Apple may again be partnering with silicon chip manufacturer TSMC to produce its own AI server processor, according to a leak from Chinese social network Weibo.
Yes, news of Apple’s next step into the world of artificial intelligence tools is unironically brought to you by, MacRumors reports, “the Weibo user known as ‘Phone Chip Expert’”, who suggests that the processor will be produced using TSMC’s state of the art 3 nanometer node.
As MacRumors points out, the Weibo user known as Phone Chip Expert has form, having correctly identified ahead of formal announcements that the iPhone 7 would be water resistant and that the A16 Bionic chip would be exclusive to the iPhone 14’s Pro variant.
Apple AI progress
The Weibo user known as Phone Chip Expert may well be about to strike again with their clairvoyant powers, but it’s unclear as to exactly when Apple would formally announce such an AI processor, let alone launch it commercially.
In an increasingly AI-crazed world driven by data centers, it doesn’t surprise us that Apple are striving to be self-sufficient in its cloud computing processes.
Apple is a behemoth large enough to run its own data centers, and as generative AI tools, such as Apple’s own upcoming on-device large language model (LLM), increasingly trickle down to B2B and consumer audiences, it may as well exert as much control and oversight as possible over how that processing is done.
It’s clear that Apple have designs in the ‘AI space’ (blech), and supposedly even have credible ideas about it might improve our lives, but neither we, you, or the Weibo user known as Phone Chip Expert will truly know what those are until, probably, the company’s Worldwide Developers Conference (WWDC) in June.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Today is Earth Day, and it’s as good a time as any to think about the way we use technology and how we can use it better and more sustainably given the myriad challenges it represents to our environment.
Major computer manufacturers like Acer, Apple, Dell, HP, Lenovo, and many others have all started moving towards more sustainable products, and not just in their use of better packing materials that reduce new material use by increasing the amount of post-consumer material in their construction.
Even the computers themselves are starting to use post-consumer materials and manufacturers are expanding opportunities for upgrading the devices to keep them current longer, thereby reducing e-waste around the world.
But one area hasn’t seen nearly enough attention: processors, specifically their power usage.
A computer’s CPU is the brains of the whole operation, so naturally, it needs a good bit of power to operate at higher levels of performance. This is even more true of dedicated GPU chips in laptops or the best graphics cards used in desktop systems.
And while better power efficiency in laptops is going to be a bigger plus in terms of battery life, desktops have seemingly gone in the complete opposite direction, with Intel, AMD, and Nvidia configured systems drawing a lot of power to run hardware that often exceeds what users need, all so they can be called the ‘fastest’ or ‘most powerful’. It’s not a sustainable approach.
We need to start emphasizing efficiency over power
There will be circumstances when a lot of power for a component is necessary to do important work, and I’m not saying that every graphics card needs to have its power consumption halved as a basic rule.
Get the hottest deals available in your inbox plus news, reviews, opinion, analysis and more from the TechRadar team.
But for consumers who are barely tapping into the performance that an Nvidia RTX 4080 Super brings to the table, much less what an Nvidia RTX 4090 or AMD RX 7900 XTX offers, you have to ask if this kind of performance is worth the cost in terms of carbon emissions.
And of course, this is only in terms of useful work, like video editing or gaming, and not for something like cryptocurrency mining, which has at best a marginal social utility, and whose cost in terms of energy usage in the aggregate far outstrips any practical benefit cryptocurrency has in the real world (unless you’re really into criminal activity or need to launder some money).
Regular old processors aren’t immune either, with the current generation of Intel processors soaking up an extraordinary amount of energy (at least in bursts) relative to competitors like AMD and especially Apple.
What you get for that energy draw is some incredible performance numbers, but for 97-98% of users, this kind of performance is absolutely unnecessary, even if users are using the appropriate processor for their needs.
The performance arms race needs to change
Ultimately, AMD, Intel, and Nvidia are locked in an arms race for making the fastest and most powerful processors and graphics cards, and it shows no signs of stopping. Meanwhile, Apple’s move to its own M-series silicon has been a major win, both for the company and for consumers.
Apple’s chips are based on Arm‘s big.LITTLE architecture, which is incredibly energy efficient from the ground up, originally having been intended for mobile devices. But now that the architecture is sophisticated enough to be used in laptops and even desktops, the energy efficiency remains while Apple has scored major performance gains over both AMD and Intel and kept energy use down.
