Graphics Cards - Page 22

AMD has scored one point over NVIDIA after company CEO Lisa Su confirmed that its next-generation MI450 AI accelerator will be built on a 2nm silicon process, beating NVIDIA's upcoming Vera Rubin architecture built on 3nm silicon.

Su confirmed the nanometer size in a recent interview with Yahoo Finance, where she was discussing the recent partnership announcement with ChatGPT-creator OpenAI. For those who don't know, OpenAI and AMD recently announced a new partnership that involves AMD providing OpenAI with its next-generation GPUs until it has completed the agreed-upon 1 gigawatt of power. In exchange, AMD will provide OpenAI with approximately 10% of its company.

During the interview, Su revealed that AMD's upcoming AI accelerator, scheduled for release in 2026, will utilize 2nm technology, which means it will leapfrog over 3nm. Notably, NVIDIA's upcoming next-generation AI GPU architecture, Vera Rubin, will utilize 3nm silicon, giving AMD a competitive edge over Team Green in this regard. Why does the nm matter? Typically, a reduction in silicon process size leads to performance gains in the form of increased efficiency and raw performance.

The latest GeForce RTX 50 Series game bundle is here, with NVIDIA offering a free Deluxe Edition copy of the highly anticipated multiplayer extraction shooter ARC Raiders with eligible purchases of desktop GeForce RTX 50 Series GPUs and laptops with the latest GeForce RTX 50 Series Laptop GPUs.

ARC Raiders is set to launch on October 30 for PC and consoles, and looks to expand on the extraction shooter genre popularized by Escape from Tarkov and Hunt: Showdown by blending exploration and other elements from the adventure genre. It's set in a future where Earth has been overrun by deadly machines, with human Raiders (i.e., other players) also presenting a threat.

The PC version of the game is optimized to run on GeForce RTX hardware, with day-one support for DLSS 4 with Multi Frame Generation, NVIDIA Reflex, and cutting-edge ray-traced global illumination lighting effects. NVIDIA's DLSS 4 trailer for ARC Raiders shows performance hitting an impressive 400+ FPS in 4K, presumably running on a GeForce RTX 5090.

GIGABYTE's new AORUS RTX 5090 AI BOX is now out in the wild, and has been tested by Notebookcheck with some rather interesting results: it's up to 27% slower than a desktop RTX 5090. Let's dive into it.

Firstly, the new AORUS RTX 5090 AI BOX is a new external GPU solution running over the new Thunderbolt 5 standard, and while in theory the new TB5 standard is one of the fastest, this solution ends up with the same bandwidth as OCulink. Thunderbolt 5 does have some things going for it over OCulink, where it can transmit power (to charge your laptop and other devices over the TB5 port), and it can be daisy-chained and used for external monitors.

The new AORUS RTX 5090 AI BOX is available right now and costs $2999, matching the price of the AORUS RTX 5090 Master White Edition graphics card (at Newegg) so you're effectively getting a cool Thunderbolt 5 external enclosure for free. On the same day that GIGABYTE launched its new AORUS RTX 5090 AI BOX, Notebookcheck published their review and it has some fantastic performance information and testing to geek out over.

MaxSun has another announcement today, with new single-slot, liquid-cooled Intel Arc Pro B60 48GB graphics cards, which allow for huge 7-way GPU installations. Check it out:

MaxSun once again teamed with Japanese hardware manufacturer Abee to use its liquid cooling solution on its Intel Arc Pro B60 48GB card, with the normal Turbo Edition featuring a dual-slot cooling solution, the new Arc Pro B60 48G Liquid Edition is a nice-and-tight single-slot GPU solution.

The GPUs themselves have connectors for the liquid loop on the right, making sure the complete cooling setup can be installed easily inside of the Abee Designer C700W PC chassis. This chassis is the same one MaxSun is using for its new AI workstation -- which has 4 x Arc Pro B60 48GB Turbo GPUs installed -- but the new single-slot cards are an even better option for a truly insane workstation PC.

The high-end NVIDIA GeForce RTX 50 series cards like the flagship RTX 5090 push that new 12V-2x6 power connector to its limit, flowing 575W+ of power through it, so adding an additional layer of protection could be a good move, and that's why this new $79 device can do.

Aqua Computer has introduced the new AMPINEL, a small dedicated hardware device for the 12V-2x6 power connector that both measures power, but also balances current flow and is capable of shutting down the GPU if those limits are exceeded. AMPINEL will continuously monitor all 6 lines of the connector, so that if the current rises on a single pin by above 7.5A, the device kicks in, redistributing load in real-time using a hybrid analog/digital control system (this is the hardware solution itself).

