Grand Theft Auto 6 for 11 years, as the artists needed to plan out the map first
• The game's roads are much narrower and denser than GTA 5's, which created "ripple effects," according to Bloomberg's Jason Schreier
• Schreier says head of development Aaron Garbut told him, "All these things sort of come [back] to bite you later"

Rockstar Games has reportedly been working on Grand Theft Auto 6 much longer than realized, since the studio needed to plan out the game's ambitious map first.

That's according to Bloomberg's Jason Schreier, who shared more details about his visit to Rockstar North in a YouTube video after releasing his behind-the-scenes GTA 6 article.

While it's common knowledge that full-scale development of GTA 6 began around 2018 and was originally planned to be made alongside Red Dead Redemption 2, pre-production on the game actually began much earlier.

"Something that I learned while talking to those folks [the developers at Rockstar North] is that there are a few artists at Rockstar who have actually been working on GTA 6 for more than 11 years since 2015, if not earlier," Schreier said. "2015 is when Rockstar started kind of planning out GTA 6 in the sense that it was conversations before like before that."

Schreier explained that the game's massive map, which is twice as big as GTA 5's world and more densely populated and interactive, was a priority for the studio and the first thing it planned.

"2015 is when they started going to Miami, actually starting working on the map of the game, which is always the first thing that Rockstar does on these big open-world games as they start planning out the map first," Schreier said.

He added that gradually over the next couple of years, top leadership then mobilized everyone to work on Red Dead Redemption 2, as the game needed a lot of help due to its ambitious systems. However, a few artists continued work on GTA 6's map.

Aaron Garbut, art director, current head of development and co-studio head at Rockstar North, also told him that "the roads are significantly narrower in GTA 6 than they were in GTA 5," which fans may have noticed in the trailers.

This created a more realistic depiction of a fictional Los Angeles, and less like a racing game with wider spaces to zip through like GTA 5. However, Rockstar's goal of making GTA 6's map have narrower roads and denser traffic from the beginning presented some technical challenges, or what Schreier called "ripple effects."

Garbut told Schreier, "All these things sort of come [back] to bite you later," because narrow, denser roads meant the team had to implement more surrounding buildings, and artists have worked on the concept for 11 years.

GTA 6 launches in under one month, on November 19, for PS5, Xbox Series X, and Series S. Rockstar recently released the cinematic GTA 6 launch trailer, which offers a deeper look into the game's story. Elsewhere, a new report claims the PC version could launch in early 2027.

Sharp packs 24 subjects into five-minute lessons with quizzes, streaks, and XP so learning fits a busy day.

The post Swap Doomscrolling for 5-Minute Lessons With This $50 Learning App appeared first on TechRepublic.

The iPhone Duo has a 4,883mAh battery. See how it compares with Samsung, Google, Motorola, and other premium phones.

The post iPhone Duo Battery Revealed: How It Compares to Samsung, Google, and More appeared first on TechRepublic.

The Justice Department has charged 28 people in California with voting crimes as Election Day nears. And, TikTok staff raised child safety concerns with executives, who failed to act, court records show.

(Image credit: Imen Ben Youssef / Hans Lucas)

CNET is partnering with sister sites PCMag, Mashable, ZDNET and others on an all-new app with news, reviews and videos. Our trusted expertise is a free download away.

The updated agent also can tell you if it’s going to rain and maybe you shouldn’t book that outdoor table.

Research on past recessions suggests downturns can accelerate automation. That raises the specter of a double whammy for workers.

It’s the year of window cams, and Baseus has my favorite model yet. Here’s why renters and second-story homes should consider it.

The new Kindle Colorsoft brings updated hardware and optional page-turn accessories, while Kobo Libra Colour offers built-in buttons and stylus support for less.

The post Kindle Colorsoft vs Kobo Libra Colour: Which E-Reader Should You Buy? appeared first on TechRepublic.

Kindle Scribe combines ebook reading with handwritten notes, while reMarkable Paper Pure focuses on writing and document workflows. We compare features, costs, and usability.