If we are going to maintain a livable planet in the future, we must go on an energy diet. Seeing the difference between what Apple’s done and what AMD, Intel, and Nvidia aren’t doing puts to bed any excuse those three latter chipmakers have for not refocusing on efficiency going forward.
As hard as it may be to hear or accept, these three chipmakers must acknowledge that we don’t have a performance problem, we have a sustainability problem. They should turn away from squeezing even more performance out of their hardware that we don’t need, and give us the efficiency that is desperately needed, especially when these marginal increases in power are coming at far too high a cost.
Some power users and enthusiasts might not like seeing decreasing power usage while maintaining roughly the same level of performance gen-on-gen, or slightly better performance but far less than with previous generations, but it’s what needs to be done, and the sooner everyone acknowledges this and adapts, the better.
Astera Labs, a relatively unknown semiconductor company with a $10 billion market value, recently showed off its new Aries 6 PCIe retimer board, which it is now sampling to leading AI and cloud infrastructure providers.
Due to the high-speed data transfer within the PCIe interface, the signals can suffer from degradation, especially over longer distances or due to interference. A PCIe retimer helps maintain the data signal integrity over the PCIe interface by cleaning, reshaping, and retransmitting the data signals.
The Aries 6 retimers, the first in Astera Labs’ PCIe 6.x portfolio, offer robust, low power, and efficient PCIe 6.x and CXL 3.x connectivity, squarely aimed at catering to the networking requirements of next-generation GPUs, accelerators, CPUs, and CXL memory controllers.
3x the reach
Based on Intel‘s Retimer Supplemental Specification, the new Aries 6 retimers build upon the company’s widely deployed PCIe 5.0 retimer portfolio and reportedly extend reach distance by three times the standard rate.
Casey Morrison, Chief Product Officer, Astera Labs, said, “PCIe 6.x technology’s superior bandwidth is required to handle data-intensive workloads and to maximize utilization of AI accelerators, but the faster speeds introduce new signal integrity issues in hyperscale platforms. Aries Smart DSP Retimers have set the gold standard for addressing critical PCIe/CXL connectivity challenges with a solid track record of robust performance and seamless interoperability. We’re proud that our third generation of Aries Retimers with support for PCIe 6.x, PCIe 5.x, and CXL 3.x have now been sampled to leading AI and cloud platform providers.”
ServeTheHometook a first look at Astera Labs’s new portfolio at Nvidia GTC 2024 (see the photograph above) and observed that the “Aries 6 was linked at PCIe Gen6 x16 speeds at 10-11W of power consumption. That is a big deal since it is lower than Broadcom is claiming with its new retimers.“
Many major AI firms are excited about the imminent arrival of Astera Labs’ new generation Aries Smart DSP Retimers. Raghu Nambiar, Corporate Vice President, Data Center Ecosystems and Solutions, AMD, said, “Our close collaboration with Astera Labs on PCIe technologies ensures our customers’ platforms continue to meet the higher bandwidth connectivity requirements of next-generation AI and HPC workloads,” while Brian Kelleher, Senior Vice President of GPU Engineering, Nvidia, added, “Astera Labs’ new Aries Smart DSP Retimers with support for PCIe 6.2 will help enable higher bandwidth to optimize utilization of our next-generation computing platforms.”
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
NVIDIA sees AI as a means of putting new tools into the hands of gamers and creators alike. NVIDIA Instant NeRF is one such tool, leveraging the power of NVIDIA’s GPUs to make complex 3D creations orders of magnitude easier to generate. Instant NeRF is an especially powerful tool in its ability to create these 3D scenes and objects.
In effect, NVIDIA Instant NeRF takes a series of 2D images, figures out how they overlap, and uses that knowledge to create an entire 3D scene. A NeRF (or Neural Radiance Field) isn’t a new thing, but the process to create one was not fast. By applying machine learning techniques to the process and specialized hardware, NVIDIA was able to make it much quicker, enough to be almost instant — thus Instant NeRF.
Being able to snap a series of photos or even record a video of a scene and then turn it into a freely explorable 3D environment offers a new realm of creative possibility for artists. It also provides a quick way to turn a real-world object into a 3D one.
