Tag: NVIDIA

Meta sheds more light on how it is evolving Llama 3 training — it relies for now on almost 50,000 Nvidia H100 GPU, but how long before Meta switches to its own AI chip?

[ad_1]

Meta has unveiled details about its AI training infrastructure, revealing that it currently relies on almost 50,000 Nvidia H100 GPUs to train its open source Llama 3 LLM.

The company says it will have over 350,000 Nvidia H100 GPUs in service by the end of 2024, and the computing power equivalent to nearly 600,000 H100s when combined with hardware from other sources.

The figures were revealed as Meta shared details on its 24,576-GPU data center scale clusters.

The company explained “These clusters support our current and next generation AI models, including Llama 3, the successor to Llama 2, our publicly released LLM, as well as AI research and development across GenAI and other areas.“

The clusters are built on Grand Teton (named after the National Park in Wyoming), an in-house-designed, open GPU hardware platform. Grand Teton integrates power, control, compute, and fabric interfaces into a single chassis for better overall performance and scalability.

The clusters also feature high-performance network fabrics, enabling them to support larger and more complex models than before. Meta says one cluster uses a remote direct memory access network fabric solution based on the Arista 7800, while the other features an NVIDIA Quantum2 InfiniBand fabric. Both solutions interconnect 400 Gbps endpoints.

“The efficiency of the high-performance network fabrics within these clusters, some of the key storage decisions, combined with the 24,576 NVIDIA Tensor Core H100 GPUs in each, allow both cluster versions to support models larger and more complex than that could be supported in the RSC and pave the way for advancements in GenAI product development and AI research,” Meta said.

Storage is another critical aspect of AI training, and Meta has developed a Linux Filesystem in Userspace backed by a version of its ‘Tectonic’ distributed storage solution optimized for Flash media. This solution reportedly enables thousands of GPUs to save and load checkpoints in a synchronized fashion, in addition to “providing a flexible and high-throughput exabyte scale storage required for data loading”.

While the company’s current AI infrastructure relies heavily on Nvidia GPUs, it’s unclear how long this will continue. As Meta continues to evolve its AI capabilities, it will inevitably focus on developing and producing more of its own hardware. Meta has already announced plans to use its own AI chips, called Artemis, in servers this year, and the company previously revealed it was getting ready to manufacture custom RISC-V silicon.

AMD’s FSR 3.1 promises improved image quality – with less flickering, better motion – and a trick that’ll delight some Nvidia GPU owners

[ad_1]

AMD has announced its next-gen upscaling box of tricks to make frame rates smoother with PC games at GDC 2024.

This is FSR 3.1, and it sounds like a big step forward – and not just for the quality of the upscaling tech, since there’s a surprise in tow for Nvidia (and Intel Arc) GPU owners.

The basics outlined by AMD are that in terms of better image quality, FSR 3.1 will boast ‘improved temporal stability’ which means fewer flickering or shimmering effects and less ‘fizziness’ around objects that are moving. There’ll be a reduction in the level of ghosting seen, too.

FSR 3.1 will provide a cleaner image overall, in short, and we’ll also get better preservation of detail, AMD asserts.

What about the surprise we mentioned? This is “decoupling FSR 3 upscaling from frame generation” so they are separate entities, and as AMD makes clear, this means that frame generation (adding additional artificial frames for a smoother frame rate) could work with other upscaling solutions.

AMD tells us that the first game to use FSR 3.1 will be Insomniac’s Ratchet & Clank: Rift Apart, and it’ll be delivered in an update for the game later this year (sadly we don’t get any idea of a precise timeframe).

FSR 3.1 should be available to game developers from Q2, meaning next month.

Support for Vulkan and the Xbox Game Development Kit is also inbound, and AMD notes a new FidelityFX API will make it “easier for developers to debug” and provide forward compatibility with updated versions of FSR.

Analysis: DLSS turbocharger for older Nvidia GPUs

Better image quality is clearly welcome with FSR 3.1 – even if some feel AMD has been pretty slow to act implementing improvements on this front – but the decoupling of frame generation is equally important.

