Latest news with #TechPowerUp
Arabian Post
16-06-2025
- Business
- Arabian Post
EdgeCortix's SAKURA‑II Elevates Raspberry Pi 5 with On‑Device Generative AI
EdgeCortix has launched its SAKURA‑II M.2 AI accelerator for the Raspberry Pi 5 and other Arm‑based platforms, enabling high‑performance, energy‑efficient execution of generative AI at the device edge. With 60 TOPS and 30 TFLOPS performance within an 8–10 W power envelope, the SAKURA‑II module supports advanced models including Llama 2, Stable Diffusion, Vision Transformers and VLMs on compact, affordable hardware. Dr Sakyasingha Dasgupta, EdgeCortix founder and CEO, highlighted that the integration 'opens the door for innovators and enterprises around the world to build smarter, faster, and more efficient edge AI‑driven devices'. This remark underscores a clear strategic pivot: migrating AI workloads away from cloud dependence and embedding them directly into low‑power devices. Venture partner Sailesh Chittipeddi echoed this view, emphasising the appeal for IoT and edge application engineers seeking scalability without datacentre overhead. The core of SAKURA‑II is EdgeCortix's DNA architecture, offering high memory bandwidth—up to 68 GB/s—and support for dual-channel LPDDR4x. This combination optimises batch‑1 inferencing for real‑time AI tasks while maintaining minimal latency and maximised compute utilisation. ADVERTISEMENT Market response has been mixed. It's FOSS notes the roughly US $349 price tag for the M.2 module, with no explicit mention of shipping costs, urging buyers to clarify before purchase. A TechPowerUp forum debate revealed cost‑sensitive hobbyists comparing it to a US $130 AI HAT offering about 26 TOPS. One user characterised SAKURA‑II as 'on another completely different ballpark' due to its RAM and bandwidth advantages for advanced applications. For industrial users, particularly those operating in space‑weight‑power‑cost constrained environments such as drones, robotics, smart agriculture or security, SAKURA‑II's offline capabilities are pivotal. By enabling autonomous AI without cloud reliance, organisations can enhance resilience and reduce latency in mission‑critical operations. Academic research on efficient edge deployment reinforces this evolution. A paper from June 10, 2025 demonstrated quantised YOLOv4‑Tiny object detection on Raspberry Pi 5, achieving 28.2 ms inference per image at 13.85 W power consumption. While this study used CPU‑based INT8 quantisation, the performance and consumption metrics set a baseline that illustrates SAKURA‑II's potential leap in efficiency and speed via dedicated silicon acceleration. EdgeCortix's positioning also aligns with wider trends in AI hardware development. Their DNA technology enables dynamic reconfiguration and mixed‑precision processing approximating FP32 accuracy—important for generative AI workloads that balance performance with model fidelity. Partners like SoftBank and Renesas have emphasised the importance of this co‑design approach, blending hardware IP with compiler‑driven software stacks to reduce TCO and accelerate time‑to‑market. Industry analysts see SAKURA‑II and similar accelerators as closing the gap between cloud‑scale AI and embedded edge use cases. By supporting multi‑billion‑parameter models on hand‑held devices, they suggest a future where even small autonomous systems can perform complex tasks like content generation, language parsing and computer vision locally—without connectivity or latency constraints. However, barriers remain. The ~$349 entry price may deter hobbyists and small‑scale developers, contrasted with cheaper model‑specific HAT solutions. Adoption may hinge on use case value—where the benefits of on‑device Generative AI outweigh acquisition and integration costs. Enterprise rollouts will need to consider software support, model compatibility, and real‑world inference benchmarks – details which are pending independent testing. EdgeCortix provides MERA, its compiler and runtime platform, enabling developers to deploy models across heterogeneous Edge AI systems, signalling strong software ecosystem support. This software‑hardware synergy contrasts with many accelerators that must rely on limited driver support or manual optimisation. The extension to Raspberry Pi 5 is significant. As one of the most accessible single‑board computers, Pi 5 offers a global developer base and extensive community support. Pairing it with SAKURA‑II could catalyse novel applications—from mobile robotics and decentralised AI devices to educational platforms that illustrate advanced AI concepts. Going forward, key indicators to watch will include independent benchmark results, broader platform support, and commercial deployments in agriculture, defence, and industrial automation. The ROI calculation will depend on whether the performance and efficiency gains translate into measurable gains—lower energy costs, reduced latency, or enhanced autonomy.
