Intel’s Arc Pro B60 Dual offers pro-grade memory at a fraction of Nvidia’s price
This dual-GPU rig from Maxsun delivers workstation power
Each GPU gets one DisplayPort and one HDMI, avoiding OS overload in multi-GPU workstations
At Computex 2025, Maxsun unveiled a striking new entry in the AI hardware space: the Intel Arc Pro B60 Dual GPU, a graphics card pairing two 24GB B60 chips for a combined 48GB of memory.
Servethehomeclaims Maxsun envisions these cards powering dense workstation builds with up to four per system, yielding as much as 192GB of GPU memory in a desktop-class machine.
This development appears to have Intel's implicit approval, suggesting the company is looking to gain traction in the AI GPU market.
A dual-GPU card built for AI memory demands
The Arc Pro B60 Dual GPU is not designed for gaming. Instead, it focuses on AI, graphics, and virtualization tasks, offering a power-efficient profile.
Each card draws between 240W and 300W, keeping power and thermal demands within reach for standard workstation setups.
Unlike some alternatives, this card uses a blower-style cooler rather than a passive solution, helping it remain compatible with conventional workstation designs. That matters for users who want high-end performance without building custom cases or cooling systems.
Still, the architecture has trade-offs. The card relies on x8 PCIe lanes per GPU, bifurcated from a x16 connector. This simplifies design and installation but limits bandwidth compared to full x16 cards.
Each GPU also includes just one DisplayPort and one HDMI output. That design choice keeps multi-GPU setups manageable and avoids hitting OS-level limits, older Windows versions, for example, may have trouble handling more than 32 active display outputs in a single system.
The card’s most intriguing feature may be its pricing. With single-GPU B60 cards reportedly starting around $375 MSRP, the dual-GPU version could land near $1,000.
If that estimate holds, Maxsun’s card would represent a major shift in value. For comparison, Nvidia’s RTX 6000 Ada, with the same 48GB of VRAM, sells for over $5,500. Two of those cards can push costs north of $18,000.
Even so, Intel’s performance in professional applications remains an open question. Many creative professionals still favor Nvidia for its mature drivers and better software optimization.
The Asus Zenbook A14 weighs less than a MacBook Air, and it's on sale for the first time ever. But the deal ends tomorrow, so you'll need to act fast to score this discount.
After many months of speculation, Google finally showed off its still-early-day Android XR smart glasses prototype. It was an impressive live demo, with a live translation portion that went off well but not without hitches. Still, it got the crowd at Google I/O going, and right after that opening keynote wrapped, I strolled around the Shoreline Amphitheater to find a pair to try.
Much like my time with Project Moohan, the prototype Android XR headset that Google and Samsung are working on, I only spent about five minutes with these prototype glasses. And no, it wasn’t a sleek frame made by Warby Parker or a wild one from Gentle Monsters – instead, it was the pair Google demoed on-stage, the prototype Android XR glasses made by Samsung.
As you can see above, much like Meta Ray-Bans and unlike Snapchat Spectacles (the first gen), these prototypes look like standard black frames. They're a bit thicker on either the left or right stems, but they’re also loaded with tech – though not in a way that screams it from the outside.
It was a short, pretty rushed demo, but certainly a compelling one.
(Image credit: Jacob Krol/Future)
The tech here is mostly hidden – there is a screen baked into the lens, which, when worn, appears as a little box when it’s showing something larger. Otherwise, when I first turned the glasses on, I saw the time and the weather hovering at the top of my field of vision.
When I pressed the button on the right stem to capture a photo, it almost flashed transparently larger in my field of vision. Neat and a bit more present way of capturing than on the screen-less Meta Ray-Bans.
These are both cool, and during the keynote, Google also shared that the screens could be used for messaging, calls, and translating as well, but I didn’t get to try that. While I couldn’t ask for directions myself, a Google rep within my demo was able to toss up what navigation would like, and this feature has me more excited about smart glasses with a screen built-in.
