With Nvidia Ace taking up 1 GB of VRAM in Inzoi, Team Green will need to up its memory game if AI NPCs take off in PC gaming
At this year's GDC event, Nvidia showcased all its latest RTX technologies, including Ace, its software suite of 'digital human technologies.' One of the first games to use it is Inzoi, a Sims-like game, and while chatting to Nvidia about it, I learned that the AI model takes up a surprising amount of VRAM, which raises an interesting question about how much memory future GeForce cards are going to have.
The implementation of Nvidia Ace in Inzoi (or inZOI, to use the correct title) is relatively low-key. The family you control, along with background NPCs, all display 'thought bubbles' which give you clues as to how they're feeling, what their plans are, and what they're considering doing in the future. Enabling Nvidia's AI system gives you a bit more control over their thoughts, as well as making them adapt and respond to changes around them more realistically.
In terms of technical details, the AI system used is a Mistral NeMo Minitron small language model, with 500 million parameters. The developers had experimented with larger models but settled on this size, as it gave the best balance between responsiveness, accuracy, and most importantly of all, performance. Larger models use more GPU resources to process and in this specific case, Inzoi uses 1 GB of VRAM to store the model.
That may not seem like very much, but this is a small model with some clear limitations. For example, it doesn't get applied to every NPC, just those within visible range and it won't result in any major transformations to a character's life. The smaller the language model, the less accurate it is, and it has the potential to hallucinate more (i.e. produces results that aren't in training data).
While Inzoi's AI system isn't all that impressive, what I saw in action at the GDC made me think that Nvidia's Ace has huge potential for other genres, particularly large, open-world RPGs. Alternatives already exist as mods for certain games, such as Mantella for Skyrim, and it transforms the dull, repetitive nature of NPC 'conversations' into something far more realistic and immersive.
To transform such games into 'living, breathing worlds,' much larger models will be required and traditionally, this involves a cloud-based system. However, a local model would be far preferable for most PC gamers worldwide, which brings us to the topic of VRAM.
Nvidia has been offering 8 GB of memory on its mainstream graphics cards for years, and other than the glitch in the matrix that is the RTX 3060, it doesn't seem to want to change this any time soon. Intel and AMD have been doing the same, of course, but where 16 GB of VRAM is the preserve of Nvidia's high-end GPUs, such as the RTX 5070 Ti and RTX 5080, one can get that amount of memory on far cheaper Arc and Radeon cards.
But if Nvidia Ace really takes off and developers start to complain that they're being restricted about what they can achieve with the software suite, because of the size of the SLM (small language model) they're having to use, then the jolly green giant will have to respond by upping the amount of VRAM it offers across the board.
After all, other aspects of PCs/computers have needed to increase the minimum amount of memory they sport because of AI, such as Apple with its base Mac spec and Lunar Lake laptops having 16 GB because of Copilot.
It's not often that one can say AI is doing something really useful for gamers but in this case, I think Nvidia's Ace and competing systems may well be what pushes graphics cards to consign 8 GB of VRAM to history. Not textures, ray tracing, or frame generation but realistic NPC responses. Progress never quite goes in the direction you always expect it to.
Best CPU for gaming: Top chips from Intel and AMD.Best gaming motherboard: The right boards.Best graphics card: Your perfect pixel-pusher awaits.Best SSD for gaming: Get into the game first.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
an hour ago
- Yahoo
Trump gave China the AI chips it wanted. Beijing isn't saying thank you
In a surprising reversal of the United States' years-long technology restrictions on China, President Donald Trump last month allowed Nvidia to resume sales of a key AI chip designed specifically for the Chinese market. Yet rather than celebrating, Beijing's response has been noticeably lukewarm, despite having long urged Washington to ease the stringent export controls. In the weeks since the policy U-turn, Beijing has called the chip a security risk, summoned Nvidia for explanations and discouraged its companies from using it. The less-than-welcoming sentiment reflects Beijing's drive to build a self-sufficient semiconductor supply chain – and its confidence in the progress its rapidly advancing chip industry has made. But the cold shoulder may also represent some political posturing. Despite significant advances in its semiconductor sector, China still needs America's chips and technology. Experts said China's national champion Huawei has developed chips with performance comparable to — and in some cases surpassing — the newly approved Nvidia chip. However, China still wants the more advanced AI processors that remain blocked under US export controls. In the years since Trump first imposed tech restrictions on Huawei during his first term, China's chip technology has made significant strides, spurred by the frustration that mounted as Washington piled on export controls, said Xiang Ligang, director-general of a Beijing-based technology industry group and an advisor to the Ministry of Industry and Information Technology. 