The post Kindle Scribe vs reMarkable Paper Pure: Which Should You Buy? appeared first on TechRepublic.

• Enterprises are being forced into software upgrades that they can't really get out of, report finds
• 91% have upgraded hardware in the past three years just to maintain software
• Enterprise protection legislation and software support obligations could be the answer

Legacy tech has become something of a hot topic in recent years with AI, but while hardware and data foundations have been core to the discussion around artificial intelligence, a new Origina report has revealed that around one-third of enterprise IT budgets are being earmarked for software upgrades.

The report found that 28% of large enterprise IT budgets are being spent on vendor-driven changes that they would not have otherwise chosen in a trend the company's calling a 'software tax'.

But more than just consuming budgets, software changes are also impacting worker time and leading to potentially lost productivity as a result.

Software upgrades cost more than you might think

"That's not a one-off cost, it's a recurring drag on capacity, and every year it repeats is a year those people aren't working on the initiatives that actually differentiate the business," former Liberty Mutual CIO James McGlennon and Origina board member James McGlennon explained.

More than four in five (84%) report that senior engineering staff always or often spend time dealing with vendor-driven software changes, taking them away from more meaningful work that they're employed to do. A similar number (80%) also reported that leaving a major software vendor could lead to widespread disruption within their business, ultimately forcing them to stay with that software vendor and enforcing vendor lock-in – with exit costs and migration technicalities to blame.

More than three-quarters now agree their software vendor dictates the timing and terms of changes, not them (78%), and that they have very little negotiating power (76%).

In many instances, enterprise customers don't even feel that an upgrade was totally necessary. Nearly all (91%) had retired fully functional hardware within the past three years just to comply with software vendor support deadlines. While Origina doesn't specifically state it, Windows 10's end of support is likely a large factor in this statistic.

Looking ahead, most (91%) support legislation to provider greater protection for enterprise customers, and four in five (82%) support software support obligations similar to what the EU has introduced for hardware repairability.

• A vulnerability in NordVPN's Windows app allowed local attackers to stop unprotected system services.
• The flaw, officially tracked as CVE-2026-106030, does not affect NordVPN version 8.12.0.0 or later
• Experts label the bug "clumsy," but the actual impact on users is minimal, requiring a device to already be compromised by malware.

Nord Security has rolled out an important security update for its Windows application to address a vulnerability that could allow a hacker to stop unprotected Windows services.

The flaw allowed a local, low-privileged attacker to act as a trusted process. From there, they could potentially stop unprotected Windows services or change their startup types to manual, leading to a denial of service for those specific applications.

When a top-tier provider, frequently ranked as the best VPN on the market, flags a security patch of this nature, it naturally turns heads. However, cybersecurity experts stress that while the coding error shouldn't have happened, the real-world danger to users is incredibly low.

The issue, officially tracked as CVE-2026-106030, was responsibly reported to CERT Polska by IT Security Expert Marcin Ressel. NordVPN appears to have fixed the vulnerability in version 8.12.0.0, and users are simply advised to ensure their app is fully up to date.

How the vulnerability works

(Image credit: NordVPN)

To understand the bug, you have to look at how Windows applications talk to each other using a feature called "named pipes."

According to an intelligence brief published by Mallory, the vulnerability allowed an attacker to connect to a specific NordVPN named pipe and invoke methods to stop services or change their start types without proper authorization.

Mike Williams, Technical Editor at TechRadar, explains the core of the issue: "Named pipes are a common means for Windows processes to communicate with each other. One process might send a message to another asking it to carry out some task."

Because this communication channel is incredibly powerful, it is a frequent target for threat actors. "Because named pipes are so sensitive, hackers often try to abuse them, to run these tasks without permission," Williams notes. "That means it's important to secure them, and make sure that only messages from legitimate processes are accepted. Any good developer knows this: it's a fundamental aspect of using named pipes."

Unfortunately, the NordVPN Windows app, specifically versions earlier than 8.12, fell short of this standard. "NordVPN has failed to do that, which makes this a clumsy bug. It should never have happened," Williams adds.