Why? Well, it was that the experience of navigating doesn’t get in the way of my field of view – I can simply still look straight forward and see at the top that in 500-feet or 50-feet that I need to make a right onto a specific avenue. I don’t need to look down at my phone or glance at my wrist, it’s all housed in just one device.
If I need more details or want to see my route, I could glance down to see a mini version of the map, which moved as I moved my head. If I wore these in NYC, I could walk normally and glance at the top to see directions, but when safely stopped and not in the way of others, I could look down to see my full route. That’s pretty neat to me.
(Image credit: Jacob Krol/Future)
The projected screen itself had good-enough quality, though I’m not sure how it performs in direct sunlight, as I tested these in a little room that Google had constructed. It’s important to remember that this is still a prototype – Google has several brands onboard to produce these, but there isn’t an exact timeframe. Developers will be able to start developing and testing by the end of the year, though.
This year, the Project Moohan headset, which also runs Android XR, will arrive. Samsung will ship the headset in a to-be-revealed final version, which could build support from third parties and let Google get more feedback on the platform.
Gemini, Google’s very wise AI assistant, blew me away on Project Moohan and was equally compelling on the Android XR glasses. I asked it for the weather, and got it to give me an audio report of the next few days, had it analyze a replica of a painting, and even look at a book, tell me the reviews, and where I could purchase it.
That power of having Gemini in my frame has me really excited for the future of the category – it’s the audio responses, the connection to the Google ecosystem, and how it plays with the onboard screen. It remains to be seen how Samsung’s final design might look, but it will likely sit alongside several other Android XR-powered smart glasses from the likes of Warby Parker, X-Real, and Gentle Monster, among others.
I’ve long worn Meta Ray-Bans and enjoy those for snapping unique shots or recording POVs like walking my dog Rosie or riding an attraction at a Disney Park. Similarly, I really enjoyed the original version of the Snapchat Spectacles, but the appeal wore off. Those both did only a short – or in the case of the Spectacles, very short – list of functions, but Android XR as a platform feels a heck of a lot more powerful, even from a short five-minute window.
While the design didn’t sell me on Samsung’s prototype, I have high hopes for the Warby Parker ones. Seeing how Gemini’s smarts can fit into such a small frame and how a screen can be genuinely useful but not overly distracting really has me excited. I have a feeling not all of the Android XR glasses will appeal to everyone, but with enough entries, I’m sure one of them will pair form with function in a correct balance.
Gemini in glasses feels less like the future, and considering this new entry, my eyes are set to see what Meta's does next and what Apple's much-rumored entry into the world of smart glasses will look like.
You might also like
from Latest from TechRadar US in News,opinion https://ift.tt/87MFPBu
Dell Pro Max Plus laptop is the first to feature an enterprise-grade discrete NPU
It features a Qualcomm AI 100 PC Inference Card with 32 AI-cores and 64GB memory
This offers the opportunity to carry out high-intensity AI tasks, even on the move
Dell has unveiled an AI PC with a never-before-seen feature it hopes will spur on the next levels of productivity.
Revealed at Dell Technologies World 2025, the new Dell Pro Max Plus laptop is the first to feature an enterprise-grade discrete NPU, offering the opportunity to carry out high-intensity AI tasks even on the move.
The mobile workstation features a Qualcomm AI 100 PC Inference Card with 32 AI-cores and 64GB memory, which Dell says should be more than enough to handle the needs of AI engineers and data scientists deploying large models for edge inferencing.
“Personal productivity is being reinvented by AI,” Dell said, “the install base of a billion and half PCs is ageing, and it’s being replaced with AI innovation.”
“The Windows 10 end of life is coming, and we are ready - Dell is the leader in commercial AI PCs, and we’re further distancing ourselves from the competition.”
The CEO highlighted the new Dell Pro Max device during his keynote address, noting it would be ideal for developers and scientists, offering up to 20 petaflops of performance due to embedded Nvidia GB300 hardware, and up to 800GB of memory - enough to run and train models with a trillion parameters.