Yahoo
09-06-2025
- Yahoo
Snapdragon X2 Elite leak teases a monster 18-core chip with 64GB RAM
When you buy through links on our articles, Future and its syndication partners may earn a commission. Just when we thought Qualcomm's Snapdragon X Elite was its boldest move, a fresh leak suggests the company's next chip is gearing up to break records. The newly uncovered Snapdragon X2 Elite variant, reportedly codenamed SC8480XP, boasts a whopping 18-core CPU configuration and up to 64GB of LPDDR5X RAM — the highest we've ever seen in Qualcomm's Windows-on-ARM efforts. The leak, which comes via a test platform uncovered by @KOMACHI_ENSAKA and reported by TechPowerUp, indicates that Qualcomm may be preparing a chip powerful enough to blur the line between ARM laptops and high-end desktops. According to the leak, the X2 chip packs 18 Oryon V3 CPU cores, a 50% jump from the already formidable 12-core Snapdragon X Elite. That means more threads, better multitasking, and higher headroom for compute-heavy apps like video editing and local AI inference. The core architecture itself has evolved. Qualcomm's third-gen Oryon design, derived from the server-class NUVIA Phoenix cores, is expected to deliver up to 30% better single-thread performance, with even bigger gains in multi-core workloads. That puts Qualcomm squarely in the fight with AMD and Intel's most efficient silicon. But the spec sheet doesn't stop at core counts. Testing configs are reportedly running with up to 64GB of LPDDR5 RAM, representing a full 2x increase over current X Elite laptops. Qualcomm is also moving toward a System-in-Package (SiP) approach, directly embedding RAM and storage onto the processor. It's a move that echoes Apple Silicon but goes one step further, minimizing latency by eliminating external buses. There is a catch, though: SiP means that memory (and possibly storage) is integrated into the chip itself, so you won't be upgrading anything. What you buy is what you get. What's interesting here is that the X2 Elite is being tested with full-blown desktop cooling solutions, including a 120mm all-in-one liquid cooler. That's not standard issue for a thin-and-light laptop; it's desktop-tier gear. And it suggests that Qualcomm may be prototyping desktop or high-TDP laptop configurations. This would mark a significant departure from Qualcomm's previous strategy, which focused on ultra-portables and long battery life. It now seems that the company wants to scale ARM performance into new product classes, potentially rivaling Apple's M-series desktops and AMD's Strix Halo APU for creative workloads and local AI inference. It's worth noting that these tests are reportedly being conducted internally, and it's unclear whether such high-TDP devices will ever hit shelves, but the fact that Qualcomm is even testing them speaks volumes. It's also worth noting that Qualcomm might be positioning itself to compete more aggressively in the high-performance laptop segment, with the increased core count and memory support potentially making Arm-based laptops more appealing to power users. Regardless of how impressive the chip sounds, Windows on Arm still has work to do. Compatibility and optimization remain sore spots, even despite Microsoft's recent Copilot+ PC initiative and more OEMs embracing Snapdragon silicon. Qualcomm's hardware might be catching up or even surpassing Apple and Intel on paper, but real-world performance will depend on software doing its part. There's also the matter of timing. Qualcomm is expected to unveil new Snapdragon X-series chips at its Snapdragon Summit 2025, scheduled for September 23-25 in Hawaii. It's unclear whether the 18-core X2 Elite will headline that event, but when Laptop Mag spoke to Qualcomm leaders at Computex in May, all questions about the X2 were met with "come to Snapdragon Summit." If leaks are true, the 18-core Snapdragon X2 Elite might just be Qualcomm's most aggressive move yet. With 64GB RAM support, an ultra-dense integrated design, and desktop-class cooling in test environments, the question now is whether the ecosystem — apps, drivers, and thermal envelopes — can keep up with the silicon. If it can? ARM might finally be ready to challenge x86 where it hurts. Qualcomm has another Snapdragon lawsuit in the works: Here's what you need to know. Make AI videos for free with OpenAI's Sora in Microsoft Bing The forces urging you to update to Windows 11 are growing — but are things as dire as they appear?
