VeriSilicon's Scalable High-Performance GPGPU-AI Computing IPs Empower Automotive and Edge Server AI Solutions
VeriSilicon's GPGPU-AI computing processors are architected to tightly integrate GPGPU computing with AI accelerator at fine-grained levels.
VeriSilicon's GPGPU-AI computing IPs are based on a high-performance General Purpose Graphics Processing Unit (GPGPU) architecture with an integrated dedicated AI accelerator, delivering outstanding computing capabilities to AI applications. The programmable AI accelerator and sparsity-aware computing engine accelerate transformer-based and matrix-intensive models through advanced scheduling techniques. These IPs also support a broad range of data formats for mixed-precision computing, including INT4/8, FP4/8, BF16, FP16/32/64, and TF32, and are designed with high-bandwidth interfaces of 3D-stacked memory, LPDDR5X, HBM, as well as PCIe Gen5/Gen6 and CXL. They are also capable of multi-chip and multi-card scale-out expansion, offering system-level scalability for large-scale AI application deployments.
VeriSilicon's GPGPU-AI computing IPs provide native support for popular AI frameworks for both training and inference, such as PyTorch, TensorFlow, ONNX, and TVM. These IPs also support General Purpose Computing Language (GPCL) which is compatible with mainstream GPGPU programming languages, and widely used compilers. These capabilities are well aligned with the computing and scalability requirements of today's leading LLMs, including models such as DeepSeek.
'The demand for AI computing on edge servers, both for inference and incremental training, is growing exponentially. This surge requires not only high efficiency but also strong programmability. VeriSilicon's GPGPU-AI computing processors are architected to tightly integrate GPGPU computing with AI accelerator at fine-grained levels. The advantages of this architecture have already been validated in multiple high-performance AI computing systems,' said Weijin Dai, Chief Strategy Officer, Executive Vice President, and General Manager of the IP Division at VeriSilicon. 'The recent breakthroughs from DeepSeek further amplify the need for maximized AI computing efficiency to address increasingly demanding workloads. Our latest GPGPU-AI computing IPs have been enhanced to efficiently support Mixture-of-Experts (MoE) models and optimize inter-core communication. Through close collaboration with multiple leading AI computing customers, we have extended our architecture to fully leverage the abundant bandwidth offered by 3D-stacked memory technologies. VeriSilicon continues to work hand-in-hand with ecosystem partners to drive real-world mass adoption of these advanced capabilities.'
VeriSilicon is committed to providing customers with platform-based, all-around, one-stop custom silicon services and semiconductor IP licensing services leveraging its in-house semiconductor IP. For more information, please visit:
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
23 minutes ago
- Yahoo
Nvidia Earnings Seen as Next Big AI Spark
This article first appeared on GuruFocus. Nvidia (NASDAQ:NVDA) is back in the spotlight with earnings coming up on August 27, and Wedbush thinks the report could be the next big spark for the stock. Analysts there say the AI story is still in the second inning of a nine-inning game, with plenty of growth left as more companies and governments pile into the spending boom. Warning! GuruFocus has detected 7 Warning Signs with AMD. The Street expects Nvidia to deliver $1.01 a share on $45.9 billion in revenue. That would underscore just how much demand has surged for its chips as Big Tech pours nearly $350 billion into AI capex this year. Wedbush argues the cavalry is still coming, with enterprises worldwide only starting to step up their AI budgets. Even after Nvidia, Palantir (NASDAQ:PLTR), and AMD (NASDAQ:AMD) sold off Tuesday in a round of profit-taking, Wedbush called the dip a buying chance. They see trillions of dollars in AI infrastructure, software, and power spend fueling the tech bull cycle for at least another two to three years volatility or not. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
Yahoo
23 minutes ago
- Yahoo
Coinbase Just Flipped the Script--Is This the Real Killer App for Crypto?
This article first appeared on GuruFocus. Coinbase (NASDAQ:COIN) is making its boldest bet yet: that stablecoinsthose tame, dollar-pegged cousins in crypto's otherwise wild Westcould become the lifeblood of the AI economy. At its recent Crypto x AI hackathon in Brooklyn, the company gave a glimpse into this future. About 100 developers, some flying in from as far as Italy and Malaysia, stayed up through the sweltering August weekend coding AI tools and bots that transact using digital dollars. At the center of it all was x402, Coinbase's new open-source protocol designed to let machines settle in stablecoins over the internetinstantly, without banks, and for pennies. Warning! GuruFocus has detected 7 Warning Signs with COIN. This isn't a marketing gimmick. For Coinbase, it's a strategic pivot away from the highs and lows of trading revenue, and into infrastructure with real-world use cases. With the U.S. now having passed its first federal stablecoin law, the regulatory fog is beginning to liftand the race is on. Stripe recently acquired Bridge and is reportedly building its own blockchain. PayPal launched its own stablecoin, leaning into distribution. But Coinbase's approach focuses on developers. It wants to be the go-to toolkitthe AWS of programmable payments. According to Nemil Dalal, who leads Coinbase's Developer Platform, this is the logical next step in co-founder Brian Armstrong's decade-old roadmap: speculation first, infrastructure second, real-world utility third. Of course, this isn't guaranteed to scale overnight. Visa and Mastercard remain formidable gatekeepers with massive reach. AI's commercial path is still taking shape. And stablecoins, while booming in countries battling inflation, remain a niche payment method in the U.S. Still, for a weekend in Brooklyn, the momentum looked real. Hackathon teams built bots that could charge per query, marketplaces that didn't need banks, and tools that paid creators instantly. The message was hard to miss: crypto might finally be shedding its casino image, and Coinbase is positioning itself to power what comes next.