The new AMPINEL uses a six-layer PCB with thick copper traces for minimal loss, paired with black-anodized aluminum housing that double as a heatsink, while the connector itself is encapsulated for a stronger mechanical fit on your high-end GPU.

China has some very interesting custom and modded graphics cards, with a new one appearing: an NVIDIA GeForce RTX 4080 SUPER graphics card with 32GB of GDDR6X, double the VRAM as the RTX 4080 SUPER ships with 16GB of VRAM.

These custom graphics cards are not official in any form, and they're not even made by legitimate manufacturers, with the sole purpose of their creation to unlock additional memory on graphics cards, when they're not used for gaming. This expanded VRAM capacity allows Chinese "startups" and AI farms to make cheaper RTX series GPUs for AI training, with more VRAM than their gaming counterparts.

NVIDIA released the GeForce RTX 4080 SUPER in January 2024 launching with 16GB of GDDR6X memory, the same VRAM capacity as the RTX 4080 non-SUPER, but the newer RTX 4080 SUPER uses faster GDDR6X memory chips with 23Gbps on the RTX 4080 SUPER and slightly slower 22.4Gbps G6X chips on the RTX 4080 non-SUPER. However, the new SUPER card has more CUDA cores and an increased TDP over the non-SUPER card.

Fengpo is a brand you've probably never heard of, but the Chinese company has just released a new GeForce RTX 3080 graphics card that rocks a larger 20GB of VRAM (RTX 3080 has 10GB) using a cooler from PNY's higher-end GeForce RTX 4090 graphics card. Check it out:

We reported all the way back in January 2021 that GeForce RTX 3080 engineering samples were discovered with 20GB of VRAM, and since then we've seen reports of RTX 3080 cards with 20GB surfacing... and now, Fengpo has officially launched a GeForce RTX 3080 graphics card with 20GB of VRAM.

In a new post on X by Uniko's Hardware, we're learning that Chinese content creator "GPU Girl" has benchmarked the RTX 3080 20GB graphics card, comparing it against the RTX 5060 Ti 16GB card in LM Studio.

Razer's new Core X V2 external Thunderbolt 5-powered GPU enclosure is now available on Amazon for $349.99, with a full Thunderbolt 5 port offering up to 64Gbps of bandwidth, all through a single cable into your laptop.

The new Razer Core X V2 external GPU enclosure might have a Thunderbolt 5 port, but you're limited to just PCIe 4.0 x4 speeds that are capped to 64Gbps, while the full bandwidth sits at 120Gbps in one direction, and 80Gbps bidirectionally.

But there are other improvements that Thunderbolt 5 delivers on this external GPU enclosure, like delivering up to 140W of power to the connected device that sees the data transfer handled, as well as charging for laptops, smartphones, and gaming handhelds.

The Steam Hardware & Software Survey results for September 2025 are in, and in the discrete GPU space, NVIDIA continues to dominate with 74% market share compared to just under 18% for AMD and around 8% for Intel. However, a significant portion of AMD and Intel's market share comes from integrated GPUs, as seen in AMD's latest Ryzen AI and Ryzen Z2 Extreme chips for laptops and gaming handhelds.

Examining the top 25 discrete graphics cards used by PC gamers on Steam reveals that they are all GeForce models, led by mainstream options such as the GeForce RTX 3060 and RTX 4060, entry-level models like the GeForce RTX 3050, and mid-range cards like the GeForce RTX 3070 and RTX 4070.

With NVIDIA's GeForce RTX 50 Series launching this year, led by the new flagship GeForce RTX 5090, every model is now listed in the Steam Hardware Survey, apart from the GeForce RTX 5050. Out of the full lineup, far and away the most popular card looks to be the GeForce RTX 5070, which has seen its market share grow to 1.62%.

GeForce Game Ready Driver 581.42 - WHQL is here, and it adds day one support for what is arguably one of the biggest PC game launches of the year, Battlefield 6. For GeForce RTX owners planning to pick up the latest installment in the acclaimed multiplayer shooter series, this is a must-install, as the game is set to utilize a wide range of RTX technology support.

Launching on October 10, Battlefield 6 is set to arrive with DLSS 4 with Multi Frame Generation and NVIDIA Reflex support on day one, as well as DLSS Frame Generation, DLSS Super Resolution, and DLAA support. "Using these technologies, all GeForce RTX gamers can accelerate and enhance their experience, making gameplay smoother and more responsive," NVIDIA writes.