Some artists are already realizing the potential of Instant NeRF. In a few artist showcases, NVIDIA highlights artists’ abilities to share historic artworks, capture memories, and allow viewers of the artworks to more fully immerse themselves in the scenes without being beholden to the original composition.
Karen X. Cheng explores the potential of this tool in her creation, Through the Looking Glass, which uses NVIDIA Instant NeRF to create the 3D scene through which her camera ventures, eventually slipping through a mirror into an inverted world.
Hugues Bruyère uses Instant NeRF in his creation, Zeus, to present a historic sculpture from the Royal Ontario Museum in a new way. This gives those who may never have a chance to see it in person the ability to view it from all angles nonetheless.
(Image credit: NVIDIA)
With tools like Instant NeRF, it’s clear that NVIDIA’s latest hardware has much more than just gamers in mind. With more and more dedicated AI power built into each chip, NVIDIA RTX GPUs are bringing new levels of AI performance to the table that can serve gamers and creators alike.
The same Tensor Cores that make it possible to infer what a 4K frame in a game would look like using a 1080p frame as a reference are also making it possible to infer what a fully fleshed out 3D scene would look like using a series of 2D images. And NVIDIA’s latest GPUs put those tools right into your hands.
Instant NeRF isn’t something you just get to hear about. It’s actually a tool you can try for yourself. Developers can dive right in with this guide, and less technical users can grab a simpler Windows installer here which even includes a demo photo set. Since Instant NeRF runs on RTX GPUs, it’s widely available, though the latest RTX 40 Series and RTX Ada GPUs can turn out results even faster.
The ability of NVIDIA’s hardware to accelerate AI is key to powering a new generation of AI PCs. Instant NeRF is just one of many examples of how NVIDIA’s GPUs are enabling new capabilities or dramatically speeding up existing tools. To help you explore the latest developments in AI and present them in an easy-to-understand format, NVIDIA has introduced the AI Decoded blog series. You can also see all the ways NVIDIA is boosting AI performance at NVIDIA’s RTX for AI page.
Microsoft and OpenAI are reportedly in the process of planning a groundbreaking data center project which would include an AI supercomputer named “Stargate”.
A report by Anissa Gardizy and Amir Efrati in The Information claims the goal of the project, which would be financed by Microsoft to the tune of over $100 billion, and which reportedly has a launch date set for 2028, is to reduce the two companies’ reliance on Nvidia, something that a lot of the tech giants involved in AI are increasingly looking to try to do.
Microsoft and OpenAI’s plan reportedly involves five phases, with Stargate being the fifth and most ambitious one.
The data center will be the supercomputer
The cost of the project is attributed to the age-old “sources familiar with the plans” (The Information says these are “a person who spoke to OpenAI CEO Sam Altman about it and a person who has viewed some of Microsoft’s initial cost estimates”), but neither Microsoft nor OpenAI have yet commented on the specifics of the project.
The new data center project is expected to push the boundaries of AI capability and could potentially exceed $115 billion in expenses. This is more than triple the amount Microsoft spent on capital expenditures for servers and equipment last year. Microsoft is currently working on a smaller, fourth-phase supercomputer for OpenAI that is expected to launch around 2026, The Information claims.
Shedding more light on the report, The Next Platform says, “The first thing to note about the rumored “Stargate” system that Microsoft is planning to build to support the computational needs of its large language model partner, OpenAI, is that the people doing the talking – reportedly OpenAI chief executive officer Sam Altman – are talking about a data center, not a supercomputer. And that is because the data center – and perhaps multiple data centers within a region with perhaps as many as 1 million XPU computational devices – will be the supercomputer.”
The Next Platform also says if Stargate does come to fruition it will be “based on future generations of Cobalt Arm server processors and Maia XPUs, with Ethernet scaling to hundreds of thousands to 1 million XPUs in a single machine,” and it definitely won’t be based on Nvidia GPUs and interconnects, which seems like a safe bet if the rumors are to be believed.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
There’s no shortage of startups pushing technology that could one day prove pivotal in AI computing and memory infrastructure.
Celestial AI, which recently secured $175 million in Series C funding, is looking to commercialize its Photonic Fabric technology which aims to redefine optical interconnects.