Yahoo
06-05-2025
- Yahoo
Gigabyte's GPU Thermal Paste Keeps Leaking
Several instances of Gigabyte graphics cards leaking thermal interface gel during operation have raised concerns over the quality control of its new graphics cards. Although Gigabyte claims the problem is merely cosmetic, the fact that these problems have arisen so soon after the cards launched isn't giving new owners much confidence in the design. All graphics cards use some form of thermal interface material (TIM) to sandwich between the big metal heatsink and individual chips and components on the card's PCB, including the GPU itself and memory modules. Traditionally that's been in the form of heat paste, but more exotic solutions are used today, including thermal pads, liquid metal, and in this instance, a "server grade" thermal gel. Where traditional concerns with TIM involved it drying out, this newer gel design appears to have a potential issue with its viscosity going in the other direction entirely. At the very least, something is causing it to slip out of its intended placement. There have now been three confirmed reports of Gigabyte RTX 50-series graphics cards leaking thermal gel from between the heatsink and various chips on the cards. It puddles around the chips, but as of yet hasn't caused any problems beyond merely looking unsavory, leading to Gigabyte's hand-waving comments about it being merely an aesthetic issue, as per TechPowerUp. But it also claims the issue only affects early production batches, suggesting that there is something about the design that has changed since then, fixing whatever the problem is. Gigabyte gel slipping out on a GPU. Credit: TechPowerUp Forum In each case of leaking gel, the cards were mounted vertically in the case, perhaps oriented in such a way that gravity contributed to the force driving out the gel in a certain direction. But Gigabyte doesn't warn about this in the official documentation for the cards, nor when these cards were first sold. That raises questions about what changed from early batches of the card, and if the change was made knowingly, why wasn't it communicated? Owners of the affected cards are understandably upset and concerned. Even if the issue is only cosmetic for now, surely leaking gel isn't as thermally effective, or will become less so over time. This raises concerns over long-term reliability, which shouldn't be something a new GPU owner has to worry about. Especially if they paid the inflated prices of Nvidia's new high-end cards, which in some cases can cost over $3,000. Gigabyte has issued a statement saying that it takes "your concerns seriously," but has made it clear that for now it won't be doing any kind of recall of potentially affected products. If you have a Gigabyte RTX 50-series graphics card, it might be worth avoiding vertical mounting. At least until this issue is cleared up or understood more.
Digital Trends
06-05-2025
- Digital Trends
I loved these speakers so much, I burned myself trying to fix them
Table of Contents Table of Contents Love of the ridiculous A glimmer of hope Maybe I can fix high-grade electronics myself… So long and thanks for all the riffs Sigh. Today's a legitimately sad day. I've had to say goodbye to my Harman Kardon GLA-55 2.0 PC speakers. I've had them for over 15 years of faithful service, but despite trying to have them repaired and my own botched attempts at fixing them, I've had to accept that their time has come and they've joined the paint tins for the next trip to the dump (for recycling). I have a new pair on the way, and they won't sound as good, nor look as iconic, but they will work properly, which is arguably more important. It's still a sad day though, and worth eulogizing. While technology tends to come and go, we all have that one gadget, peripheral, or component which has a special place in our hearts and for me, my ridiculous Harman Kardon speakers were one of them. Recommended Videos I literally just wrote a piece for PCWorld about why these speakers were so great, too. A fitting send off, but an ironic one. Let's give them one more. Love of the ridiculous The GLA-55s are a unique set of speakers that were gifted to me in the early 2010s by my then boss who had no need for them in his even-more-impressive audiophile setup. They're huge and heavy and completely ridiculous. They didn't ever fit into my setup or look right next to my more modern, and less-crystally monitors and other hardware. But they sounded great and were a brilliant conversation starter. Out of all the strange gadgets, gizmos, and components I've had over the last 20 years of writing about PCs and other hardware, my Harman Kardon speakers always drew the most interest from visiting friends and family. They were as much a work of art as capable PC speakers with strong bass and clear highs. TechPowerUp even called them the coolest speakers ever in its coverage of them way back when. But after