This means that a PC gamer using Nvidia’s DLSS upscaling could draft in AMD’s frame generation to further rev up frame rates. Okay, so Lovelace GPU owners get DLSS 3 frame generation anyway, but those who don’t have an RTX 4000 graphics card can use DLSS but layer AMD’s frame generation on top, at least in theory. (Remember, while DLSS 3.5 may support older Nvidia GPUs with ray reconstruction, Team Green’s frame generation is still only available to RTX 4000 graphics cards).

Or those with an Intel Arc graphics card making use of XeSS upscaling could do the same thing, of course. AMD has tried to stay as open as possible with FSR and this is another commendable move in that respect.

We’ll need to see all this in action before we can really draw conclusions about how far AMD has advanced with version 3.1 of FSR, and it might be a bit of a wait yet before that happens.

Via VideoCardz

[ad_2]

Source Article Link

Tags AMDs, delight, flickering, FSR, GPU, image, Improved, motion, NVIDIA, Owners, promises, Quality, Thatll, trick

Featured

Nvidia CEO says don’t give up learning new skills — just maybe leave programming to AI

Post author By lisa nichols
Post date March 21, 2024
No Comments on Nvidia CEO says don’t give up learning new skills — just maybe leave programming to AI

[ad_1]

Nvidia CEO Jensen Huang has clarified comments he made about the supposed “death of coding”.

Huang had been criticized in the past for saying on several occasions that as AI platforms would soon be doing a lot of the heavy lifting when it comes to coding, young people today should not necessarily consider learning it as a vital skill.

Speaking at the company’s Nvidia GTC 2024 event in San Jose, Huang was asked at a press Q&A if he still believed this was the case – and it seems not much has changed.

Death of coding?

“I think that people ought to learn all kinds of skills,” Huang said, comparing learning to code to skills such as juggling, playing piano or learning calculus.

However, he did add that, “programming is not going to be essential for you to be a successful person…but if somebody wants to learn to do so (program), please do – because we’re hiring programmers.”

In the past, Huang had said that time otherwise spent learning to code should instead be invested in expertise in industries such as farming, biology, manufacturing and education, and that upskilling could be a key way forward, helping provide the knowledge of how and when to use AI programming.

Huang did also add that generative AI would require a number of new skills in order to close the technology divide.

“You don’t have to be a C++ programmer to be successful,” he said. “You just have to be a prompt engineer. And who can’t be a prompt engineer? When my wife talks to me, she’s prompt engineering me.”

“We all need to learn how to prompt AIs, but that’s no different than learning how to prompt teammates.”

These skills could be vital for younger people entering the workforce at an auspicious time, Huang went on to add.

“It (AI) Is a new industry – that’s why we say there’s a new industrial revolution,” he declared, In the future, almost all of our computing will be generated.”

NVIDIA partners with Ubisoft to further develop its AI-driven NPCs

Post author By lisa nichols
Post date March 19, 2024
No Comments on NVIDIA partners with Ubisoft to further develop its AI-driven NPCs

[ad_1]

NVIDIA has been working on adding generative AI to non-playable characters (NPCs) for a while now. The company is hoping a newly-announced partnership with Ubisoft will accelerate development of this technology and, ultimately, bring these AI-driven NPCs to modern games.

Ubisoft helped build new “NEO NPCs” by using NVIDIA’s Avatar Cloud Engine (ACE) technology, with an assist from . The end result? Characters that don’t repeat the same phrase over and over, while ignoring the .

These NEO NPCs are said to interact in real time with players, the environment and other in-game characters. NVIDIA says this opens up “new possibilities for emergent storytelling.” To that end, Ubisoft’s narrative team built complete backgrounds, knowledge bases and conversational styles for two NPCs as a proof of concept.

A lady NPC talking about a bank heist or something. — Ubisoft/NVIDIA

These are fully-fleshed out characters with “environmental and contextual awareness” along with conversational memory. NVIDIA says each character excels with “collaboration and strategic decision-making.” The company also incorporated its Audio2Face technology to allow for appropriate facial animations and lip syncing. NVIDIA suggests that these characters push “the boundaries of game design and immersion.”