Celestial AI’s foundational technology is designed to disaggregate AI compute from memory to offer a “transformative leap in AI system performance that is ten years more advanced than existing technologies.”
Lower energy overhead and latency
The company has reportedly been in talks with several hyperscale customers and a major processor manufacturer, about integrating its technology. Though specific details remain under wraps, that manufacturer is quite likely to be AMD since AMD Ventures is one of Photonic Fabric’s backers.
As reported by The Next Platform, the core of Celestial AI’s strategy lies in its chiplets, interposers, and optical interconnect technology. By combining DDR5 and HBM memory, the company aims to significantly reduce power consumption while maintaining high performance levels. The chiplets can be used for additional memory capacity or as interconnects between chips, offering speeds comparable to NVLink or Infinity Fabric.
“The surge in demand for our Photonic Fabric is the product of having the right technology, the right team and the right customer engagement model”, said Dave Lazovsky, Co-Founder and CEO of Celestial AI.
“We are experiencing broad customer adoption resulting from our full-stack technology offerings, providing electrical-optical-electrical links that deliver data at the bandwidth, latency, bit error rate (BER) and power required, compatible with the logical protocols of our customer’s AI accelerators and GPUs. Deep strategic collaborations with hyperscale data center customers focused on optimizing system-level Accelerated Computing architectures are a prerequisite for these solutions. We’re excited to be working with the giants of our industry to propel commercialization of the Photonic Fabric.”
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
While Celestial AI faces challenges in timing and competition from other startups in the silicon photonics space, the potential impact of its technology on the AI processing landscape makes it a promising contender. As the industry moves towards co-packaged optics and silicon photonic interposers, Celestial AI’s Photonic Fabric could play a key role in shaping the future of AI computing.
Horizon Forbidden Westhas come to PC, and it’s given me another reason not to buy a PS5. I’ve bought every generation of PlayStation console since the OG model, but with Sony‘s shift to (belatedly) porting most of its exclusives to PC, it just doesn’t seem worth splashing out on a new console when I can just wait for the games I want to play to come to me.
So, I was very happy to hear that Horizon Forbidden West was going to be ported to PC. As a big fan of the original game, which I played on PS4, I’d been looking forward to playing it.
Of course, as a visually-impressive first-party game from Sony, I was also keen to see how it performed on our 8K rig. As you can see in the specs box on the right, our rig has remained largely unchanged for over a year. This is because it remains a formidable machine – and, crucially, the Nvidia GeForce RTX 4090 graphics card that does the bulk of the work when gaming has yet to be beaten. It remains the best graphics card money can buy.
With rumors swirling that Sony is planning on releasing a more powerful PS5 Pro console in the near future that could target 8K resolutions through a mix of more powerful hardware and upscaling technology, Horizon Forbidden West at 8K on PC may give us an idea of the kind of visuals future PlayStation games may offer.
It also suggests what obstacles Sony will face if the PS5 Pro will indeed target 8K resolutions. Despite being almost two years old, the RTX 4090 GPU still costs more than its original launch price, hovering around $2,000/£2,000. While the PS5 Pro will likely be more expensive than the standard PS5, there’s no way it’ll be even half the price of Nvidia’s GPU – and that’s before you add in the cost of the other PC components required. Basically, you can’t currently buy an affordable 8K gaming machine that is priced for mainstream success. That’s the scale of the challenge Sony faces.
(Image credit: Future)
Spoilt for choice
One of the best things about Sony’s initiative to bring its games to PC, apart from giving me an excuse not to spend money I don’t have on a PS5, is that they usually come with an excellent choice of PC-centric options, including support for upscaling technology from Nvidia and support for ultrawide monitors.
Horizon Forbidden West continues this streak, and the PC port has been handled by Nixxes Software, which has handled many previous PlayStation to PC ports.
Get the hottest deals available in your inbox plus news, reviews, opinion, analysis and more from the TechRadar team.
This latest release is particularly noteworthy as not only does it support DLSS 3 for Nvidia RTX graphics, but it also supports competing upscaling tech in the form of AMD FSR 2.2 and Intel XeSS.
All three of these features allow the game to run at a lower resolution, with the images upscaled so that the game appears at a higher resolution, but without the additional strain on your PC’s graphics card.