'We have this capability, it's not as they imagine – that if China is blocked, China won't be able to function, or that China will be finished,' he said. To him, the policy about-face only reflects the importance of having a wholly homegrown chip supply chain. 'For Chinese companies, we may only have one choice if we wish to ensure a relatively secure supply of chips – that means relying on our own domestically produced chips,' Xiang said. That may be China's goal, but in the high-stakes AI race, with all its national security implications, the US remains the leader, at least for now. China is not 'naive' The chip in focus is Nvidia's H20, which was released by the AI chip leader last year to maintain access to the Chinese market following strict export controls put in place under the Biden administration that stopped the export of chips with high processing power. Last month, Trump greenlit the sales of the chip to China after banning it in April as the US trade frictions with China deepened. Trump has justified his decision by calling the chip 'obsolete,' as it lags behind the company's cutting-edge AI processors like Blackwell or H100, from which H20 is derived. He and his officials appeared to have embraced a view long promoted by Nvidia's CEO Jensen Huang – that US can maintain its tech leadership only through ensuring its chips remain the global standard. 'You want to sell the Chinese enough that their developers get addicted to the American technology stack,' Commerce Secretary Howard Lutnick said last month. But the dramatic reversal has fueled questions about Trump's transactional approach to national security – once considered off-limits to bargaining. China, on the other hand, is alarmed by the alleged security risks of Nvidia's H20s like 'tracking and positioning' and 'remote shutdown' features, capabilities that some US lawmakers have called for but Nvidia denies it has placed in its chips. China's cyberspace watchdog and industry ministry have since summoned the American chip giant over the security concerns and urged firms to avoid H20 chips, a development which was previously reported by Bloomberg. One major Chinese tech company which has developed its AI models has received notice from the authorities urging it to exercise caution in the use of H20s, and advising it not to purchase them, a company insider said on the condition of anonymity. CNN has reached out to the ministry and the cyberspace authorities for comment. An Nvidia spokesperson told CNN that NVIDIA 'does not have 'backdoors' in our chips that would give anyone a remote way to access or control them.' 'Banning the sale of H20 in China would only harm US economic and technology leadership with zero national security benefit,' the spokesperson added. But China believes the US isn't playing fairly, Xiang said. 'What we actually want, you refuse to sell us. For the things you already consider obsolete, you still want to dump them into our market and occupy our market. Do you really think we're that naive?' he said. Still coveted Despite Beijing's concerns and the H20's reduced performance, the chips remain highly sought after by Chinese companies. Equity research firm Bernstein estimated that shipment of the chips to China this year would have reached 1.5 million units, or about 23 billion in revenue, without Trump's export restrictions. Major buyers include Chinese tech giants such as TikTok owner ByteDance, Alibaba and Tencent. While Huawei's top AI chips excel in computing power – one of the key measures in evaluating processors' performance – in comparison with H20, they fall short in terms of memory bandwidth, which determines how much data can move between a chip's memory and computing unit. That bandwidth depends on a technology known as High Bandwidth Memory (HBM) used in AI chips to ensure efficient data transmission in AI model training. China's top HBM maker CXMT, or ChangXin Memory Technologies, is still about three to four years behind industry leaders like South Korea's SK Hynix and Samsung, and American Micron, according to MS Hwang, research director at Counterpoint Research, a research firm. Last year, the Biden administration further tightened export controls on China, including restrictions on HBM sales, forcing Chinese companies to rely on existing stockpiles. Beijing has requested Washington to lift restrictions on HBM as part of the trade deal negotiations, Financial Times reported this week. Key appeal of H20 for Chinese companies also lies in Huawei's limited production capacity and Nvidia's well-established ecosystem, said Qingyuan Lin, senior analyst at Bernstein focusing China's semiconductor industry. 'Even when you want to completely replace the H20 demand with the local guys, they're not able to deliver the amount of chips that's needed,' he said. The supply bottlenecks stem from constraints in scaling up production of both the manufacturing of computing units of the AI chips and the integration of various components in them, a technology known as advanced packaging in the industry, Lin said. Bernstein estimated that Huawei's shipments of its advanced AI chips in 2025 would amount to around 700,000 units, still far short of the demand in the country. CNN has reached out to Huawei for comment. Meanwhile, Nvidia's powerful ecosystem, which integrates its chips with its software platform, has created what experts call a 'moat,' making it difficult and costly for AI developers who train models on its software to switch to alternatives. 