The company didn’t explicitly announce that this technology would be coming to Ubisoft titles in the nearish future, but it’s a safe bet it’s on the roadmap. Perhaps soon you’ll be able to see what the sweet old man outside of the inn thinks about you absolutely obliterating the whole town. Won’t that be fun?

This isn’t the only big AI-adjacent announcement from NVIDIA this week. The company just , an evolution of the chips that are used to train large language models. The company promises that its Blackwell chips are seven to 30 times faster than the H100 series and use 25 times less power.

This article contains affiliate links; if you click such a link and make a purchase, we may earn a commission.

[ad_2]

Source Article Link

Tags AIdriven, develop, NPCs, NVIDIA, partners, Ubisoft

Business Industry

MediaTek partners with Nvidia to make car chips to rival Qualcomm, Samsung

Post author By miranda cosgrove
Post date March 19, 2024
No Comments on MediaTek partners with Nvidia to make car chips to rival Qualcomm, Samsung

[ad_1]

MediaTek has announced that it has partnered with Nvidia to make four new automotive chips for connected and self-driving cards. The company’s Dimensity Auto Cockpit series chips will compete with automotive chips from Qualcomm and Samsung.

MediaTek Dimensity Auto Cockpit chips to compete with Samsung’s Exynos Auto chips

The company’s new Dimensity Auto Cockpot series has four chips: CM-1, CV-1, CX-1, and CY-1. These 3nm chips use ARM’s V9-A CPU cores and Nvidia’s RTX GPU. They support Nvidia’s DRIVE OS and will power the infotainment system and other intelligence inside connected cars. Nvidia’s technology powers AI processing on these chips, and it will be used for entertainment, navigation, and general information needs.

The GPU can drive four high-resolution screens. In fact, MediaTek and Nvidia claim that the new chips can even support in-vehicle gaming, complete with ray-traced graphics. Media streaming and audio/video calls with AI-enhanced clarity are also supported. They can even monitor people inside the car for natural controls and gaze-aware UI.

The CV-1 is a chip for entry-level cars. The CM-1 is for mid-range cars and the CV-1 is for high-end cars. The CX-1 is for premium car offerings. These chips can simultaneously run multiple operating systems, including Android Auto, Linux, and QNX, in virtualization.

The new chips also support up to ten cameras and other sensors for assisted driving. The built-in ISP can process HDR scenes in real-time. Their connectivity features include a built-in 5G modem (sub-6GHz), NTN (for direct satellite connectivity for emergency calls and messages), Wi-Fi 7, GNSS, and Bluetooth.

[ad_2]

Source Article Link

Tags Car, chips, MediaTek, NVIDIA, partners, Qualcomm, Rival, Samsung

Entertainment

NVIDIA says its Blackwell GPUs are the world’s most powerful chips

Post author By lisa nichols
Post date March 19, 2024
No Comments on NVIDIA says its Blackwell GPUs are the world’s most powerful chips

[ad_1]

NVIDIA’s H100 chips are used by nearly every AI company in the world to train large language models hooked into services like ChatGPT. It’s been great for business. Now, the company is ready to make those chips look terrible, announcing a next-generation platform called Blackwell.

Named for David Harold Blackwell, a mathematician who specialized in game theory and statistics, NVIDIA claims Blackwell is the world’s most powerful chip, reaching speeds of 20 petaflops compared to just 4 petaflops the H100 provided. Yeah, throw it in the trash. You need new chips.

And if you didn’t know how powerful NVIDIA is, its press release for this new platform includes quotes from the CEOs of OpenAI, Microsoft, Alphabet, Meta and Tesla — yes, all CEOs you probably know the names of.

— Mat Smith

The biggest stories you might have missed

You can get these reports delivered daily direct to your inbox. Subscribe right here!

The tournament is postponed until further notice.