This mainly allows less powerful GPUs to hit resolutions with graphical effects enabled that they usually wouldn’t be able to handle. It also allows the mighty RTX 4090 to reach the demanding 8K resolution (7680 × 4320) in certain games while maintaining a playable framerate.
By supporting the three major upscaling tools, Horizon Forbidden West gives users much more choice (both FSR and XeSS work for a range of GPUs, while DLSS is exclusive to recent Nvidia GPUs) – and it also gives me a chance to see which upscaling tech performs the best.
(Image credit: Sony)
First up: DLSS
First, I played Horizon Forbidden West at the 8K resolution of 7680 × 4320 and the graphics preset at ‘Very High’ – which is the highest quality on offer. With DLSS turned off (so the game is running at native 8K), my 8K test rig managed to run Horizon Forbidden West at an average of 32 frames per second (fps).
Considering that this is a graphically-intensive game and running at the highest graphics and at a resolution that’s pushing around 33 million pixels, this is very impressive, and is a testament to the raw power of the RTX 4090, the rest of the components inside the rig built by Stormforce Gaming, and the talents of Guerrilla Games (developers of the game) and Nixxes Software.
I feel that 30fps is the minimum frame rate for a playable game, so if you wanted to play Horizon Forbidden West at a native 8K resolution, that’s certainly possible. If you drop the graphics preset, then the frame rate will go up – though at the cost of graphical fidelity.
Of course, you don’t spend around $2,000 on a GPU to get 32fps in a game, so I turned on DLSS and set it to ‘Quality’, which minimizes the amount of upscaling performed to preserve image quality as much as possible. This led the average framerate to jump to 45fps, with a maximum frames per second of 60.7fps.
One thing to note with my results, which you can view in the chart above, is that because Horizon Forbidden West doesn’t have a built-in benchmark tool, I had to play the same section over and over again, using MSI Afterburner to record my framerate. I chose a section of the game with large open spaces, water effects and a combat encounter, and I tried to make each playthrough, lasting around eight minutes, as similar as possible. However, my playthroughs weren’t identical, as some things, such as enemy attacks, would change, and this explains why there are some discrepancies between results. Still, it should give you a good idea of the difference each setting makes.
Next, I turned ‘Frame Generation’ on. This is a new feature exclusive to DLSS 3 and Nvidia’s RTX 4000 series of cards. It uses AI to generate and insert frames between normal frames rendered by the GPU. The goal is to make games feel even smoother with higher, more consistent framerates while maintaining image quality.
As the chart shows, this gave the game another bump in frames per second. I then tested the other DLSS settings with Frame Generation left on.
With DLSS set to Ultra Performance, I hit 59.3fps at 8K – basically the 60fps goal I aim for in these tests, which offers a balance of image quality and performance. With Ultra Performance, the RTX 4090 is rendering the game at a much lower resolution, then using DLSS to upscale to 8K, and this reliance on upscaling can lead to an image quality that can suffer from a lack of sharpness and detail, and graphical artifacts. The good news is that DLSS 3 is a big improvement over previous versions, and the hit to graphic quality is far less noticeable these days.
So, thanks to DLSS, you can indeed play Horizon Forbidden West at 8K. But how does AMD and Intel’s rival technologies cope?
(Image credit: Sony Interactive Entertainment)
AMD FSR 2.2 tested
AMD’s FSR 2.2 technology isn’t as mature as Nvidia’s DLSS 3, but it has a noteworthy feature that DLSS lacks: it’s open source and doesn’t just work with AMD graphics cards – Nvidia and Intel GPUs can make use of it as well.
This makes it far more accessible than DLSS, which is exclusive to new and expensive Nvidia GPUs, and for many people this flexibility makes up for any shortfall in performance.
As you can see from my results above, FSR 2.2 provides a decent jump in frame rates compared to running Horizon Forbidden West natively at 8K, though at each quality setting, it doesn’t quite keep up with DLSS 3’s results.
The best results I managed was with FSR set to ‘Ultra Performance’, where it hit 55.2fps on average. Below DLSS 3’s best results, but certainly not bad, and close to doubling the performance of the game compared with playing it natively.
As well as being unable to hit the same highs as DLSS 3, AMD FSR 2.2’s image quality at Ultra Performance isn’t quite as good as DLSS 3 at similar settings, with a few instances of shimmering and ghosting becoming noticeable during my playthrough.