In addition to the game's multiplayer modes, which led to the recent Battlefield 6 Open Beta becoming the largest in the franchise's history, the game is also set to include a single-player campaign. In addition to Battlefield 6, the latest GeForce Game Ready Driver also adds support for FBC: Firebreak's major Breakpoint update, fixes a few bugs, and expands Project G-Assist controls for laptops.

MSI unveiled a brand-new GPU design at the recent Tokyo Game Show 2025 event, featuring a 16-pin power connector that is hidden inside a magnetically detachable back panel.

The design is reminiscent of SAPPHIRE's Radeon RX 9070 XT NITRO+ graphics card that we reviewed earlier this year. Not only is it one of a handful of RDNA 4 GPUs to adopt the new 16-pin power connector for power delivery, but it also introduced an innovative way of hiding the power connector inside a removable panel on the backplate.

The new MSI design, as seen here via Japanese YouTube creator @harucan5492, is for a new GeForce RTX 5070 Ti Ventus 3X model featuring a hidden power connector and a removable backplate. It's also a White Edition GPU, which is one of the most popular colors for graphics cards in the Asia region.

AMD Fluid Motion Frames (AFMF) is the company's driver-based Frame Generation technology that adds the option to generate frames and boost performance/smoothness in thousands of games. As a driver-based tool that is a part of the Radeon Adrenaline Software's HYPR-RX suite, AFMF 2 launched last year, bringing improvements to latency, image quality, and stability.

The technology also pairs with AMD Radeon Anti-Lag or Radeon Anti-Lag 2 to reduce latency, with AMD touting the technology as a viable solution for boosting performance on mobile devices with lower-powered integrated Radeon graphics.

AMD has continued to update the technology, with AFMF 2.1 launching earlier this year, receiving generally positive feedback due to its further improvements in latency and image quality by reducing ghosting. And now, thanks to some driver sleuths over at the Guru3D forums, The Creator and ecffg2010, it appears that AMD is preparing to launch AFMF 3, also known as AMD Fluid Motion Frames 3.

In recent times, the idea of a new GPU being available for its launch MSRP (the recommended price point set by companies like NVIDIA or AMD) has become something of a myth for most people. With the launch of the new GeForce RTX 50 Series and Radeon RX 9000 Series earlier this year, although some MSRP models have been available, picking up a brand-new GPU for less than a 10 to 30% markup over the suggested retail price has been challenging.

As we head toward the holiday season, it's now possible to pick up a brand-new GeForce RTX 50 Series graphics card at or even below the MSRP price point. These price drops aren't limited to selected models either, as you can easily find GeForce RTX 5060, RTX 5060 Ti, RTX 5070, RTX 5070 Ti, and even RTX 5080 graphics cards for a decent price. At a time when the cost of all PC, console, and gaming hardware is on the rise due to the current economic climate and AI boom, this is both surprising and welcome.

Examining retailers like Newegg and Amazon, let's take a look at some of the GeForce RTX 50 graphics cards currently available at a price point that better reflects the launch-day MSRP, starting with the GeForce RTX 5060.

With several credible leaks from various sources in recent months, right now it's only a matter of time before NVIDIA officially announces the GeForce RTX 50 SUPER Series refresh. For those that need a refresher, the lineup includes three models: the GeForce RTX 5070 SUPER, RTX 5070 Ti SUPER, and RTX 5080 SUPER.

The key takeaway is that these GPUs are expected to offer a substantial 50% increase in VRAM capacity. Alongside increases to clock speeds and some modest specification increases for the GeForce RTX 5070 SUPER 18GB GPU, overall performance boosts compared to the non-SUPER variants will likely be modest, and in the 5-10% region. Unless, of course, you're looking at a workload that takes advantage of the massive boost to VRAM capacity.

The only real question right now is when we can expect to see them. The latest rumor, which originated from Benchlife (via Videocardz), suggests that the GeForce RTX 50 SUPER Series may arrive later in 2026 than initially expected.

ASRock and Capcom have partnered to release a custom Monster Hunter-themed gaming GPU with the arrival of the new Radeon RX 9070 XT Monster Hunter Wilds Edition 16GB. The physical design looks to be based on the company's impressive Steel Legend line-up, albeit presented in a very different light, thanks to the "Arkveld" design inspired by one of the game series' most iconic monsters.