Yeah, this is bad. Respawn, the EA-owned studio behind Apex Legends, has postponed the North American Finals tournament after hackers broke into matches and equipped players with cheats. Footage of the hacks on Twitch show players being able to see their opponent’s location through walls, while notable player (and one of the best) ImperialHal was gifted an aimbot to hit enemies more easily. Respawn says it would share more information soon, but as of time of writing, the studio hasn’t elaborated.

Continue reading.

The Mevo Core has improved built-in mics and works with any MFT lens.

Logitech is expanding its Mevo lineup of livestreaming cameras. The company’s new Mevo Core shoots in 4K, a big upgrade from the 1080p Mevo Start camera kit I tested a few years back. However, the trade-off is pricing as the new model will set you back three times as much for a three-camera setup. $999. So yes, this is probably for the pro streamers.

To emphasize that, the Core ships as a body only, but Logitech will sell lens bundle kits through Amazon and B&H Photo Video. You will need to buy an additional lens just to make it work. And it’s only compatible with — so there’s a high chance you’ll have to buy one.

Continue reading.

It’s like Google search on Safari all over again. Plus 15 years.

Apple is reportedly in talks with Google to integrate its Gemini AI in iPhones, according to Bloomberg. Gemini could be the cloud-based generative AI engine for Siri and other iPhone apps, while Apple’s models could be woven into the upcoming iOS 18 for on-device AI tasks.

There are regulatory concerns to consider—the Department of Justice has already sued Google over its search dominance, including the way it pays Apple and other companies to use its search engine. But given how Microsoft and OpenAI’s partnership turned the Bing search engine into something people were actually talking about, the team-up might be worth the risk.

Continue reading.

[ad_2]

Source Article Link

Tags Blackwell, chips, GPUs, NVIDIA, Powerful, Worlds

Featured

“The world’s most powerful chip” — Nvidia says its new Blackwell is set to power the next generation of AI

Post author By lisa nichols
Post date March 19, 2024
No Comments on “The world’s most powerful chip” — Nvidia says its new Blackwell is set to power the next generation of AI

[ad_1]

The next generation of AI will be powered by Nvidia hardware, the company has declared after it revealed its next generation of GPUs.

Company CEO Jensen Huang took the wraps off the new Blackwell chips at Nvidia GTC 2024 today, promising a major step forward in terms of AI power and efficiency.

The first Blackwell “superchip”, the GB200, is set to ship later this year, with the ability to scale up from a single rack all the way to an entire data center, as Nvidia looks to push on with its leadership in the AI race.

Nvidia Blackwell

Representing a significant step forward for the company’s hardware from its predecessor, Hopper, Huang noted that Blackwell contains 208 billion transistors (up from 80 billion in Hopper) across its two GPU dies, which are connected by 10 TB/second chip-to-chip link into a single, unified GPU.

This makes Blackwell up to 30x faster than Hopper when it comes to AI inference tasks, offering up to 20 petaflops of FP4 power, far ahead of anything else on the market today.

Nvidia GTC 2024

(Image credit: Future / Mike Moore)

During his keynote, Huang highlighted not only the huge jump in power between Blackwell and Hopper – but also the major difference in size.

“Blackwell’s not a chip, it’s the name of a platform,” Huang said. “Hopper is fantastic, but we need bigger GPUs.”

Despite this, Nvidia says Blackwell can reduce cost and energy consumption by up to 25x, giving the example of training a 1.8 trillion parameter model – which would previously have taken 8,000 Hopper GPUs and 15 megawatts of power – but can now be done by just 2,000 Blackwell GPUs consuming just four megawatts.

The new GB200 brings together two Nvidia B200 Tensor Core GPUs and a Grace CPU to create what the company simply calls, “a massive superchip” able to drive forward AI development, providing 7x the performance and four times the training speed of an H10O-powered system.

The company also revealed a next-gen NVLink network switch chip with 50 billion transistors, which will mean 576 GPUs are able to talk to each other, creating 1.8 terabytes per second of bidirectional bandwidth.

Nvidia has already signed up a host of major partners to build Blackwell-powered systems, with AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure already on board alongside a host of big industry names.