“Today’s PCs are becoming AI workstations - blazing fast, all-day battery life powered by NPU and GPU innovation," Dell declared.
Intel’s four-year-old Optane P5800X still outpaces the SN8100 in real-world speed tests
SanDisk’s new WD Black SN8100 PCIe Gen5 SSD is fast, efficient, and engineered to meet the demands of gamers and power users alike.
The drive uses a PCIe Gen5 x4 interface and is available in 1TB, 2TB, and 4TB capacities. Built around SanDisk's in-house 8-channel controller and BiCS 3D TLC NAND, it supports read speeds of up to 14.5 GB/s and write speeds up to 12.7 GB/s, placing it among the fastest Gen5 drives currently available.
However, despite the SN8100’s cutting-edge design and impressive benchmarks, Intel’s now-defunct, four-year-old Optane P5800X still holds the crown as the fastest SSD in real-world use.
Benchmarks suggest top speeds - but not across the board
In synthetic benchmarks like CrystalDiskMark and ATTO, the SN8100 breaks lab records for sequential throughput and random reads, reaching up to 2.3 million IOPS.
According to TweakTown, “this SSD is like none other; it’s at least 20% more powerful than any flash-based SSD we’ve ever encountered.”
It also demonstrates notable efficiency, consuming just 7 watts under load and requiring no active cooling, making it a serious contender for best SSD or the best portable SSD for enthusiast builds.
Still, synthetic benchmarks don’t always reflect real-world performance. In practical transfer tests, the SN8100 ranked ninth overall, indicating that while it's extremely fast, it's not without limitations, and it doesn't dethrone the Intel Optane P5800X.
Launched in 2021, the P5800X remains unmatched in real-world responsiveness and latency. While its sequential read speeds top out at 7.2 GB/s - slower than the SN8100 - its random read/write IOPS exceed 4.5 million, and latency frequently drops below 10 microseconds. That’s where it truly shines.
Flash-based SSDs like the SN8100 still rely on garbage collection and page-level management, leading to occasional latency spikes during small, random workloads. In contrast, the P5800X maintains consistent performance under heavy load, with no significant dips, a key reason why it’s still regarded as the fastest SSD ever made.
That said, the SN8100 is an impressive drive in its own right. It's a customized version of Silicon Motion’s SM2508 controller, enhanced with proprietary technologies like nCache 4.0 and WD Black Gaming Mode.
It also fits into the Sony PlayStation 5’s expansion slot, achieving read speeds of 6,550 MB/s in that setup, well above the console’s minimum requirement. However, with a price tag of $280 for the 2TB model, it clearly belongs in the premium tier.
If you’ve ever wondered how much one of these giant-capacity SSDs might set you back, the answer is: maybe not quite as much as you’d expect, as although early estimates placed its price close to $14,000, but you can actually pick up the drive from Tech-America for a much more affordable $12,399.
Obviously, this isn’t a drive for your typical PC rig - it uses a PCIe 4.0 interface and comes in U.2 (available now) and E1.L (expected later this year) form factors. It’s aimed at enterprise storage environments handling large-scale AI, machine learning, and data-intensive workloads.
Longer lasting QLC
The drive is built with 192-layer QLC NAND. With endurance rated at 0.60 drive writes per day and a total of 134.3 petabytes written over five years, the 122.88TB model is designed to last longer than earlier QLC offerings.
Solidigm, a US-based subsidiary of SK Hynix, reportedly tested the drive under extreme conditions. Running 32KB random writes at full load, the drive operated continuously for five years and retained around 5 percent of its life.
Performance claims include up to 930,000 IOPS for 4K random reads and 7.4GBps for sequential reads.
Solidigm markets its large SSD as a solution to space and power constraints in data centers, claiming that replacing traditional hybrid systems with its all-QLC drives could reduce rack usage from nine to one and cut power consumption by around 90 percent.
