Graphics Cards - Page 25

With the rise of PC cases with panoramic views, alongside the increased size of enthusiast graphics cards in recent years, the idea of vertically mounting a GPU is becoming increasingly common. And when it comes to brackets and riser cables to support the latest PCIe 5.0 cards like the GeForce RTX 50 and Radeon RX 9000 Series, Lian Li's new 'Multi-Directional Vertical GPU Bracket Kit' might just be the ultimate solution.

This all-in-one solution not only supports vertical orientation for GPUs, but it also allows forward and backward tilting, as well as a three-level height adjustment (with 10.16mm increments) for improved airflow for bottom-mounted fans or radiators.

The Multi-Directional Vertical GPU Bracket Kit mounts directly to the motherboard standoffs for secure installation and is built to support all modern GPUs, with its reinforced support arm rated up to 10kg. Lian Li notes that this will prevent GPU sag, while also supporting the largest quad-slot GPUs.

A couple of weeks ago, ASUS and Noctua's latest GPU collaboration launched with the release of the ASUS GeForce RTX 5080 Noctua Edition. Sporting three of the company's second-generation Noctua NF-A12x25 G2 120mm fans, an advanced vapor chamber, and other high-quality thermal bits and pieces, it's reportedly one of the quietest high-end GeForce RTX 50 Series graphics cards on the market.

It's also one of the largest, as putting three Noctua fans on a GPU that you'd typically associate with PC and radiator cooling means that the GeForce RTX 5080 Noctua Edition is a big boy. With dimensions of 385x151x80 mm, it's a rare quad-slot GPU. However, it also ships with a generous +115 MHz increase to the Boost Clock, so you're getting overclocked performance.

As a premium and custom GPU with cooling designed in collaboration with Noctua, if you're wondering why ASUS and Noctua didn't team up to create a custom (and near-silent) GeForce RTX 5090, the answer is simple. Getting access to the RTX 5090's GB202 chip is increasingly difficult due to shortages and the insatiable demand for high-end NVIDIA hardware for AI.

A US retailer is selling a blower-style RTX 5090 graphics card which is a dual-slot effort, but if you want this spin on the Blackwell flagship, it's seriously pricey - and the GPU comes with other caveats, too.

This is the AFOX GeForce RTX 5090 32GB Professional graphics card and it is being sold by HydraCluster Tech Builds for a towering $5,999 - which is supposedly $500 off the retail price (add seasoning now, and lots of it).

Tom's Hardware picked up on VideoCardz originally noticing the new RTX 5090, which as noted is a two-slot affair - but it's anything but compact in its other dimensions, certainly in terms of its beefy width.

NVIDIA has reportedly halted production of its H20 AI chip that is designed specifically for the Chinese market, with NVIDIA sending notifications to Arizona-based Amkor Technology and Samsung Electronics to stop manufacturing.

NVIDIA's AI chips, specifically the H20, are yet again up in the air. The sale of AI chips to China has been nothing short of a fiasco, with NVIDIA initially set to lose a staggering $5.5 billion over trade restrictions implemented on China for high-powered AI chips. These trade restrictions were set to ensure the US remains at the top of the global AI race, and at the time cited national security risks as China would use NVIDIA's GPUs to develop a sophisticated AI that the Chinese military could then utilize.

Only a month later, the H20, which is a heavily modified version of the H100 GPU, was banned from entering China. However, by July, the US government had changed its stance on NVIDIA chips sold to China. NVIDIA was able to secure a license, which would approve exporting the H20 to China, provided that 15% of the revenue from these sales would be directly allocated to the US government.

AMD could change the game with its next-gen Radeon GPU release, renaming the new RDNA 5-based graphics cards to the Radeon PTX series... and there's a (pretty) good reason for that.

Moore's Law is Dead covers the topic in his new video, pondering whether AMD should redo its branding with the new Radeon PTX naming scheme, given the ray tracing and path tracing performance inside of RDNA 5. In his video, MLID says that we could see AMD change up the naming scheme for the next-gen RDNA 5-based Radeon GPUs, with the AT3 GPU called the Radeon PTX 1060 XT, and the AT4 GPU called the Radeon PTX 1050 XT.

You might be thinking what the hell "PTX" is, well... NVIDIA rebranded from GeForce GTX to GeForce RTX for the rise of Ray Tracing (RT + X = RTX) so AMD could do something similar with PTX (PT = Path Tracing + X).