Nvidia has virtually recreated the entire planet — and now it wants to use its digital twin to crack weather forecasting for good

[ad_1]

Faster and more accurate weather forecasts are about to become a real possibility across the globe thanks to a new release from Nvidia.

The computing giant has announced a new digital twin cloud platform that it says will help meteorologists and weather experts create richer and more detailed simulations and more.

The new Earth-2 APIs, unveiled at Nvidia GTC 2024, can be utilized to help address the $140 billion cost in losses around the world due to extreme weather brought on by climate change, with Nvidia saying work can now begin on an “unprecedented scale”.

Whatever the weather

“Climate disasters are now normal — historic droughts, catastrophic hurricanes and generational floods appear in the news with alarming frequency,” said Jensen Huang, founder and CEO of NVIDIA. “Earth-2 cloud APIs strive to help us better prepare for — and inspire us to act to moderate — extreme weather.”

Nvidia says the new models are set to be used by governments and organizations across the world, including the Taiwan Central Weather Administration, which aims to utilize better detection of typhoon landfall, with earlier predictions of such incidents meaning citizens can be evacuated quicker.

The Earth-2 cloud APIs will run on Nvidia DGX Cloud, opening them up to all kinds of users to create high-resolution simlations.

They use a new Nvidia generative AI model called CorrDiff which is able to generate 12.5 times higher resolution images than current numerical models 1,000x faster and 3,000x more energy efficiently, as well as correcting inaccuracies from previous models, bringing together multiple sources of information to create much more accurate and focused forecasts.

Nvidia is taking the Apple Vision Pro to the Omniverse

Post author By lisa nichols
Post date March 18, 2024
No Comments on Nvidia is taking the Apple Vision Pro to the Omniverse

[ad_1]

The Apple Vision Pro could soon be a lot more helpful for businesses around the world thanks to an upgrade from Nvidia.

At its Nvidia GTC 2024 event today, the computing giant announced the launch of a new software framework, built on its own Omniverse Cloud APIs, that will make Apple‘s mixed reality device much more intuitive for industry and business use.

The new Omniverse Cloud APIs will integrate with Nvidia’s own Graphics Deliver Network (GDN), greatly streamlining the process of creating and generating 3D experiences to be visible on an Apple Vision Pro device.

Nvidia Omniverse

Nvidia also showed off a demo of how the technology will work, via a fully accurate and interactive digital twin of a car, streamed directly to an Apple Vision Pro device. This digital twin was shown to be particularly useful for design and customization of the vehicle via previews of the color and trim, right down to a virtually sitting within the finished vehicle before it enters production.

Users can render fully interactive experiences via a single application available from Apple’s native SwiftUI and Reality Kit, with the Omniverse RTX Renderer streaming from GDN.

The company adds that because the platform is based on its Nvidia RTX Cloud offering, such renderings, edits and more can be seamlessly shared to an Apple Vision Pro device, with updates showing as they happen.

“The breakthrough ultra-high-resolution displays of Apple Vision Pro, combined with photorealistic rendering of OpenUSD content streamed from NVIDIA accelerated computing, unlocks an incredible opportunity for the advancement of immersive experiences,” added Mike Rockwell, vice president of the Vision Products Group at Apple.

“Spatial computing will redefine how designers and developers build captivating digital content, driving a new era of creativity and engagement.”

Nvidia says such demos show off the huge potential of spatial computing for businesses across multiple industries, offering boosted customization but also cutting down on the need for real-world prototypes.

Designers could also use such APIs to build applications that allow them to see their 3D data like never before, creating more accurate and robust simulations that allow them to push the limits of what is possible. Factory operators could also build richer digital twins to spot issues before they happen, boosting productivity, and designers can offer unbridled customization and creativity.

“Apple Vision Pro is the first untethered device which allows for enterprise customers to realize their work without compromise,” added Rev Lebaredian, vice president of simulation at Nvidia. “We look forward to our customers having access to these amazing tools.”

Nvidia GTC 2024 day one — all the news and announcements as it happens

Post author By lisa nichols
Post date March 18, 2024
No Comments on Nvidia GTC 2024 day one — all the news and announcements as it happens

[ad_1]

Refresh