The drive joins other high-capacity SSDs announced in 2024, including models from Phison, Samsung, and Western Digital. Phison’s SSD supports PCIe Gen5 and offers faster peak throughput, though the D5-P5336 delivers a higher endurance rating and greater storage density.
Kioxia CM9 Series SSDs use 8th gen BiCS FLASH for enterprise performance
Faster NAND speeds and power efficiency support AI and data centers
Offers 61.44TB max, dual-port design, and massive write improvements
Kioxia has announced its CM9 Series PCIe 5.0 NVMe SSDs, marking the first enterprise drives built using its 8th generation BiCS FLASH 3D TLC memory.
With PCIe 5.0 and NVMe 2.0 support, the CM9 SSDs are designed to meet modern standards for data center storage by offering high-efficiency storage capable of supporting AI, machine learning, and high-performance computing.
These new SSDs feature CMOS directly bonded to array (CBA) architecture, an update designed to improve performance, power efficiency, and memory density. Kioxia’s use of CBA-based flash architecture promises faster NAND interface speeds and lower latency, which helps the drives deliver quicker data access and improved power efficiency.
Top-tier bit density
Compared to the previous CM7 series, the CM9 line shows increases of about 65% in random write speeds, 55% in random read, and 95% in sequential write speeds.
The CM9 SSDs, currently sampling to select customers, are built to handle read-intensive and mixed-use workloads in enterprise data centers and offer capacities of up to 61.44TB in 2.5-inch form and 30.72TB in E3.S configurations.
The drives are compatible with both the NVMe-MI 1.2c and OCP Datacenter NVMe SSD 2.5 specifications, and support dual-port configurations, making them suitable for enterprise environments where reliability and continuous access are critical.
Kioxia, which recently helped Linus Tech Tips smash the Pi calculation world record, says gains in power efficiency include roughly 55% better sequential read and 75% better sequential write performance per watt.
Although it’s early in the lifecycle of the CM9 Series, the specs and performance numbers suggest the company is aiming to strengthen its position in high-performance enterprise storage.
Axel Stoermann, Vice President and CTO for Embedded Memory and SSD, Kioxia Europe GmbH, said, “Alongside processing power and energy efficiency, memory is fundamental to enable AI, machine learning, and high-performance computing applications. The CM9 Series powered by our BiCS FLASH generation 8, is designed to address these storage demands, providing top-tier bit density, rapid data transfer, and outstanding power efficiency, all of which contribute to the superior performance of our SSDs."
Google I/O events are an often frustrating glimpse of the near future, with a lot of shiny software toys scheduled to land sometime "in the coming months". That often means a long wait of up to a year, so for Google I/O 2025 we've rounded every new announcement that you can actually try today.
Naturally, some of the features below come with restrictions – a few are only available to try now in the US, while some are restricted to subscribers of Google's AI Pro or AI Ultra tiers. But many have also rolled out worldwide, so there are new features to take for a spin even if you don't currently pay Google a cent.
What's missing from the list below and coming at a later date? Quite a bit actually, including some of the more futuristic ideas like Google Beam and Android XR, and it also isn't clear how long we'll have to wait for a worldwide rollout of AI Mode for Search, Veo 3, Flow, Virtual Try On in the Shopping app, and Google's top-tier AI Ultra plan.
Still, there are quite a few things from Google I/O 2025 to keep us amused in the meantime, so here's a list of the ones that are available to try today...
Google completely upended its golden goose, Search, at I/O 2025 this week, announcing several new features to stave off the threat of ChatGPT – and the biggest was arguably the US rollout of AI Mode.
If you're in the US and aren't seeing the new tab in Search (or in the search bar of the Google app), it's likely because Google said it'd be a gradual roll-out "over the coming weeks".