so many years of use they had started to show their age. Scratches in the crystal-like plastic, dents in the soft-surface tweeter from errant kid fingers, and finally, a hole in the rubber surrounding the main drivers. While manoeuvring them for some shots for the aforementioned article I wrote about their longevity, I brushed my finger against the rubber and it just disintegrated into a sticky, powdery mess. At the lightest of touches I had inadvertently destroyed the airtight chamber around the driver which helped deliver such thumping good sound. In an instant of carelessness I'd ruined this speaker. But apparently this wasn't my first time. Looking closer at the other speaker revealed similar stress points on the rubber and in actuality, it had already started to come away from the plastic surround too. Neither speaker was in great shape and some quick Googling suggested that at best, I shouldn't be using them until they were fixed. If they could be fixed at all. A glimmer of hope My first contact with Harman Kardon was a dead end. These speakers had been end of life for years, so no, they couldn't fix them or provide any repair aid. They did have a third-party company they used for repairs in the UK, though. They might be able to do it, I was told. But no. The parts to replace this very specific speaker surround no longer existed, so there was no way they could do it either. Perhaps a third party alternative? I looked, but the drivers on these speakers were a very specific 3.1-inches in diameter. I couldn't find anything close to the right size… Until I did. It turns out that a Dutch speaker repair shop had a few foam replacement driver surrounds for my very specific Harman Kardon GLA-55 speakers. What luck! I ordered a pair to be shipped over in due haste and eagerly awaited their arrival two weeks later. In the meantime I contacted the original repair shop, but balked at the potential bill. It was $150 (equivalent) to even look at the speakers to see if they could or could not repair them. The estimates for the actual repair were several hundred more, plus shipping. While these speakers might have been worth $1,000 when new, but I could get a comparable set of speakers for the same price of a repair. But I'm an industrious chap. Before I confine these speakers to the trash heap I should take one more shot at it. Maybe I can fix high-grade electronics myself… I've been taking apart graphics cards, laptops, and other complicated PC hardware for years. Surely with due care I could have a good go at fixing these myself. So I set to work. I peeled off the underside rubber feet, I removed the screws. I took apart the base housing, and undid the screws on the circuit board and… solder. Unleaded solder at that, so my hobbyist soldering iron couldn't quite get it hot enough to melt it effectively. I didn't have a solder wick. I burned myself twice in my inexperience. Two hours into this mess of melted skin and stubbornly solid solder, I considered my options. I was on the cusp of diving down an (admittedly autistically driven) rabbit hole. There was still a chance I could fix this, but I was about to order a new soldering iron and some other kit to make this one repair easier, that I wasn't confident of performing effectively. I touched the rubber on the speaker surround again, as if to confirm how bad it was, and it melted away in my hand in a sticky mess that would have made fitting the foam even after an effective teardown difficult. This wasn't going to work, was it? I could spend hours of my time and tens or hundreds more of my dollars to fix these old, busted, dented, beautiful speakers. Or I could go and write an article about it and earn enough to pay for their replacements. So, here we are. So long and thanks for all the riffs Reluctantly, I spent the next 20 minutes tidying up the mess I'd made. The blobs of solder on the table, the seemingly-liquifying bits of speaker rubber on the floor, put the tools into the toolbox, my blistered finger under the tap. I thought about screwing the speakers back together so they could at least take their funeral march to the dump in full form. Like preparing them for an open casket. But they're speakers, not a person or pet. I stuck them next to the pile of other junk outside and ordered some perfectly fine Edifier R1280DBs. They're not super exciting, don't look weird and wonderful, and probably won't sound as good as the Harman Kardon's did back in their heyday. But there is Bluetooth support, and they'll fit better on my desk, and won't have a weird popping sound when I'm instead using my wireless headphones. I will miss the GLA-55s though. They weren't just a great piece of hardware, but an iconic part of my PC gaming and working setup for over a decade. It honestly feels a bit weird typing away like this without them flanking my monitors. Something is amiss. I suspect it'll feel that way for quite some time.