Why the real-world impact is minimal

While the technical details of the flaw, which carries a CVSS v4 base severity score of 6.8, sound alarming, the practical implications for the average user are far less dramatic.

In a LinkedIn post, the researcher who discovered the flaw, Marcin Ressel, described it as "a local privilege escalation with limited impact" that he found while "playing around with NordVPN for Windows."

Crucially, this exploit can't shut down critical, system-protected security services. Built-in defenses like Windows Defender and processes guarded by Windows Tamper Protection are entirely immune to this specific attack.

"There's no impact on the VPN itself, and you would have to have malware running on your PC to even cause the issue"

Mike Williams, Technical Editor

"But the impact of the bug is very small," Williams clarifies. "All an attacker can do is change the status of non-critical Windows services. That is, it can't turn off Windows Defender, but maybe it could disable something related to media playback. But even if it did that, it's only flipped a Windows setting, and once you understood the cause, you could just flip it back."

Furthermore, this vulnerability cannot be triggered remotely over the internet. An attacker would already need to have a local foothold on your machine to exploit it. There has also been no active exploitation of this flaw reported in the wild.

"In other words, there's no impact on the VPN itself, and you would have to have malware running on your PC to even cause the issue," Williams concludes. "(And if there's malware running on your device then you've much bigger problems than this.)"

If you are running NordVPN on Windows, your best defense is simple: open the app and confirm you are running version 8.12.0.0 or later.

Having been at Nvidia's RTX Spark event in San Francisco earlier this month, I was lucky enough to get a firsthand walk through of the many things the 'superchip' can do. Since then I've spent almost every waking moment putting the ARM-based RTX Spark through its paces and seeing what makes it so different from other chipsets that have come before.

The RTX Spark is Nvidia’s first System on a Chip (SoC). Combining an 18-core CPU and 5120 Cuda core Blackwell RTX GPU all in one, or a 20-core CPU and 6144-core GPU for its meatier version, it offers an integrated chipset in the vein of Apple’s M series chips. While Apple’s silicon is known for its oomph, that Blackwell RTX GPU means you’ll also get considerable gaming performance here to boot.

How does this translate in terms of raw numbers? It’s certainly got a lot going on under the hood, as the RTX Spark N1X chip I’ve been playing around with squeezes 20 cores into its CPU. While the 2,623 single-core performance I clocked on Geekbench 7.1 falls somewhat short of the 4,288 we got for Apple’s M5 chip, its 21,899 multi-core score massively outstripped the latter’s 17,926. That means that for tasks like content creation, software development and AI, you should be able to expect more oomph in your corner.

So given the potential it promises across agentic AI, creative workflows and gaming, let’s take a look at how it performs across each of those areas in turn.

Hybrid agentic AI

(Image credit: Future)

During the launch of Nvidia’s RTX Spark SoC, there was a huge focus on its potential for hybrid agentic AI. Now, cloud-based AI models are undoubtedly pretty powerful but I absolutely have security concerns about uploading my most private data to the cloud and I’m not entirely sure I want to blindly trust the companies behind some of these models.

Using local, open-source models that I can vet myself and that never transmit sensitive data outside of my network feels like a much safer alternative — but not everybody necessarily wants to cut themselves off from the power of frontier models. That’s why Nvidia has introduced a hybrid-system for the RTX Spark instead.

Instead, the hybrid-workflow that the RTX Spark supports allows you to select your choice of local models and ring-fence them for certain parts of an agentic workflow, while outsourcing specific, demanding tasks to cloud frontier models. As a result, users should be able to take a far more granular approach, deciding exactly how much they want to rely on public vs private models.

For people who like to vibe-code complex products or are generating demanding 3D scenes, not being forced to choose between binary security options of a lax, YOLO approach and a full-blown lockdown will likely be a valuable tool.