Yahoo
23 minutes ago
- Yahoo
Google's Pixel 10 phones raises the ante on artificial intelligence
Google on Wednesday unveiled a new line-up of Pixel smartphones injected with another dose of artificial intelligence that's designed to do everything from fetch vital information stored on the devices to help improve photos as they're being taken. The AI expansion on the four Pixel 10 models amplifies Google's efforts to broaden the use of a technology that is already starting to reshape society. At the same time, Google is taking a swipe at Apple's Achilles' heel on the iPhone. Apple so far has only been able to introduce a few basic AI features on the iPhone while failing to deliver on last year's promise to deliver a more conversational and versatile version of its often-blundering virtual assistant Siri. Without mentioning the iPhone by name, Google has already been mocking Apple's missteps in online ads promoting the four new Pixel models as smartphones loaded with AI technology that consumers won't have to wait for more than a year to arrive. 'There has been a lot of hype about this and, frankly, a lot of broken promises, too,' Google executive Rick Osterloh said during a 75-minute presentation in New York about the new Pixel phones. The event was emceed by late-night TV show host Jimmy Fallon. Google, in contrast, has been steadily increasing the amount of AI that it began to implant on its Pixels since 2023, with this year's models taking it to another level. 'We think this yeasr we have a game-changing phone with game-changing technology,' Osterloh said. Taking advantage of a more advanced processor, Google is introducing a new AI feature on the Pixel 10 phones called 'Magic Cue' that's designed to serve as a digital mind reader that automatically fetches information stored on the devices and displays the data at the time it's needed. For instance, if a Pixel 10 user is calling up an airline, Magic Cue is supposed to instantaneously recognize the phone number and display the flight information if it's in Gmail or a Google Calendar. The Pixel 10 phones will also come with a preview feature of a new AI tool called 'Camera Coach' that will automatically suggest the best framing and lighting angle as the lens is being aimed at a subject. Camera Coach will also recommend the best lens mode to use for an optimal picture. The premium models — Pixel 10 Pro and Pixel 10 Pro XL — will also include a 'Super Res' option that deploys a grab bag of software and AI tricks to zoom up to 100 times the resolution to capture the details of objects located miles away from the camera. The AI wizardry could happen without users even realizing it's happening, making it even more difficult to know whether an image captured in a photo reflects how things really looked at the time a picture was taken or was modified by technology. The Pixel 10 will also be able to almost instantaneously translate phone conversations into a range of different languages using the participants own voices. Google is also offering a free one-year subscription to its AI Pro plan to anyone who buys the more expensive Pixel 10 Pro or Pixel 10 Pro XL models in hopes of hooking more people on the Gemini toolkit it has assembled to compete against OpenAI's ChatGPT. The prices on all four Pixel 10 models will remain unchanged from last year's Pixel 9 generation, with the basic starting at $800 and the Pro selling for $1,000, the Pro XL at $1,200 and a foldable version at $1,800. All the Pixel 10s expect the foldable model will be in stores on August 28. The Pixel 10 Pro Fold will be available starting October 9. Although the Pixel smartphone remains a Lilliputian next to the Gulliverian stature of the iPhone and Samsung's Galaxy models, Google's ongoing advances in AI while holding the line on its marquee devices raise the competitive stakes. 'In the age of AI, it is a true laboratory of innovation,' Forrester Research analyst Thomas Husson said of the Pixel. Apple, in particular, will be facing more pressure than usual when it introduces the next-generation iPhone next month. Although the company has already said the smarter Siri won't be ready until next year at the earliest, Apple will still be expected to show some progress in AI to demonstrate the iPhone is adapting to technology's AI evolution rather than tilting toward gradual obsolescence. Clinging to a once-successful formula eventually sank the BlackBerry and its physical keyboard when the iPhone and its touch screen came along nearly 20 years ago. Apple's pricing of the next iPhone will also be under the spotlight, given that the devices are made in China and India — two of the prime targets in President Donald Trump's trade war. But Apple appeared to gain a reprieve from Trump's most onerous threats earlier this month by adding another $100 billion on top of an earlier $500 billion investment pledge to the U.S. The tariff relief may enable Apple to minimize or even avoid price increases for the iPhone, just as Google has done with the Pixel 10 models. Michael Liedtke, The Associated Press