Yahoo
24-02-2025
- Yahoo
Blackwell launch gets even worse as Nvidia confirms some RTX 5090 and RTX 5070 Ti GPUs are underperforming in games – here's how to check if you're affected
When you buy through links on our articles, Future and its syndication partners may earn a commission. Nvidia has confirmed that approximately 1 in 200 of its RTX 5090 and RTX 5070 Ti GPUs have an issue with their graphics chips The problem is a loss of ROPs, a key element of the inner workings of the GPU Nvidia says those with an affected graphics card should contact the maker to arrange a replacement Following reports of some RTX 5090 GPUs failing to perform as well as they should in gaming, Nvidia has confirmed that there is an issue with the chips in the Blackwell flagship, as well as the newly arrived RTX 5070 Ti GPU. This is a hardware-level problem, meaning it's a fault deep in the chip which can't be fixed, and it's slowing down these graphics cards by an appreciable (albeit variable) amount. In a statement addressing the matter, Nvidia told The Verge: "We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified. "The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected." Your first question may well be: what's a ROP, then? ROP stands for Raster Operations Pipeline, and this is hardware that's a key part of the process of rendering the graphics for your PC games. (It's a lot more complicated than that, in reality, but that's all you really need to know). With fewer of those pipelines available to deal with the relevant graphics processing tasks during gameplay, unsurprisingly, performance is a bit slower. Also, if you're wondering about the mentioned RTX 5090D, that's the variant of the Blackwell flagship sold in China, which was involved in the initial reports of this issue – notably the RTX 5070 Ti wasn't, though. This whole episode unfolded yesterday, having first emerged courtesy of TechPowerUp's review of a Zotac RTX 5090 Solid graphics card (via VideoCardz). In its review, the tech site found that this third-party model was somehow underperforming versus an Nvidia RTX 5090 Founders Edition (the performance baseline used by TechPowerUp in gauging the relative power of the flagship GPU variants). Indeed, the Zotac RTX 5090 was around 5% slower than Nvidia's own model, while running at the same clock speeds, which obviously didn't make much sense. Not until TechPowerUp investigated and found this wasn't an issue pertaining to faulty cooling (or other likely-seeming root causes), but in fact that the Zotac GPU was missing ROPs. The RTX 5090 graphics card was showing 168 ROPs enabled (in the GPU-Z utility) rather than the expected count (and official spec) of 176 ROPs. All vendors are potentially affected by this gremlin in the works, of course, as this is an issue with the chips produced by Nvidia, and sent to third-party partners to be used in the manufacturing of their graphics cards. That was swiftly shown yesterday as reports started to come in, as folks started checking their boards for this issue. While in its statement, Nvidia mentions the lack of one ROP, it is referring to one block of them, so as observed, the ROP count is reduced by eight (the number in a block) with graphics cards that have this issue. To check your RTX 5090 or 5070 Ti, you can fire up a tool that peers deep into the innards of your hardware, monitoring and reporting back on multiple elements of the spec. Obviously what you're looking for is the ROPs count, and that can be provided by GPU-Z as already mentioned, or an alternative utility like HWiNFO (and probably other software out there, no doubt). In GPU-Z, you'll find the ROPs number listed in the Graphics Card tab, on the seventh line down, over on the left-hand side (we've got an explainer here, if you want further details on GPU-Z). For the RTX 5090, the number should be 176, whereas 168 is what the impaired flagship models are showing. With the RTX 5070 Ti, the correct spec is 96 ROPs, so in theory, it will be reduced to 88 ROPs (but I haven't seen confirmation of that yet, so perhaps it could have less of an impact). If you do have an RTX 5090 or 5070 Ti with this problem, how much will it affect you in practical terms? Well, that varies as I already mentioned, although as stated the average impact is a performance loss of something in the order of 5% (or thereabouts – Nvidia is saying 4%). However, you may not notice any difference at all in some cases, as one game may use the mentioned pipelines (ROPs) more heavily, while another may hardly touch them at all. So some games could be slowed down by more than 5%, and others may have a negligible loss in frame rates (such a low impact you'd never be able to tell). However, before you go thinking that maybe this isn't such a big deal after all, rest assured, it is. A fault like this should not have cleared quality assurance and made it into production hardware in the first place. And when you recall how much buyers have forked out for the RTX 5090 in particular – the MSRP is a true wallet-worrier, and many folks have overpaid beyond that – well, you can start to see how this is a big letdown. If you have an RTX 5090 or RTX 5070 Ti, check your graphics card in GPU-Z as outlined above. If your model is showing a loss of ROPs, as advised by Nvidia, contact your board manufacturer and begin the process of having the graphics card replaced. That could be a troublesome matter, though, for those who may have sold their old GPU when they upgraded (if they need to send the faulty Blackwell graphics card back, before receiving a new one – and are left with a gaming PC without an engine, essentially). The other worry is that it's not like fresh stock is going to be easy to come by, either, right now. Future GPUs shouldn't suffer from this issue because as Nvidia observes, the 'production anomaly' here has been fixed, as you would hope. Nvidia is investigating reports of crashes plaguing RTX 5090 and 5080 GPUs, with possible driver issues maybe hitting RTX 4000 models too Finally, some good news about Nvidia's new GPUs: RTX 5090 stock levels rumored to surge in a month or so Where to buy an Nvidia RTX 5070 Ti - your best bets for getting the upper-mid-range GPU