So now that I’ve spent some hands-on time with the RTX Spark, how has its agentic performance delivered on these promises? To get a broad picture of I ran the MLPerf Client benchmark to simulate data analyst and SWE agentic workflows using the Qwen 3 8B local model, with each run hitting an average duration of 63.7 and 50.8 seconds and processing up to 14,005 and 10,238 tokens respectively.

I also benchmarked some generative image and video workflows using ComfyUI. When testing workflows using Flux2.Dev (NVFP4 and Flux2.Klein 4B Base) NVFP4 models for image generation, it took 46.5 and 16.2 seconds to generate respectively, while the LTX 2.5 (Distilled) NVFP4 model took 88 seconds. That’s broadly in line with what Nvidia has been promising and shows how rapidly the RTX Spark can generate media using local models.

Raw data will only tell you so much though. That’s why I spent days training my own AI agent using Hermes and the Qwen3.8 27B open-source local model to see what the RTX Spark could do.

Certainly, it seems to have absolutely ripped through some tasks on its Extra High setting. Feeding it some analytics data in CSV form, I asked it to tidy up the spreadsheet, analyze the data and present its key findings and after 12 minutes of pondering and tweaking, it happily spat out a pivot table and some observations on the data. While I wasn’t hugely surprised by any of its conclusions, I’ve spent months poring through this data and the fact it could identify the most essential points without any of my secure information leaving the laptop was pretty impressive.

Some experiments I conducted involved a little more patience and tweaking of the agent’s workflow to get the best results though. The first time I asked my Hermes agent to connect to my online email client and provide a digest of the day’s emails, it spent 75 minutes trying various methods to exchange tokens with Outlook’s online client, before crashing unexpectedly. However, the second time I played a slightly more active role, providing it feedback when I felt it was drifting or had come up against hurdles and it successfully set up the brief in 70 minutes.

So with some patience and judicious prompting, you can definitely get some good results utilizing a local model. Additionally, you don’t have to go through this full process every time — ask Hermes to save something as a skill and it will automate this process in future or even schedule it as a recurring Cron Job. You can even ask it to work out ways to improve the efficiency of existing skills, meaning it should be easier to minimize the run time and resources they require in future.

Creative workflows

(Image credit: Future)

You’d be forgiven for thinking this means the RTX Spark laptops are predominantly focused on AI tools. But that couldn’t be further from the case: while a lot of the press surrounding the chip has been on its agentic AI tools, this really isn’t the sole string to its bow.

As you can tell from the fact that the RTX Spark is used in the latest Asus ProArt P16, creative work is a big focus for the new SoC. Even if you’re not interested in the power it offers for generative AI, the sheer number of RTX-compatible creative apps out there, whether that’s Blender, Adobe Creative Suite, Canva or Unity shows how many creative giants are throwing their weight behind the new SoC.

Eager to test how well the RTX Spark handles creative work, I used Puget Bench to automate a range of tasks in Photoshop and Premiere Pro. PugetBench’s standard Photoshop workflow took 10.83 minutes and netted a score of 11,801, while its standard Premiere Pro pipeline took 8.12 minutes and achieved a score of 76,234, putting it firmly in the range of a true creative workstation. That means it sits in the rough region of Apple’s M5 chip, which scored 13,755 and 69,887 respectively, besting it on complicated video work but not quite equaling it for still imagery.

For those who use generative AI in their work, I also benchmarked some generative image and video workflows using ComfyUI. When testing workflows using Flux2.Dev (NVFP4 and Flux2.Klein 4B Base) NVFP4 models for image generation, it took 46.5 and 16.2 seconds to generate respectively, while the LTX 2.5 (Distilled) NVFP4 model took 88 seconds. That’s broadly in line with what Nvidia has been promising and shows how rapidly the RTX Spark can generate media using local models.

While I’m neither an aspiring cinematographer or a budding Blender artist, I’ve spent decades tinkering with static images so I dug out an old 4K, 300dpi Photoshop file. It was already pretty huge as it featured lots of hi-res layers, clipping masks and adjustment layers but I tripled the number of layers for good measure and added a bunch of filters and effects. Processing time remained less than a second and even halving the image size only took about five, which is pretty impressive given the sheer number of layers and groups I was working with.