We've been using it for a while, though, and have put together a guide on how to master the new AI mode. It shouldn't be your go-to for everything, but we've concluded that "if you’re researching, planning, comparing, or learning, AI Mode can be a real comfort". Google hasn't yet commented on when it'll get a worldwide launch, but we'd imagine it'll be sometime this year.
Arguably the biggest breakthrough moment at Google I/O 2025, Veo 3 is the first AI video generator that can deliver synchronized audio (including speech) alongside its video creations. And it's available to try now for a lucky few, if you're in the US and on the new Gemini Ultra plan.
Granted, that is a pretty small group of people, but we had to include it in this list because it is actually available today for those lucky peeps, and US enterprise users on the Vertex AI platform.
The amount of processing power required for Veo 3 could mean a relatively slow rollout elsewhere, and Google has hinted as much by also releasing new features for Veo 2 like the ability to give it reference scenes.
Not sure how to weave all of your AI videos together into a cohesive whole? Google also addressed that issue with a new AI video editor called Flow – and like Veo 3, it's out now for AI Pro and Ultra subscribers in the US.
It's a bit like a Premiere Pro that you can operate entirely with natural language, to avoid learning keyboard shortcuts or complex menus. To get an idea of how it works, check out Google's short tutorial.
Impressively, it goes as far as giving you menus of camera moves like 'dolly out' and 'pan right', so you don't even have to describe them. Google has also at least promised that it's "coming soon" to more countries, so we're hopeful of a wider rollout in 2025.
The big smartphone story of Google I/O 2025 was the full rollout of one of the best AI tools around on Android and iOS – Gemini Live.
Like ChatGPT's Advanced Voice Mode, Gemini Live is an AI assistant that you can chat to using your voice. The most useful part, though, is that you can also give it eyes using your phone's camera to get help with whatever's in front of you or on your screen.
To conjure the assistant, open the Gemini app on iOS or Android, tap the Gemini Live icon (on the far right of the text input box), and start chatting away.
available worldwide in the Gemini app, Whisk, Vertex AI
(Image credit: Google)
Google didn't just level-up its AI-generated video at I/O 2025 – we also got a new Imagen 4 model for whipping up still images in higher resolution (now up to 2K) than before.
The latest Imagen (which is available now in the Gemini app, Whisk, Vertex AI and across Google Workspace) also showed that it's been working hard on one of its main weaknesses – handling text.
This means that scenes involving typography should no longer be a jumbled mess of weird characters and look more realistic. While Imagen 4 is available to use for free, it does come with usage limits – you can expect 10-20 image generations on a free plan, while Gemini subscribers get a more generous 100-150 generations a day.
Okay, Gemini 2.5 Flash isn't brand new, but it was given a big upgrade at Google I/O 2025 – and it's now available to everyone to dabble with in the Gemini app.
In fact, Gemini 2.5 Flash is now the default model in Google's Gemini chatbot, because it's apparently the fastest and more cost-efficient one for daily use. Some of the specific improvements, over its 2.0 Flash predecessor, include a greater ability to understand images and text.
Wondering how it compares to ChatGPT 4o? We've already compared the two to help you see which might be the best for you. Spoiler: it's a close call, but Gemini 2.5 Flash is particularly appealing if you live in Google's world of apps and services.
Need a coding assistant to speed up your workflow? Google has just given Jules (first introduced as a Labs experiment last December) a wider public beta rollout, with no waiting lists.
Jules is a bit more than a coding copilot – it can autonomously beaver away on fixing bugs, writing tests and building new features without any input from you. It works 'asynchronously', which means it can work on various tasks without waiting for them to finish.
Google says Jules isn't trained on your private code and that your data stays within its private environment. With autonomous agents on the rise, it certainly looks worth dabbling with if you could do with some coding assistance.
Google Shopping has had a 'Try On' feature for clothes since 2023, but it got a big upgrade it got at Google I/O 2025. Rather than using virtual models to show you how your chosen clothes might fit, it now lets you upload a photo of yourself – and uses AI to help you avoid the hassle of changing rooms.
