Graphics Cards - Page 87

The rumor mill is spinning up for NVIDIA's new GeForce RTX 40 SUPER series of graphics cards, with some new information suggesting the RTX 4080 SUPER will have 20GB of VRAM on a wider memory bus, which should position the card smack bang in the middle of the performance of the RTX 4090 and the RTX 4080.

Benchlife is now reporting new information from their AIB sources, which points to NVIDIA's new GeForce RTX 4080 SUPER graphics card featuring 20GB of GDDR6X memory. The AIB partner didn't say it would definitely be the "RTX 4080 SUPER," but it could be another RTX 4080 with 20GB of VRAM. I think that would be graphics card naming suicide, especially after the "unlaunched" GeForce RTX 4080 12GB.

NVIDIA's new GeForce RTX 4080 SUPER would make sense, especially with 20GB of GDDR6X memory on a wider 320-bit memory bus, up against the RTX 4080 and its 16GB GDDR6X on a 256-bit memory bus. The flagship GeForce RTX 4090 has 24GB of GDDR6X on an even larger 384-bit memory bus.

MaxSun has officially confirmed the price of its upcoming flagship GeForce RTX 4090 MegaGamer graphics card, the top-of-the-mountain version of the new MGG series the company launched earlier this year.

It is unfortunate that MaxSun doesn't have its graphics cards in more countries across the planet, as its GeForce RTX 4090 MegaGamer looks like an interesting RTX 4090 to enter the ring. It'll mostly be available to China and select Asian countries, while the rest of the world has to skip this one.

MaxSun has announced an official MSRP of $1999 on its new GeForce RTX 4090 MegaGamer, a $400 premium over teh reference price of the RTX 4090, which is $1599. MaxSun's new GeForce RTX 4090 MegaGamer is unique in the fact that it doesn't just have a couple of fans; there are FIVE cooling fans on this card. There's a pretty nice-looking two-piece aluminum body shell and a customizable RGB light bar. It does look cool, I'll give it that.

Samsung formally announced its next-gen GDDR7 memory back in October 2022, so we're about a year away from the announcement now, with the South Korean giant hosting its annual Memory Tech Day in San Jose on October 20.

At the event, Samsung announced its next-gen Shinebolt HBM3E memory, which is destined for AI GPUs and HPC. But the company also provided some small but important updates on its new GDDR7 memory. Samsung said that it's not just improving the active power consumption of GDDR7 over GDDR6, but its new GDDR7 memory will use 50% less standby power than GDDR6 does, which is thanks to additional clock controls.

We already know that Samsung will have 16Gbit (2GB) modules of GDDR6 at up to 32Gbps of projected speed, which represents a 33% improvement in bandwidth per pin over GDDR6. This means on a 256-bit memory interface, we should see a hefty 1TB/sec of memory bandwidth... while on a much wider 512-bit memory interface, that bandwidth skyrockets to 2TB/sec of memory bandwidth.

We heard it was coming in a leak not too long ago now, but now it's official: AMD has announced its new Radeon RX 7900M mobile GPU, its most powerful GPU inside of a laptop ever released.

The new AMD Radeon RX 7900M uses the Navi 31 GPU with 4608 Stream Processors, with a Game Clock of 1825MHz and a Boost Clock of up to 2090MHz; we have some powerhouse Radeon RX mobile GPU goodness in a new wave of Radeon RX 7900M-powered gaming laptops. There's also 16GB of GDDR6 memory clocked at 18Gbps on a 256-bit memory bus, while the Total Graphics Power (TGP) sits at up to 180W.

AMD is really pushing the limits with 180W TGP and up to 200W using SmartShift Max technology, while NVIDIA's flagship GeForce RTX 4090 Laptop GPU with 16GB GDDR6 memory also has a TGP of 150W but can hit up to 175W with Dynamic Boost technology.

Well, this isn't a surprise: NVIDIA's flagship GeForce RTX 4090 24GB graphics card has doubled in price in China because of US sanctions banning high-end GPUs from being exported to China.

The new rules set by the Biden administration will go into effect on November 16, 2023, after which NVIDIA won't be allowed to ship most of its high-end AI GPUs, and for gamers: NVIDIA can't ship its GeForce RTX 4090 24GB graphics card to China. Well, given that the GeForce RTX 4090 is effectively banned in China in a few weeks' time, the price of the RTX 4090 graphics cards has skyrocketed, doubling in price within hours.