And the RTX Spark should even be able to speed up these applications’ built-in AI tools, thanks to the recently announced collaboration between Nvidia and Adobe. Optimizing the apps to take full advantage of the chip’s GPU, Adobe promises that RTX Spark laptops offer 2x AI processing, editing, color and effects. That’s a real advantage for anyone that uses these AI tools as part of their workflow.

Gaming performance

(Image credit: Future)

Given Nvidia’s history, gaming is naturally also a big focus for the RTX Spark chipset. As most games run on x86 architecture, SoCs haven’t been the most natural home for gaming in the past. However, Nvidia’s stature in this space means plenty of developers launched RTX Spark native versions of their games when the chip was launched on October 16, including Cyberpunk 2077, Doom: The Dark Ages, and 007 First Light — while the Call of Duty series is due to land in 2027. And for those that aren’t yet compatible, Prism emulation will mean you’ll be able to run many titles with only a slight impact on your CPU performance.

When it comes to gaming performance, the RTX Spark is certainly no slouch. It netted a 3DMark Port Royal score of 8,133, Steel Nomad score of 2,846 and a Speed Way score of 2,872. That’s a touch below Asus Zephyrus G14 with Nvidia GeForce RTX 5070 Ti’s respective 10,119, 3,488 and 4,031 scores. But it’s considerably better performance than you’d get out of most laptops with an integrated GPU such as the pre-RTX-Spark Dell XPS 16, which secured values of 3,888, 1,488, and 916.

But while these simulated benchmarks are helpful for getting a broad view of gaming performance, in-game benchmarks are the best way of gaining real insight into how actual titles perform. While testing the RTX Spark, Cyberpunk 2077 delivered a stable 30.16fps at 1080p on Raytracing: Ultra settings and Doom: The Dark Ages averaged out at 48.58fps at 1080p — very similar to the 32.33fps and 55.63fps we got out of the Asus TUF Gaming A14 with Nvidia GeForce RTX 5060.

Nothing compares to actually trying out games first hand though, so I fired up a few of my favorites and tried them out. Cyberpunk 2077 honestly felt pretty playable even using the above Raytracing Ultra setting and once I’d switched down to Ray-tracing Medium, it ran at a wonderfully fluid 60fps.

Meanwhile, running Gears of War: E-Day at Ultra settings felt as smooth as tahini drizzled over a marble countertop. As I crunched through waves of the Horde, there wasn’t any noticeable slowdown even at Ultra settings, with the action never dipping below a seriously fluid 115fps. So it might not match the RTX 5090 for sheer power but anyone barring elite gamers will find a lot to like here still.

So even though it’s naturally not at the level of dedicated gaming laptops like the MSI Vector A18 HX A9W, I’m still pretty impressed with gaming on the RTX Spark chip. And that only becomes more pronounced when you compare it to similar lightweight SoC systems. While the MacBook Pro M5 is similarly impressive in terms of creative performance, it was barely able to scrap 11.5fps in Cyberpunk 2077 on Raytracing: Ultra settings.

As a result, if you’re looking for a chipset that will give you not just great performance in one specific task but great performance across everything you do, the RTX Spark has proven it will be a solid choice.

The RTX 5080 offers more processing power, but the RTX 5070 Ti shares its 16GB VRAM and DLSS capabilities. Here's how performance, power, and 2026 pricing compare.

The post RTX 5070 Ti vs RTX 5080: Which GPU Is Better Value in 2026? appeared first on TechRepublic.

The new scholarship tax credit program could help public school families pay for some educational services, but advocates worry it could do more harm than good.

Here's what's behind the fluctuating costs of everyday grocery items such as ground beef, eggs and chocolate chip cookies.

The warnings surfaced in a trove of unreported court records. "I have been fighting about fixing this for months with people who didn't think the harms are severe enough," wrote one executive.

(Image credit: Imen Ben Youssef)