Graphics Cards - Page 12

NVIDIA's latest DLSS 5 technology has been nothing if not controversial, with some hailing it as one of the most transformative leaps in real-time graphics, while others dismiss it as little more than 'AI slop'. In a conversation with YouTuber Daniel Owen, NVIDIA's Jacob Freeman has revealed a lot more about the specifics of the technology; details that would otherwise be veiled behind layers of marketing language.

NVIDIA has now confirmed that DLSS 5 takes only a scene's 2D-rendered frame and its motion vectors as input. This stands in contrast to marketing claims that the technology is 'anchored to source 3D content,' a phrase that suggested a deeper understanding of the game engine. In reality, as the model sits at the end of the graphics pipeline, it only sees the 2D frame and remains blind to the 3D geometry of objects.

Likewise, the model lacks access to PBR (Physically Based Rendering) properties provided by the engine. As a result, it is forced to infer what the material is supposed to look like rather than reading these properties directly from the game engine. This forces the model to rely on semantic labeling to identify clusters of pixels as eyes, cheeks, lips, etcetera. If the training data is biased toward perfect faces, the model risks reinterpreting or 'yassifying' a character's face to a generic standard, rather than preserving the developer's original intent.

AMD Software: Adrenalin Edition 26.3.1 is here, and it's a big one for PC gaming as it adds support and optimization to two of the biggest releases for the month, Crimson Desert and Death Stranding 2: On the Beach. It's also a must for RDNA 4 gamers with a Radeon RX 9000 Series graphics card, as it introduces FSR Upscaling 4.1, the latest version of AMD's AI- or ML-based Super Resolution solution.

According to AMD's Jack Huynh over on social media, FSR 4.1 brings "Sharper image quality for ML-based upscaling, delivering finer details and smoother camera motion across all ML-powered AMD FSR games." Compared to the already impressive FSR 4.0, the new FSR 4.1 improves the quality of objects in motion and their overall detail.

As seen in the Crimson Desert video above, AMD Software: Adrenalin Edition 26.3.1 also includes FSR Ray Regeneration 1.1 to enhance ray-traced reflections and global illumination. Two notable updates that help AMD's FSR remain a competitive alternative to NVIDIA DLSS for Radeon RX 9070 XT and Radeon RX 9060 XT owners.

Microsoft recently announced that it plans to address shader compilation stutter by bringing Advanced Shader Delivery for Windows to PC. After NVIDIA confirmed that it will be available for GeForce RTX gamers, Intel has now joined the list of GPU vendors working on enabling this technology for its Arc GPUs.

Intel has rolled out its Graphics Shader Distribution Service in the 32.0.101.8626 WHQL Game On driver, which the company says will improve first-time load times by up to 2x on Arc B-series graphics cards and Intel Core Ultra Series 2 and Series 3 processors with integrated Arc graphics.

These significant performance bumps come from automatically downloading pre-compiled shaders for your games, so you don't have to stare at your screen each time the game compiles shaders. There is cloud infrastructure in the background doing the heavy lifting, but for gamers, it all comes down to one thing: reduced stuttering on first launch.

NVIDIA has updated its data center roadmap for its GPU and CPU lineup at GTC 2026, with the company focused on updating its AI-focused GPU architecture every year. With NVIDIA set to roll out its Vera Rubin architecture and platform this year, which includes the Vera CPU and Rubin GPU, we also got details on its next major GPU architecture, called Feynman, set to arrive in 2028.

The successor to Rubin and 2027's Rubin Ultra, Feynman, will represent a major step for NVIDIA by introducing advanced 3D stacking technology to increase density and improve communication among the various hardware components that power the next generation of AI. In addition, Feynman will be paired with the new Rosa CPU for integrated systems, building on the foundation of Grace in the current Blackwell era.

NVIDIA's roadmap, with Feynman coming in 2028, also lists "custom HBM," which would most likely refer to the next memory evolution beyond HBM4 and HBM4e. We'll have to wait for more details on this new high-bandwidth memory solution. Still, it's safe to assume that its bandwidth and performance will be tailored for increasingly more complex AI models and systems where memory bandwidth and capacity are key.

At NVIDIA GTC 2026, the company has expanded its server lineup with the new efficient single-slot RTX PRO 4500 Blackwell Server Edition GPU. NVIDIA describes it as an "energy-efficient multi-workload accelerator designed to deliver breakthrough performance across a broad range of enterprise workloads," which, of course, includes AI inference as well as other high-end computing tasks, such as video processing.

With 10,496 CUDA Cores, which is slightly lower than the number featured in the gaming-class GeForce RTX 5080 GPU featuring the same GB203 chip, it's also packed with 32GB of GDDR7 memory on a 256-bit interface to deliver memory bandwidth of 800 GB/s. Power consumption-wise, it lives up to its efficient label with the single-slot passively cooled GPU drawing up to 165W.