NVIDIA's flagship GeForce RTX 4090 24GB graphics card launched in China back in October 2022, priced at 12,999 RMB. But now, the current price as of October 19, 2023 -- is around 26,000 RMB. You can see in the above screenshot that some of these Chinese outlets are selling a GeForce RTX 4090 for up to 29,999 RMB. A quick currency exchange turns that 29,999 RMB to around $4000 USD.

NVIDIA's blazing-fast Grace Hopper GH200 Superchip will never see the shores of China, with the US government applying sanctions on the country, stopping them from acquiring, buying, or touching any high-end supercomputer or semiconductor technology. This means GPUs, and this means NVIDIA losing parts of its market.

The US government has just imposed new restrictions, which are different to the first restrictions, which looked at the on-device connection speed. Still, with the new rules, it'll see NVIDIA's gaming-focused GeForce RTX 4090 24GB graphics card being banned from entering China. High-end enthusiast PC gamers will miss out on the best gaming GPU money can buy.

The Biden administration has pushed new rules that see limits on high-performance computing chips, with the new sanctions looking to destroy any ability of China being able to advance its high-performance computing chip technology. This new set of requirements will go into effect in 29 days, meaning NVIDIA has a very short amount of time -- while head-first into the holiday season -- to adapt to the Biden administration's new sanctions and regulatory changes.

Samsung's upcoming 5th generation HBM3e memory product has been codenamed "Shinebolt," with the marketing and development of Samsung's new HBM3e memory catching up to the great work SK hynix has been doing with its own next-gen memory.

According to sources at Business Korea, Samsung Electronics is currently shipping out HBM3e prototypes to some of its clients for quality approval (QA) testing. Samsung is reportedly shipping out 24 gigabit (Gb) chips in 8 layers (8-Hi), but will soon have a 36GB HBM3e product with 12 layers (12-Hi).

Samsung's new HBM3e "Shinebolt" memory will have a 50% performance boost over HBM3, with a huge 1.228TB/sec (12280MB/sec) memory bandwidth. The future is AI; we all know it, with NVIDIA's crazy-fast AI GPUs needing more and much faster VRAM. This is where Samsung's new HBM3e "Shinebolt" memory shines -- pun not intended, but it just works so well here -- capacity and bandwidth.

AMD has launched new GPUs this week, but in a strange move, it's two new models in the RDNA 2 powered Radeon RX 6000 Series, the Radeon RX 6750 GRE 10GB ($269) and Radeon RX 6750 GRE 12GB ($289).

These GRE or Golden Rabbit Edition graphics cards are available now in China, with AMD also providing prices in USD - pointing to these cards making their way to more regions soon. Like the Radeon RX 7900 GRE, these cards will likely be limited to system integrators outside China.

What makes this release confusing, outside of the fact that AMD is releasing previous-gen RDNA 2 hardware, is that although they share the same name, these new Radeon RX 6750 GRE 10GB and 12GB GPUs feature completely different specs and hardware. It's reminiscent of the GeForce RTX 4080 12GB and 16GB reveal from NVIDIA, which saw the company rename the 12GB model to become the GeForce RTX 4070 Ti to avoid confusion.

NVIDIA is expected to add 3 more members to the GeForce RTX 40 "Ada Lovelace" GPU family, with the purported GeForce RTX 4080 SUPER, GeForce RTX 4070 Ti SUPER, and finally, the GeForce RTX 4070 SUPER graphics cards.

The new rumor is coming from @hongxing2020, which reliably leaked that NVIDIA would be renaming its GeForce RTX 4080 12GB graphics card, which was correct in the end and was then "unlaunched" and then turned into the GeForce RTX 4070 Ti in the end. Now, the same leaker is talking about the refreshed "SUPER" family of graphics cards in the GeForce RTX 4080 SUPER, GeForce RTX 4070 Ti SUPER, and finally, the GeForce RTX 4070 SUPER.

NVIDIA's flagship GeForce RTX 4090 is a big chunk faster than the GeForce RTX 4080, with no GPU in between those two gaming beasts. The introduction of the GeForce RTX 4080 SUPER -- or even the GeForce RTX 4080 Ti -- makes sense in this position. We've heard rumors of the RTX 4080 Ti using the AD102 GPU, packing 20GB of GDDR6X, which positions it between the 16GB on the RTX 4080 and the 24GB on the RTX 4090.