(Image credit: Sony)
Intel XeSS results
Finally, I tested out Intel’s XeSS technology. While there is a version of XeSS designed to run with Intel Arc graphics cards, as with FSR you can use XeSS with various GPU brands, so there is yet another upscaling tool that gamers can try out. As with most things, the more choice there is for consumers, the better.
XeSS hasn’t been around for as long as DLSS or FSR, and as you can see from the results above, it wasn’t able to match either of Nvidia or AMD’s solutions. There’s no ‘Ultra Performance’ mode either, so XeSS hits its highest framerates with XeSS set to ‘Performance’, with an average of 50.6fps. This leads to a perfectly playable experience at 8K, but it’s noticeably more sluggish than when playing with DLSS at Ultra Performance.
However, it still gives you a decent fps bump over native 8K, and with Intel being one of the biggest proponents of artificial intelligence, I’m pretty confident that XeSS performance will improve as the technology matures. The fact that you can use it with GPUs from Intel’s rivals is also a big plus.
(Image credit: Sony)
Conclusion: DLSS for the win (again)
Once again, DLSS 3 has proved to be the best way of getting a game to run at 8K and 60fps with minimal compromises.
Not only did it allow the RTX 4090 to hit 59.3fps on average while playing Horizon Forbidden West, but it also looked the best with minimal impact to image quality.
This may not come as too much of a surprise – DLSS has been around for quite a while now, and Nvidia has been putting a lot of work into improving the technology with each release.
Also, while Nvidia’s preference for proprietary tech means you need the latest RTX 4000 series of GPUs to get the most out of it, this does at least mean Team Green can make use of exclusive features of its GPUs such as Tensor Cores. With AMD and Intel’s more open implementations, they are unable to target specific hardware as easily – though FSR and XeSS are available to a much wider range of PC gamers.
And, while FSR doesn’t quite match DLSS performance with Horizon Forbidden West, it comes close, and if you don’t have an Nvidia GPU, this is a fine alternative. As for XeSS, it shows plenty of promise.
So, upscaling tech has made gaming at 8K on PC achievable, and it’s great to see increased choices for users. So, if Sony is indeed working on a PS5 Pro that aims to run games like Horizon Forbidden West at 8K, it’s going to have to come up with its own upscaling tech (or adapt FSR or XeSS) if it wants to compete.
According to Israeli startup NeuReality, many AI possibilities aren’t fully realized due to the cost and complexity of building and scaling AI systems.
Current solutions are not optimized for inference and rely on general-purpose CPUs, which were not designed for AI. Moreover, CPU-centric architectures necessitate multiple hardware components, resulting in underutilized Deep Learning Accelerators (DLAs) due to CPU bottlenecks.
NeuReality’s answer to this problem is the NR1AI Inference Solution, a combination of purpose-built software and a unique network addressable inference server-on-a-chip. NeuReality says this will deliver improved performance and scalability at a lower cost alongside reduced power consumption.
An express lane for large AI pipelines
“Our disruptive AI Inference technology is unbound by conventional CPUs, GPUs, and NICs,” said NeuReality’s CEO Moshe Tanach. “We didn’t try to just improve an already flawed system. Instead, we unpacked and redefined the ideal AI Inference system from top to bottom and end to end, to deliver breakthrough performance, cost savings, and energy efficiency.”
The key to NeuReality’s solution is a Network Addressable Processing Unit (NAPU), a new architecture design that leverages the power of DLAs. The NeuReality NR1, a network addressable inference Server-on-a-Chip, has an embedded Neural Network Engine and a NAPU.
This new architecture enables inference through hardware with AI-over-Fabric, an AI hypervisor, and AI-pipeline offload.
The company has two products that utilize its Server-on-a-Chip: the NR1-M AI Inference Module and the NR1-S AI Inference Appliance. The former is a Full-Height, Double-wide PCIe card that contains one NR1 NAPU system-on-a-chip and a network-addressable Inference Server that can connect to an external DLA. The latter is an AI-centric inference server containing NR1-M modules with the NR1 NAPU. NeuReality claims the server “lowers cost and power performance by up to 50X but doesn’t require IT to implement for end users.”