Once you've uploaded a full-length photo of yourself, you'll start to see little "try it on" buttons when you click on outfits that are served up in the Shopping tabs search results. We've taken it for a spin and, while it isn't flawless, it does give you a solid idea of what some clothes will look like on you. And anything that helps us avoid real-world shopping is fine by us.
Google brought its 'Deep Research' feature to Gemini Advanced subscribers (now Gemini Pro) in late 2024. And now the handy reports tool has given a particularly useful upgrade – the ability to combine its research of public data from the web with any private PDFs or images that you uploads.
Google provided the example of a market researcher uploading their own internal sales figures so they could cross reference them with public trends. Unfortunately, you can't yet pull in docs or data from Google Drive and Gmail, but Google says this is coming "soon".
10. Gemini quizzes
available worldwide on Gemini desktop and mobile
college students in the US and UK can also get a free Gemini AI Pro upgrade for the whole school year
Google is particularly keen to get students using its Gemini app – not only did it extend its free access to Google AI Pro for school and university students to new countries including the UK, it also added a new quiz feature to help with revision.
To start a quiz, you can ask Gemini to "create a practice quiz" on your chosen subject. The most useful part is that it'll then make a follow-up quiz based on your weaknesses in the previous test. Not that you have to be studying to make use of this feature – it could also be a handy way to sharpen your pub quiz skills.
If you're a student in the US, Brazil, Indonesia, Japan and the UK, you can get your free year of Gemini AI Pro by signing up on Gemini's students page – the deadline is June 30, 2025 and you will need a valid student email address.
11. Google Meet speech translation
available to Google AI Pro and AI Ultra subscribers
initially only in English and Spanish, more languages coming soon
We're particularly looking forward to trying out Google Beam this year, with the glasses-free 3D video calls (formerly known as Project Starline) heading to businesses courtesy of HP's new hardware. But a new video calling feature you can try now is Google Meet's near real-time translations.
Available now for AI Pro and Ultra subscribers in beta, the feature will provide an audible translation of your speech (currently in English to Spanish, or vice versa) with a relatively short delay. It isn't seamless, but we imagine the delay will only reduce from here – and Google says more languages are coming "in the next few weeks".
Google switched up its AI subscription plans at Google I/O 2025, with 'Gemini Advanced' disappearing and being replaced by AI Pro and new 'VIP' tier called AI Ultra.
The latter is currently US-only (more countries are "coming soon") and costs a staggering $250 a month. Still, that figure does give you "the best of Google AI", according to the tech giant, with AI Ultra including access to Veo 3 with native audio generation, Project Mariner, and the highest usage limits across its other AI products. You also get YouTube Premium and 30TB of storage thrown in.
The AI Pro tier ($20 a month) still gets you access to Gemini, Flow, Whisk, NotebookLM and Gemini in Chrome, but with lower usage limits and cloud storage of a mere 2TB.
If you're an AI power user and like the sound of AI Ultra, Google is currently offering it at 50% off for your first three months. Don't tempt us, Google...
Maybe I'm reading too much into things, but deciding to include the ever-stressful clock could indicate we're counting down to the finale.
There are also moments of Carmy (Jeremy Allen White) trying to remember why he loves what he does and approaching a meeting with his mother (Jamie Lee Curtis), so even if it isn't the final season, there's still plenty of drama ahead, and it seems to be building to something big.
An official plot for The Bear season 4 reveals: "Carmen 'Carmy' Berzatto (Jeremy Allen White), Sydney Adamu (Ayo Edebiri), and Richard 'Richie' Jerimovich (Ebon Moss-Bachrach) push forward, determined not only to survive, but also to take The Bear to the next level. With new challenges around every corner, the team must adapt, adjust and overcome."
There's still quite a lot to speculate on regarding one of the best Hulu shows, so we'll have to wait patiently to find out more about season 4. It won't be long, though, and much like the clock in the trailer, I'm counting down the hours.