'The H20 comes with a complete ecosystem covering both hardware and software support, ensuring better compatibility and ease of integration,' said Brady Wang, associate director at Counterpoint. 'This ecosystem maturity is still a challenge for many Chinese-developed chips, making the H20 more attractive despite its cost disadvantage.' 'Very close' Still, experts said China's rapid progress in semiconductor technology should not be underestimated. Years of tightening export controls have injected both urgency and opportunity into Beijing's push for self-sufficiency, Lin said. While chipmaking technology appeared to stall after Huawei's 2023 flagship smartphone showcased advanced chips that American officials had deemed extremely difficult to produce, domestic chipmaking equipment companies have been steadily gaining ground, he said. 'The local guys actually had very little chance to gain share from the global players because of the technology gap, but export controls created a market that didn't exist before and accelerated the domestic substitution,' he said. Bernstein projects that the percentage of homemade AI chips in China will surge from 17% in 2023 to 55% by 2027, while American suppliers like Nvidia and AMD will shrink to 45% from 83%. In April, Huang of Nvidia met with Trump in Washington, urging the administration to loosen export controls on chips and saying that the diffusion of American AI technology around the world needs to be accelerated. 'There's no question that Huawei is one of the most formidable technology companies in the world…they made enormous progress in the last several years,' he said. 'China is right behind us. We're very, very close.' CNN's Hassan Tayir and Fred He contributed reporting. 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
2 hours ago
- Yahoo
S&P 500 Hits 6,400 on AI Boom: ETFs in Focus
The S&P 500 reached a new milestone on Aug. 13, 2025, closing above 6,400 for the first time. The rally was driven by large-cap technology stocks. While Trump's tariffs caused some upheavals in the early phase of the year, the tech stocks once again have led the charge. Jessica Rabe, co-founder of DataTrek Research, noted that investors are continuing to favor large-cap U.S. tech stocks over large caps in general — and that this trend is not done yet. Top 20 Stocks Outperforming the Index Rabe highlighted that the 20 largest companies on the S&P 500 have risen at an average of 40.6% since the market bottom compared to the index's overall 27.9% gain. This means these top holdings have pulled the index higher, while the remaining 480 stocks have been a relative drag. Most of the outperformers — including NVIDIA (NVDA), Microsoft (MSFT), Apple (AAPL), Amazon (AMZN), Alphabet (GOOGL, GOOG), Meta (META), Broadcom (AVGO), Tesla (TSLA), JPMorgan (JPM), Netflix (NFLX), Oracle (ORCL), and Palantir (PLTR) — share a common chord: AI-backed fundamentals. Industrials Also Riding the AI Wave Citi US equity strategist Scott Chronert, who recently boosted his year-end S&P 500 target to 6,600, said the Industrials rally is also tied to AI spending, with companies benefiting from the technology's infrastructure demands, as quoted on Yahoo Finance. Chronert believes the real opportunity lies in the longer term, as more companies will be adopting AI to improve margins and productivity. AI ETFs in Focus Against this backdrop, below we highlight a few AI-based exchange-traded funds (ETFs) that deserve a place in your portfolio. Global X Artificial Intelligence & Technology ETF AIQ The underlying Indxx Artificial Intelligence & Big Data Index is designed to provide exposure to exchange-listed companies in developed markets that are positioned to benefit from the further development and implementation of artificial intelligence technology, as well as to companies that provide critical technology and services for the analysis of large and complex data sets. The fund charges 68 bps in fees. Invest in Gold Thor Metals Group: Best Overall Gold IRA Priority Gold: Up to $15k in Free Silver + Zero Account Fees on Qualifying Purchase American Hartford Gold: #1 Precious Metals Dealer in the Nation iShares Exponential Technologies ETF XT The underlying Morningstar Exponential Technologies Index measures the performance of equity securities that are involved with the creation of groundbreaking technologies or that are users that apply such technologies within their businesses. The fund charges 46 bps in fees. Global X Robotics & Artificial Intelligence ETF BOTZ The underlying Indxx Global Robotics & Artificial Intelligence Thematic Index invests in companies that potentially stand to benefit from increased adoption and utilization of robotics and artificial intelligence, including those involved with industrial robotics and automation, non-industrial robots, and autonomous vehicles. The fund charges 68 bps in fees. ARK Autonomous Technology & Robotics ETF ARKQ The ARK Autonomous Technology & Robotics ETF is an actively managed ETF that seeks long-term growth of capital by investing under normal circumstances primarily in domestic and foreign equity securities of autonomous technology and robotics companies that follow the theme of disruptive innovation. The fund charges 75 bps in fees. ROBO Global Robotics & Automation Index ETF ROBO The underlying ROBO Global Robotics and Automation Index measures the performance of companies that derive a portion of their revenues and profits from robotics-related or automation-related products or services. The fund charges 95 bps in fees. Want the latest recommendations from Zacks Investment Research? Today, you can download 7 Best Stocks for the Next 30 Days. Click to get this free report ARK Autonomous Technology & Robotics ETF (ARKQ): ETF Research Reports ROBO Global Robotics and Automation Index ETF (ROBO): ETF Research Reports Global X Robotics & Artificial Intelligence ETF (BOTZ): ETF Research Reports iShares Exponential Technologies ETF (XT): ETF Research Reports Global X Artificial Intelligence & Technology ETF (AIQ): ETF Research Reports This article originally published on Zacks Investment Research ( Zacks Investment Research