With a mix of silver and blue tones, it also includes a custom metal backplate with both Arkveld and the Monster Hunter Wilds logo joining the 'AMD Radeon' branding. Like ASRock's Steel Legend design, there's a dedicated ARGB switch, with premium cooling that includes high-quality striped ring fans, a nickel-plated copper base for maximum GPU contact, and a phase-change thermal pad.

Outside of the GPU, the Radeon RX 9070 XT Monster Hunter Wilds Edition 16GB ships with a limited edition "Arkveld" packaging and a custom Monster Hunter-themed version of ASRock's POLYCHROME SYNC software. Interestingly, this custom GPU variant isn't an OC model as it ships with the Radeon RX 9070 XT's reference Boost Clock speed of 2970 MHz.

A gamer has purchased an NVIDIA GeForce RTX 5080 graphics card on Amazon for $1000... but received an actual BRICK instead.

Reddit user "GlassHistorical5303" posted that he ordered a PNY GeForce RTX 5080 from Amazon for around $1000, but got a physical brick instead, safely wrapped in an anti-static bag of course. His Reddit post caused quite the stir, questioning Amazon's retail policies, with the retail giant refunding his order.

There is a particular reason why the user received a brick versus the RTX 5080 he ordered, as it's the aftermath of Amazon's internal FBA logistics. Say you've ordered an RTX 5080 instead of the Redditor, Amazon would search for a totally different seller with the same stock and then ship out the product... all without checking the contents, apart from the weight of the product and the desired packaging.

Seasonic has updated its PSU calculator with two unreleased graphics cards: NVIDIA's purported GeForce RTX 5070 Ti SUPER and RTX 5070 SUPER.

In the last few months we've been hearing rumors that NVIDIA was working on its beefed-up RTX 50 SUPER series graphics cards, with the RTX 5080 SUPER, RTX 5070 Ti SUPER, and RTX 5070 SUPER... and now Seasonic is getting ahead of the game with an upgrade to its PSU calculator.

The update to the drop down list on Seasonic's PSU calculator was spotted by VideoCardz, with the RTX 5070 Ti SUPER and RTX 5070 SUPER now included on the list, even though they're not out yet. The PSU calculator lists the RTX 5070 Ti SUPER with a 350W TDP, and the RTX 5070 SUPER with a 275W TDP. This is compared to the RTX 5070 Ti with a 300W TDP and the RTX 5070 with a 250W TDP.

If you're in the market for a high-end graphics card, one of Walmart's latest deals in the US may tempt you, namely a 7% reduction on NVIDIA's RTX 5080.

This deal, spotted by Tom's Hardware, is on the PNY GeForce RTX 5080 OC graphics card, which is down from the MSRP of $999 to $929, a saving of $70.

Obviously I would hesitate to call this affordable - there's nothing affordable about top-tier GPUs these days - but it is at least more affordable than this graphics card has previously been.

NVIDIA's RTX 5060 Ti is a tempting pick out of the Blackwell line-up, in the 16GB flavor of course - and we've seen a new low in pricing for this model.

Tom's Hardware picked up on Walmart in the US offering the PNY RTX 5060 Ti Epic-X variant - a compact model that's suitable for a small form-factor PC - at a hefty discount.

You can buy this RTX 5060 Ti for $379 at the time of writing, which is considerably less than the launch price - the original MSRP was $429. Indeed, it's the lowest we've seen this graphics card at, equalling the launch price you'd pay for the 8GB model of the RTX 5060 Ti (if you were unfortunate enough to grab one of those).

Micron has teased that it's well-positioned with its GDDR7 memory, which is designed to deliver ultra-fast performance with pin speeds exceeding 40Gbps.

NVIDIA's current GeForce RTX 50 series "Blackwell" gaming GPUs ship with GDDR7 memory with speeds of 28Gbps to 30Gbps, but in its recent earnings call, Micron confirmed it has 40Gbps+ GDDR7 memory ready. The company has been working in close collaboration with NVIDIA on pioneering the adoption of LPDRAM for servers, and now that NVIDIA's launched its LPDRAM in their GB (Grace Blackwell) AI server family, Micron has been the sole supplier of LPDRAM in the data center.

Micron President & CEO Sanjay Mehrotra said: "In close collaboration with NVIDIA, Micron has pioneered the adoption of LPDRAM for servers, and since NVIDIA's launch of LPDRAM in their GB product family, Micron has been the sole supplier of LPDRAM in the data center. In addition to our leadership in HBM and LP5, Micron is also well-positioned with our GDDR7 products, which are designed to deliver ultra-fast performance with pin speeds exceeding 40 Gbps, along with best-in-class power efficiency to address needs of certain future AI systems".