AMD's next-generation RDNA 5-based discrete GPUs have had some very interesting new leaks surface, with more details on the "Alpha Trion 3" and "Alpha Trion 4" otherwise known as AT3 and AT4 in recent leaks.

In his latest video, Moore's Law is Dead discusses some new information he has on AMD's next-gen RDNA 5-based Alpha Trion 3 and Alpha Trion 4 GPUs. Starting with the Alpha Trion 3 "AT3" which is said to have 48 CUs of RDNA 5 with 20MB of L2 cache, either a 384-bit LPDDR6 or 256-bit LPDDR5X memory controller, 8 PCIe Gen5 lanes, and performance somewhere between the desktop GeForce RTX 4070 and Radeon RX 9070 graphics cards.

The Alpha Trion 4 "AT4" GPU is more cut down than AT3, with 24 CUs of RDNA 5 and 10MB of L2 cache, a 128-bit LPDDR5X memory controller, the same 8 PCIe 5.0 lanes, and performance between the desktop RTX 3060 and RTX 4060. Both AT3 and AT4 will have much higher performance in ray tracing and path tracing, with releases in 2027.

We have heard about 12VHPWR connectors melting on NVIDIA's RTX 40 series graphics cards, but never on AMD cards.

That discrepancy was due to AMD never adopting the newer power connector standard with the company's latest generation of graphics cards, as AMD explicitly stated it was sticking with the 2x 8-pin connectors for their reliability. After the fiasco of melting connectors with NVIDIA's high-end RTX 40-series GPUs, this even became a selling point for AMD graphics cards.

However, with the RX 9000 series, some partners, such as ASRock, decided to adopt the 12VHPWR connector, and now we have the first report of a melting 12VHPWR connector on an AMD Radeon graphics card. The user reporting the melting was rocking an ASRock Radeon RX 9070 XT Taichi OC, and judging by the details from the user, there were a few reasons as to why it melted.

With the release of AMD's FidelityFX SDK 2.0, FSR 4 is finally available for all game developers, with integration available via signed DLLs. And to make things super easy for developers creating titles with Unreal Engine 5, this update also includes an Unreal Engine 5 plugin for developers to add FSR 4 support in games.

AMD's FSR 4 is the latest iteration of the company's FidelityFX Super Resolution technology that moves to an AI solution similar to NVIDIA DLSS. As such, it's currently exclusive to AMD's RDNA 4 desktop GPUs, like the Radeon RX 9070 XT, which include advanced AI hardware. Per our reviews and hands-on time with FSR 4, it's safe to say that it presents a massive leap forward in image quality compared to FSR 3, and finally closes the gap between FSR and DLSS.

However, unlike previous versions of FSR, FSR 4 is not officially 'open source' yet, with code available to all developers. Well, with the release of FidelityFX SDK 2.0, it was for a moment, as AMD briefly released the FSR 4 source code before taking it down.

MaxSun is prepping its new Intel Arc Pro B60 Dual 48GB Turbo graphics card for release, featuring two Arc Pro B60 GPUs each packing 24GB of VRAM for a total of 48GB GDDR6. Check it out:

We've been keeping a close eye on this graphics card, as I'm personally a long-time (30 years+) fan of all things GPU, and especially dual-GPU offerings. MaxSun's upcoming Arc Pro B60 Dual 48GB Turbo graphics card fits into a regular PCIe slot that splits into two 8-lane connections, with a GPU assigned to each of those 8 lanes.

The PC detects the card as two GPUs, even though it's physically connected to a single PCIe slot, and unlike other dual-GPU graphics cards, this isn't for gaming. MaxSun's new Arc Pro B60 Dual 48GB Turbo graphics card is for professionals, and more so, AI workloads.

GIGABYTE has just unveiled its new AORUS RTX 5090 AI BOX, its first external graphics solution powered by the flagship NVIDIA GeForce RTX 5090 graphics card. We've got a liquid-cooled RTX 5090 inside, with the super-fast Thunderbolt 5 interface.

Inside, the AORUS RTX 5090 AI BOX has GIGABYTE's in-house AORUS WATERFORCE liquid cooling system, featuring a large copper base plate, 240mm radiator, dual 120mm fans, and an optimized pump. This makes sure that users have stable operation, lower temperatures, and quiet performance while it sits on your desk.