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
“Investing in more and more DLAs, GPUs, LPUs, TPUs… won’t address your core issue of system inefficiency,” said Tanach. “It’s akin to installing a faster engine in your car to navigate through traffic congestion and dead ends – it simply won’t get you to your destination any faster. NeuReality, on the other hand, provides an express lane for large AI pipelines, seamlessly routing tasks to purpose-built AI devices and swiftly delivering responses to your customers, while conserving both resources and capital.”
NeuReality recently secured $20 million in funding from the European Innovation Council (EIC) Fund, Varana Capital, Cleveland Avenue, XT Hi-Tech and OurCrowd.
In a move to cut its dependency on Nvidia‘s high-cost AI chips, Naver, the South Korean equivalent of Google, has signed a 1 trillion won ($750 million) agreement with Samsung.
The deal will see the tech giant supply its more affordable Mach-1 chips to Naver, by the end of 2024.
The Mach-1 chip, currently under development, is an AI accelerator in the form of a SoC that combines Samsung’s proprietary processors and low-power DRAM chips to reduce the bottleneck between the GPU and HBM.
Just the start
The announcement of the Mach-1 was made during Samsung’s 55th regular shareholders’ meeting. Kye Hyun Kyung, CEO of Samsung Semiconductor, said the chip design had passed technological validation on FPGAs and that finalization of SoC was in progress.
The exact volume of Mach-1 chips to be supplied and prices are still under discussion, but The Korea Economic Daily reports that Samsung intends to price the Mach-1 AI chip at around $3,756 each. The order is expected to be for somewhere between 150,000 and 200,000 units.
Naver plans to use Samsung’s Mach-1 chips to power servers for its AI map service, Naver Place. According to The Korea Economic Daily, Naver will order further Mach-1 chips if the initial batch performs as well as hoped.
Samsung sees this deal with Naver as just the start. The tech giant is reportedly in supply talks with Microsoft and Meta Platforms who, like Naver, are actively seeking to reduce their reliance on Nvidia’s AI hardware.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
With the Naver deal, Samsung is also looking to better compete with its South Korean rival SK Hynix, which is the dominant player in the advanced HBM segment. Samsung has been heavily investing in HBM recently and at the start of March announced the industry’s first 12-stack HBM3E 12H DRAM. This reportedly outperforms Micron’s 24GB 8H HBM3E in terms of capacity and bandwidth and is expected to begin shipping in Q2 this year.
Nvidia has been slowly launching its Super refreshes of the RTX 4000-series of graphics cards, updated versions with much-improved performance. And, according to a new rumor, the mid-range models might also be getting its own refresh.
There has been an alleged leak of plans for refreshed versions of RTX 4060, 4060 Ti, and 4070 graphics cards, which was revealed by Twitter/X user Zed__Wang and reported on by PCGamesN. Thanks to the table, we know the three brand-new dies for the cards as well as the release windows: March 2024 for the RTX 4070 and April 2024 for both the RTX 4060 and 4060 Ti.
(Image credit: Zed__Wang)
While details for these newer versions are sparse, we normally could assume that a larger die comes with superior memory bandwidth, VRAM capacity, and performance. However, according to the same leaker, this is not the case.
When asked about the spec difference between the original and the refreshed GPUs, they responded “only change the chip; the spec keeps the same.” It seems like, at most, we might see a few additional CUDA, RT, and Tensor cores but that’s all.
These mid-range refreshes could be a dud
If this leak is to be believed, then this refresh sounds like a whole lot of nothing. If the broader specs aren’t even being upgraded, then what’s even the point of releasing them in the first place? The only way it would make sense is if the new cards were cheaper than the basic versions.
One good thing about the Super series that have previously launched is that it’s already been lowering the prices of AMD’s RDNA 3 graphics cards, going by the dropping prices of the AMD Radeon RX 7700 XT and the RX 7900 XT once the RTX 4080 Super, 4070 Ti Super, and RTX 4070 Super hit retailers. So if these newer mid-range cards end up launching, then that could mean the matching AMD cards could also see a price decrease.
It’s rather sad that these are the only positives of getting RTX 4060, 4060 Ti, and 4070 refreshes if the rumors are actually true. Here’s hoping that if Nvidia does decide to release them, they’ll be worth investing in.
Get the hottest deals available in your inbox plus news, reviews, opinion, analysis and more from the TechRadar team.