AMD's new Anti-Lag+ technology has been a mess for weeks now, seeing Radeon RX 7000 series GPU owners experiencing game crashes, problems, and even VAC banning on Steam for Counter-Strike 2 players. It's been a headache, and now so much so that AMD has completely disabled Anti-Lag+ tech on its recent Radeon GPU drivers. Download the new 23.10.2 drivers here.

The company previously removed the old drivers -- 23.10.1 driver -- causing the issues with Anti-Lag+ and multiple games, while the new 23.10.2 driver disables Anti-Lag+ completely. This means that games that were experiencing problems with Anti-Lag+ were enabled, including Call of Duty: Modern Warfare II, PUBG, Apex Legends, and Counter-Strike 2.

Anti-cheating algorithms in games were detecting Anti-Lag+ as hacking-related, banning gamers across the world with an AMD Radeon RX 7000 series GPU. AMD's move of removing Anti-Lag+ from its Radeon GPU drivers is somewhat strange but hints at what should be the fact that AMD shouldn't have launched Anti-Lag+ in this state without knowing the consequences of its action for gamers.

The US Department of Commerce has announced new plans to prevent more AI GPUs from shipping to China, with the Biden administration implementing new rules to close up some of those loopholes after the US sanctions on China were placed on AI chips last year.

The initial AI GPUs that were banned were the NVIDIA H100, the number one choice for any firm working on AI, while Chinese companies weren't restricted from buying cut-down NVIDIA GPUs in the form of the H800 and A800 AI GPUs. The US restrictions didn't apply to these GPUs, but the new rules are banning those lower-end AI GPUs as well.

NVIDIA won't be the only company affected, but Intel and AMD will feel a little pain from the latest additions to the US sanctions on AI GPUs shipping into China. Not only that, but senior Biden administration officials have eluded to plans to extend the list of semiconductor manufacturing equipment subject to US restrictions. Gaming consoles, smartphones, and other consumer products with high-end chips inside of them will remain unaffected by US sanctions.

NVIDIA drops new Game Ready drivers for GeForce GPU owners fairly regularly, bringing day-one support to new games, adding to the number of DLSS-supported games, and fixing a few bugs here and there. The GeForce Game Ready 545.84 WHQL driver, available now through GeForce Experience or direct download, is slightly different.

In addition to Naraka: Bladepoint and Warhammer: Vermintide 2 getting DLSS 3 support, it also significantly boosts AI acceleration, which means running Large Language Models on Windows PCs with GeForce RTX GPUs has gotten considerably faster.

NVIDIA has added TensorRT acceleration to Stable Diffusion's Web UI for its Generative AI tool to generate images - and for the flagship GeForce RTX 4090, the result is up to 2X the AI performance. Or up to 7X when compared to a top-of-the-line Mac with an Apple M2 Ultra chip.

There is still hope that Intel's Battlemage graphics cards, theoretically due next year, could be a more compelling offering than some rumors have suggested.

According to the latest video from Moore's Law is Dead, the second generation of Arc products might offer up a beefier stack of GPUs than the YouTube leaker has previously asserted.

If you follow Moore's Law is Dead, you'll likely recall he has claimed that Battlemage may well end up as lower-end GPUs only when it arrives in 2024.

Intel's new Arc A580 graphics card is not popular with retailers, if the latest from a well-known YouTube leaker is anything to go by.

Moore's Law is Dead discussed the A580 GPU as part of his latest video, noting that despite some positive buzz around the A580, retailers don't want to stock the card.

A few sources from the retail world are cited by Moore's Law is Dead, and they include an insider at a major outlet who observes that they ordered "almost no A580s" and that maybe some might pop up in late October.

Earlier this year, NVIDIA launched RTX Video Super Resolution for GeForce RTX 30 and 40 Series GPUs, which is like a version of DLSS for video streaming on platforms like YouTube - take a lower-resolution video and use AI to upscale it to your native resolution. Like DLSS, it uses the Tensor Core AI hardware in GeForce RTX graphics cards.

That was back in February, and now, after several more months of training the RTX VSR AI model, NVIDIA is launching RTX Video Super Resolution or RTX VSR version 1.5 - now available as part of the latest GeForce Game Ready Driver. And support is expanding to the entire GeForce RTX family - including the Turing-based GeForce RTX 20 Series.

So, what's new? Alongside improvements to image quality, there's a new option that will de-artifact video played at your display's native resolution.