GIGABYTE is providing the super-fast Thunderbolt 5 standard with up to 80Gbps of bandwidth, and 64Gbps of dedicated bandwidth to the PCIe Gen4x4 for the RTX 5090. The new AORUS RTX 5090 AI BOX features a USB Type-C cable with support for up to 100W Power Delivery, so you can charge your laptop (and any other USB-C device) super-fast.

ASUS is celebrating its 30th anniversary with a brand new GeForce RTX 5090 graphics card, with its ambitious new design and limited edition prowess being unveiled at Gamescom 2025.

We caught teasers of ASUS's new GPU a few days ago, following the company saying it was working on something "top secret" in early July. Now that the "top secret" project has been officially unveiled, and the world can bask in the glory of the ROG Matrix GeForce RTX 5090 30th Anniversary Limited Edition. Quite a mouthful of a name.

The card features both 12V-2x6 and GC-HPWR, or BTF connector power options, so users can place the card in any orientation they like. Additionally, users are able to take advantage of both power connectors simultaneously to expand the available power to 800W, which is notably higher than the standard RTX 5090, which is rated for between 575 and 600W.

NVIDIA has clarified that its event later today will not feature the launch of any new hardware - so if you were hoping for something 'Super' then you can forget that idea.

And yes, there probably were a fair few gamers out there who believed that maybe NVIDIA might sneak in a new GPU with its 'GeForce On Community Update' event which is coming later today (at mid-day, US Pacific Time).

In a post on X which was flagged up by VideoCardz, Team Green said that the community update would focus on games and features, and "not hardware", clearly ruling out any possibility that an RTX 5080 Super graphics card might be revealed (or even teased, perhaps).

NVIDIA's unreleased RTX TITAN ADA prototype for the RTX 40 Series Ada Lovelace generation has always felt like more of a prototype than something that would see the light of day. This beast of a GPU, sometimes referred to as the GeForce RTX 4090 Ti, sports a massive quad-slot Founders Edition-style cooler that is reportedly designed to handle a GPU that can draw up to 900W of power.

A new deep dive video into the unreleased RTX TITAN ADA prototype from der8auer gives us another closer look at the GPU, however, this time highlighting its custom power connector. With the TITAN ADA powered by dual 12VHPWR connectors rated at up to 600W each, NVIDIA produced a custom adapter and cable for TITAN ADA that includes six 8-pin PCIe power connectors converted into a single large plug with dual 12VHPWR.

Yeah, one look at the cable and it looks even more over-the-top than the GeForce RTX 5090's four 8-pin connectors converted into a single, updated, 16-pin power cable. This prototype cable for RTX TITAN ADA isn't as flexible as the RTX Blackwell series adapters designed by NVIDIA. Still, it does include safety pins to ensure all cables are plugged in correctly.

The most expensive GeForce RTX 5090 graphics card is now available to purchase, with the ASUS ROG Astral GeForce RTX 5090 Dhahab Edition on sale in Dubai for $9205.

The ASUS ROG Astral GeForce RTX 5090 Dhahab Edition features 5kg of real gold and is now available through a UAE-based retailer. The card launched for $8000 and is now available for purchase through Microless but it'll cost you a pretty penny: $9205.48 to be exact. This price is after a 20% discount, which still makes it more expensive than MSRP.

ASUS first revealed the exotic GPU in China, after exclusively launching it in the Middle East, but Microless will ship it worldwide... at a cost, obviously. The ROG Astral GeForce RTX 5090 Dhahab Edition features a GPU boost clock of up to 2580MHz, and OC clocks of up to 2610MHz, which is identical to the ROG Astral RTX 5090 OC Edition. You're just paying for the super-exotic design, and that 5kg of gold.

MaxSun is about to start shipping its new Intel Arc Pro B60 Dual graphics card, which packs 48GB of GDDR6 memory, in China next week for $1200.

The new MaxSun Arc Pro B60 Dual 48GB is a dual-GPU graphics card powered by the BMG-G21 GPU, yet Intel hasn't officially confirmed the launch of its B60 or B50 graphics cards, but they're expected to hit system integrators (SIs) in the weeks ahead.

This isn't a regular dual-GPU graphics card of years gone, but rather the board is host to two GPUs, each with eight PCIe Gen5 lanes, functioning as two separate graphics adapters.

