Latest news with #OpenCL
Tom's Guide
28-07-2025
- Tom's Guide
Leaked Nvidia N1X CPU benchmark hints at RTX 5070 power — with 20 CPU cores
Rumors are heating up around Nvidia's N1X Arm-based CPU, and thanks to a leaked benchmark, we may have an idea of its integrated GPU capabilities — and it's hinting at RTX 50-series GPU power. A Geekbench benchmark listing for the Nvidia N1X CPU has popped up, showing its OpenCL score, which measures GPU performance. Its results show a 46,361 OpenCL score, which puts it around an RTX 2050 GPU. Of course, this isn't ideal, but this early on, it is an early prototype of the chip, so don't expect this result in the final rumored chip. More impressively, the Geekbench entry shows the specs of the N1X SoC, including a 20-core CPU that's split into two 10-core clusters, and 48 compute units. These are the Streaming Multiprocessors, and Nvidia offers 128 CUDA cores per unit. This translates to 6,144 CUDA cores, which is the same number as an RTX 5070 desktop GPU. We've heard that the N1X may deliver RTX 4070 power in a previous benchmark leak, but seeing this key spec shows what Nvidia may be aiming to deliver in its rumored Arm-based chip. What's more, it shows similar specs to the Nvidia GB10 Superchip, seen in DGX Spark-equipped PCs. As rumored, the N1X may use a pared-down version of the GB10, and as that chip combines a Blackwell GPU with a Grace Arm CPU to make a custom SoC, we can expect this chip to offer RTX 50-series performance. Now, despite the weaker OpenCL result, it still pushes integrated GPU performance in Arm-based chips. Even though it's likely not to be the final results, the N1X chip in this benchmark still pushes past Apple's M3 Max performance, which usually sees a score of 37,500 (as per our sister site, Tom's Hardware). All in all, with these expected specs now known, it's looking like Nvidia will have a strong chip on its hands, putting Apple's own M-series chips on notice. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. Of course, we won't know what Nvidia's rumored N1X chip delivers until it's officially announced, and it may be a while until we see it. Recently, the Nvidia N1X CPU was reportedly delayed due to a number of factors, with one of them having to do with the launch of Microsoft's next-gen operating system (a possible hint at Windows 12). Along with apparently being hit with problems that require engineers to make design changes to the silicon, it's now expected to arrive in 2026. Many believe it could be around early 2026, which points to a big announcement at CES 2026, but it could come later, with some reports stating it could arrive in late 2026 instead. Nvidia's N1X and N1 CPUs are expected to power next-gen desktops and laptops, with the latter bringing more ultraportable gaming laptops without the need for a discrete GPU. It's been tipped that Dell's Alienware gaming brand will be the first to receive these chips. Only time will tell what Nvidia's N1X-series CPUs will deliver, but if we're getting performance akin to an RTX 5070, then these chips should be worth waiting for. Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.
Arabian Post
14-07-2025
- Arabian Post
Open‑source digiKam add‑on sharpens AI handling in 8.7 release
digiKam's 8.7.0 update foregrounds artificial intelligence enhancements, integrating tools that tackle everyday pain points in photo management. Key features include bolstered facial recognition workflows, hardware‑accelerated processing, and a novel AI‑driven image auto‑rotation utility — moves that position digiKam as a free contender in the AI photo management space. Face recognition has undergone targeted refinement. The system now initiates automatic scans when a user confirms or tags a face, streamlining management through proactive detection. In cases of misidentification, rejected suggestions are retained by the software, enabling it to propose the next-best match — an increment in precision notably absent in prior versions. Under the surface, digiKam continues to leverage its cross‑validating KNN and SVM classifier ensemble, originally introduced in version 8.6.0, which now exhibits further acceleration and reduced false positives. The adoption of GPU acceleration marks another stride. Users can now verify compatibility and performance via a built‑in OpenCL/CUDA test in Settings, ensuring that pipelines such as face detection, resizing, and colour conversion tap into GPU performance where available. The update supports OpenCV's OpenCL layer and NVIDIA's cuDNN via CUDA. A toggle allows disabling GPU use if driver hiccups are suspected. Performance gains from GPU utilisation may significantly reduce processing times in extensive photo batches. ADVERTISEMENT Addressing a common hassle for event photography or mixed portraiture outputs, the new AI Auto‑Rotation tool analyses content and applies the correct orientation using deep‑learning techniques. Unlike earlier manual methods that rotated entire batches uniformly, this plugin scrutinises each frame for visual cues, autonomously rotating images with varied orientations — a major boon for photo collections with inconsistent metadata. Beyond AI, the release refines foundational components. The RAW engine Libraw is updated to the May 12, 2025 snapshot, extending compatibility to more than 1,260 camera models. ExifTool is upgraded to version 13.29 for more reliable metadata handling. The Qt framework jumps to 6.8.3 on Windows/Linux and 6.9.0 on macOS. The G'MIC‑Qt plug‑in, now at 3.5.0, gains a new 'Montage' layer‑mode filter via G'MIC Generic, improving stack editing and creative use cases. Stability improvements feature strongly: more than 240 bug fixes resolve issues in face workflow operations, memory leaks, installation across platforms, and database stability. The internationalisation effort reaches 61 interface languages, with documentation localised into 16 languages. Looking ahead, the developers preview ambitions beyond current enhancements. The October 2025 update may introduce AI‑assisted noise reduction and colour correction. Plans are also underway to implement natural‑language querying via large‑language models, facilitating intuitive database searches like 'sunset beach photos'. OpenCovering both fundamentals and future innovation, digiKam 8.7.0 builds solidity through stability and sharpens its AI edge without locking features behind a paywall. By delivering intelligent face management, GPU efficiency, and autonomous photo-orientation, it caters to enthusiast photographers and professionals alike, while maintaining zero-cost access across Linux, Windows, and macOS platforms. The update is available now for direct download, with cross‑platform packaging via AppImage, installers, and bundles.