There are a number of reasons why a graphics card might give up the ghost, if you're unlucky, but this is a new one, with an NVIDIA RTX 4090 appearing to have died thanks to an insect introduced on the assembly line.

As highlighted by Ghost_motley on X (formerly Twitter), the discovery of the very literal hardware bug was made by NorthWestRepair and shared in a YouTube clip (see above).

The RTX 4090 in question would spin up its fans, but was producing no video output, with the monitor remaining blank.

We're all expected to find peace with the fact that high-end enthusiast-class GPUs are massive units. Triple slot thickness, three fans, chunky heatsinks, and 'pipes everywhere' are required to keep the high-end 2023 graphics card from getting so hot that it melts into a pool of lava. At least, that's the idea for a traditional air-cooled card.

But maybe not. Over on Reddit, a GPU enthusiast named TechTaxi took it upon themselves to reduce the physical footprint of a powerful GeForce RTX 4080 and get its size down to a more case-friendly two-slot thickness with two fans for cooling. They took the 3-slot triple fan action of the Gainward GeForce RTX 4080 Phoenix "GS" and put the GPU and RTX 4080 hardware inside the smaller, sleeker GeForce RTX 4070 Ghost.

Regarding physical real estate, the mod takes the GeForce RTX 4080 dimensions of 328.9 x 131.1 x 63.6mm down to an impressive 269.1 x 131.8 x 40.1 mm. Yean, amazing stuff - but what about performance, temps, and overall thermals?

When NVIDIA announced and launched the GeForce RTX 40 Series last year, it also introduced the ground-breaking new DLSS 3 - which offered AI-powered Frame Generation technology to boost performance. The only catch is that the new tech is exclusive to the latest GPU generation because it uses new and updated AI hardware as part of the architecture.

And with that, there has been a long-standing belief in some corners of the tech community that DLSS 3 and Frame Generation are artificially locked to the new GPUs and that there's no reason why it shouldn't also be able to run on GeForce RTX 20 and 30 Series hardware. These, too, feature AI-based Tensor Cores, so why not?

Recently, the discussion surrounding DLSS 3 Frame Generation and older GeForce RTX hardware reignited because it was found that in Portal Prelude RTX - an RTX Remix community mod - GeForce RTX 20 and 30 Series owners could enable Frame Generation. And see their FPS effectively double.

AMD has updated the AMD Software: Adrenalin Edition Technical Preview Driver for AMD Fluid Motion Frames, with Version 23.30.01.02 adding a new and notable update over the weekend. Per the release notes, AMD Fluid Motion Frames, the company's frame generation technology, is similar in functionality to NVIDIA DLSS 3 Frame Generation and can now be "globally enabled on all DirectX 11 and 12 titles."

This is a pretty major update as it allows the driver-level implementation of AMD's frame generation tech to be enabled across all major PC games - with the option to turn AMD Fluid Motion Frames on or off per app. The other major update related to the quality of Fluid Motion Frames, with improvements made to frame pacing "resulting in an overall smoother gameplay experience and improved image quality." Finally, the update also adds HDR support to AMD Fluid Motion Frames.

Frame pacing has been one of the issues for those testing Fluid Motion Frames, alongside having to enable v-sync for the best results. As with NVIDIA DLSS 3, the technology generates frames - a process that adds latency - so AMD recommends using AMD Radeon Anti-Lag and the newer Anti-Lag+. But based on recent reports, you should hold off on enabling the technologies in online multiplayer games like Counter-Strike 2.

Valve has only had its huge release of Counter-Strike 2 in the last couple of weeks, but it's just surfaced that AMD Radeon RX 7000 series graphics card owners enabling AMD's new Anti-Lag+ technology will see themselves VAC banned.

Why is this happening all of a sudden? AMD's new Anti-Lag+ technology has been messing around with DLL files, and that's a big no-no for Valve and why the VAC banning is taking place for Radeon RX 7000 series GPU owners using Anti-Lag+ technology on Counter-Strike 2. But now, it's not just Counter-Strike 2, with Anti-Lag+ technology causing issues for Call of Duty: Moden Warfare and Warzone, as well as Apex Legends.

Anti-cheating technology has evolved over the years, and now has a direct feed into the graphics pipeline, meaning that game developers have had to be stricter with their anti-cheating methods and toolsets used to find any weird things going on with game files. Well, AMD's new Anti-Lag+ tech does just that: messes with game files (in this case, DLL files).