The new ASUS Dual GeForce RTX 5060 White Edition, available in OC and non-OC variants, sees the company introduce a white variant for its popular, compact, and affordable Dual lineup. The 2.5-slot design sports twin axial-tech fans with 0dB modes, a vented backplate, Dual BIOS support for Performance and Quiet modes, and an all-white version of the standard black Dual shroud.

The ASUS Dual GeForce RTX 5060 White Edition is for those out there putting together or looking for a mainstream gaming GPU for their white colored rig, which is increasingly becoming one of the most popular choices among gamers in the Asia market.

The OC Edition ships with a 2535 MHz overclocked Boost Clock speed that reaches 2565 MHz in OC Mode, with the non-OC standard edition shipping with the 2497 MHz reference Boost Clock speed and a 2527MHz OC Mode.

Arm is bringing DLSS-like AI Super Resolution to mobile gaming with its new and impressive Arm Neural Super Sampling (NSS) technology. It builds on Arm Accuracy Super Resolution (Arm ASR) technology that is already available, improving image fidelity and performance by leveraging neural accelerators on Arm GPUs that are on track to launch sometime in 2026.

Even though the hardware for this is still a year away, the company is all-in on bringing AI-powered rendering to mobile gamers on day one, as NSS already has its own Unreal Engine plugin with PC-based Vulkan emulation and Vulkan ML extensions. The NSS model, which Arm says delivers DLSS 2-like image quality on a low-power mobile device, is compact and open-source.

Arm notes that developers and companies like Enduring Games, Epic Games, NetEase Games, Sumo Digital, Tencent Games, and Traverse Research are already working to add NSS support to games and apps. As a mobile-focused tech, it's designed to reduce GPU workloads by up to 50% while maintaining native-like image fidelity.

AMD recently launched the Radeon RX 9060 non-XT graphics card; however, the fifth RDNA 4 GPU to hit the scene has arrived without a retail release. Instead of being marketed as an off-the-shelf entry-level 1080p gaming GPU, it's currently being sold exclusively as an OEM device for prebuilt systems.

Which is why South Korean creator Technosaurus (very cool name by the way) decided to buy one of these Radeon RX 9060-powered desktop rigs, remove the GPU, and compare the performance to the Radeon RX 9060 XT 8GB, the GeForce RTX 5050, and the GeForce RTX 5060.

The RTX 5050 comparison, which is NVIDIA's mainstream RTX Blackwell GPU, is the most interesting as we see the Radeon RX 9060, on average, beating the entry-level RTX GPU by around 20%. This is a decent lead, and the sort of performance you'll be able to notice when gaming at 1080p.

A week before Gamescom, where we can expect a slew of new gaming-related announcements, ASUS will be launching a new RTX 50-series GPU, and its design is unlike any other 50-series graphics card we have seen so far.

The new ROG Matrix will be released in celebration of the company's 30th anniversary, which we heard about in early July as the company took to its website to tease that it was working on something "top-secret" that's related to an RTX 50-series GPU.

Now we have a name for that project, the ROG Matrix GeForce RTX 5090, which is a level up from the ROG Matrix Platinum GeForce RTX 4090 the company released in 2023. That RTX 4090 Matrix edition featured a 360mm radiator, and it appears from the leaked images of the RTX 5090 Matrix that component has been dropped for standard cooling fans.

NVIDIA's latest GeForce Game Ready 580.97 WHQL driver is here, and this one adds optimizations for two existing games getting a DLSS 4 update. The first is Senua's Saga: Hellblade II Enhanced, which includes a new 'Very High' graphics setting, a new game mode called Dark Rot, developer commentary, and an updated photo mode.

Although the critical and fan response to Senua's Saga: Hellblade II was mixed, there's no denying that the game's lifelike visuals and detailed characters still look stunning. The new Enhanced version of the game also supports DLSS 4 with Multi Frame Generation, which includes the latest 'Transformer' model version of DLSS Super Resolution.

As a visually demanding title, NVIDIA notes that with DLSS 4 and Multi Frame Generation, frame rates are increased by a factor of 3X. And with NVIDIA Reflex, PC latency is reduced by up to 57%. So yeah, if you plan on playing Hellblade II, be sure to grab the latest GeForce Game Ready driver.