Yahoo
2 hours ago
- Yahoo
The iPhone 17 Pro's Pixel-like camera bar could house a new camera with a much longer zoom
When you buy through links on our articles, Future and its syndication partners may earn a commission. The latest iPhone 17 Pro rumors point to Apple's premium flagship handset getting a new telephoto camera with 8x optical zoom, up from 5x zoom on the iPhone 16 Pro. What's more, this new camera is tipped to make use of moving components that could allow for continuous optical zoom between an as-yet unknown range of focal lengths. That's rarely-seen technology, usually reserved for phones that are specifically designed for photography like the Sony Xperia 1 VI, and the idea of such niche tech coming to one of the world's most popular phones is properly exciting. So why aren't I jumping for joy? On paper, I should be. Most flagship phones – and some particularly well-specced mid-range handsets like the Oppo Reno 13 Pro – come with a telephoto camera, offering optical zoom through the addition of a specialized lens and sensor. However, in order to create a continuous zoom between different lenses, phones typically use digital zoom to close the distance between their various lenses. For example, a phone with a main camera at 1x zoom and a telephoto camera at 3x optical zoom will digitally zoom the main camera to achieve 2x zoom. If those aforementioned iPhone 17 Pro rumors hold true, the new telephoto camera will add even more reach and flexibility to what is already one of the best camera phones. It could also explain the new camera housing that many rumors and leaks predict the iPhone 17 Pro and iPhone 17 Pro Max to launch with. We previously thought that this design change was a purely aesthetic choice, but this Pixel-style camera bar could, in fact, house a periscope lens or other tech to support the new zoom camera. With that all said, I can't help but wonder whether I'd actually use this (admittedly very cool) new feature. Once the excitement around the iPhone 17 Pro's possible 8x zoom camera settles, will it still be worth using? Three times is the charm I'm a maximalist when it comes to tech – I like to have the best specs possible, and for a long time, I found myself pining after the 5x or even 10x telephoto lenses fitted to some of the best camera phones. I specifically remember feeling amazed when the Samsung Galaxy S21 Ultra launched with a 10x optical zoom camera. But to be honest, it's rare that I find myself zooming in that much – apparently I'm not the only one, either, as Samsung scaled down its long-range zoom camera from 10x to 5x with the release of the Galaxy S24 Ultra. As I mentioned before, unless you're zooming into or beyond the 10x mark, a 10x zoom camera will be of no use. If you ask me, the sweet spot is actually a 3x telephoto camera, like the one the iPhone 15 Pro had before it was upped to 5x on the iPhone 16 Pro. It's not that I don't see the point in longer zoom lenses. In fact, last year, I wrote that the Oppo Find X8 Pro's 6x zoom camera had opened up new possibilities when it comes to mobile photography. It's just that 3x strikes a better balance between reach and everyday usefulness; a 3x zoom lens brings details forward without feeling cramped, and the smaller lens distortion is ideal for more natural portraits. I'm not the only 3x zoom fan in the TechRadar office, either – last year, Phones Editor Axel Metz wrote that the iPhone 15 Pro's 3x telephoto camera was his go-to for natural and more professional-looking social media pics, like those below. Image 1 of 5 Image 2 of 5 Image 3 of 5 Image 4 of 5 Image 5 of 5 An 8x zoom camera – or even better, a lens that continuously zooms from, say, 5x to 8x – would be a win for certain users: concertgoers, nature photographers, maybe even portrait photographers who don't mind standing pretty far away. But without an intermediate telephoto lens to bridge the long gap between 1x and 8x, for me, this won't be a selling point. Even the Galaxy S21 Ultra switched to a 3x camera before you got to the full 10x zoom, and while the iPhone's digital zoom is excellent, there'd surely be a noticeable loss of quality at zoom levels that don't activate the new camera. I'd be reluctant to give up my 3x lens without some kind of software magic to get around that. We're expecting the iPhone 17 series to be revealed in September 2025, so we hopefully won't have to wait long to see what Apple's cooking up. What do you think about this rumor? Would an 8x telephoto camera be a boost to the iPhone, or has tech gone too far? Let us know in the comments below. You might also like Is this the iPhone 17 Pro? Leak appears to show secret flagship in the wild with new camera setup If you're having problems searching for public transit routes on Google Maps, you're not alone The Samsung Galaxy S26 could come with multiple AI assistants – while the Ultra model is tipped for a major camera upgrade