Naturally, the single-slot form factor is designed for servers and data center racks with multi-GPU setups, where a dozen GPUs could be installed in a single system. The RTX PRO 4500 Blackwell Server Edition GPU is similar, spec-wise, to the dual-slot actively cooled RTX PRO 4500 Blackwell, albeit with a lower power rating and slightly reduced memory bandwidth.

Since the debut of NVIDIA's "25-year graphics breakthrough," the response has been resoundingly negative. Gamers are complaining that the technology is adding an AI beauty filter to their favorite characters, ruining the game's original art style and calling it "AI Slop."

Jensen Huang, founder and CEO of NVIDIA, has come forward to address these remarks, immediately stating, "They're completely wrong."

NVIDIA had previously suggested that developers would be able to tune the overall intensity of how much DLSS 5 changes. Huang reiterated once again that developers are in full, direct control of the technology. He also added that DLSS 5 is different from generative AI, calling it "content control generative AI" that adds generative capability to the game's existing geometry and "doesn't change the artistic control."

NVIDIA has unveiled the new Vera Rubin AI platform, specifically designed to advance AI into its next phase of reasoning, automation, agentic-based workloads, and continuous responses. However, not all that power will be allocated to the numerous AI server farms around the world; some of it is destined for off-planet use.

During NVIDIA's GTC 2026 presentation, company CEO Jensen Huang unveiled the Vera Rubin Space Module, which Huang says will deliver up to 25x times more AI compute than Hopper-generation H100. According to the announcement, six commercial space companies have already adopted the new Vera Rubin Space Module, which NVIDIA says is specifically designed for orbital data centers that are running LLMs in space.

Huang explained during the keynote that NVIDIA is seeking a way to expand its AI computing stack beyond Earth, leading to the creation of "AI factories in space". The Vera Rubin Space Module is designed to solve a current problem in data transfer in space: satellites capture large swaths of data and then have to send it back to Earth for processing. That process introduces latency and bandwidth requirements, but with the Vera Rubin Space Module, the idea is that inference compute is much closer to where the data is being generated, reducing latency, processing time, etc.

In a new report in Taiwan's United Daily News publication (auto-translated), MSI's General Manager Huang Jinqing has warned investors in a recent briefing that 2026 is shaping up to be the "most severe year since the company was founded." Naturally, this is due to current DRAM, storage, and even GPU supply and pricing issues driven by demand for all things AI and AI infrastructure.

And when it comes to graphics cards, of which MSI is a key GeForce RTX partner, the company estimates a supply gap of around 20%, meaning NVIDIA isn't able to provide the stock and GPUs needed to maintain the broader consumer PC market. Not only that, but MSI also predicts this will lead to the PC market shrinking by 10 to 20% in 2026, a larger figure than the one we recently reported from the International Data Corporation (IDC).

When it comes to the company's plans to weather the storm, so to speak, in 2026, it will focus on the mid-range and high-end GeForce RTX market, with cards like the GeForce RTX 5060 and RTX 5070, and the price of its graphics cards will increase by anywhere between 15 and 30%. The goal is to increase profit and profit margins as the overall volume decreases.

Saying one thing is a million times more something than another thing is often hyperbole, but during a recent GDC 2026 presentation, NVIDIA's John Spitzer said exactly that when it comes to path tracing performance on future GeForce RTX graphics cards. Of course, this comes with a caveat: performance compared to NVIDIA's pre-RTX Pascal-era GeForce GTX graphics cards that lacked dedicated ray-tracing and AI hardware. Yes, it's all thanks to AI.

Real-time path tracing or full ray tracing is so demanding on GPU hardware that it's only possible thanks to a wide range of AI-powered rendering technologies such as DLSS Super Resolution, Frame Generation, RTX Mega Geometry, and more. As these features are available on the current RTX Blackwell-powered GeForce RTX 50 Series, NVIDIA says it has already achieved 10,000X faster path-tracing performance than in the Pascal era.

"If you look at the performance there with just a software RT core to today, where we have fourth-generation RT cores, we have third-generation Tensor cores, we have DLSS 4.5, which is able to infer 23 out of 24 pixels rendered," NVIDIA VP of Developer & Performance Technology, John Spitzer, said. "These are multiplicative, that you can multiply them all together to get a scaling factor that, combined with the algorithm, eventually gave a 10,000 times that we've improved the performance over the last 10 years."

COLORFUL has announced a new line-up of GeForce RTX 50 Series graphics cards, powered by the latest RTX Blackwell architecture. The new iGame GeForce RTX 50 Ultra Series blends a similar hip-hop-inspired design and aesthetic to last year's Ultra lineup, albeit in a darker black colorway that highlights the purple flourishes.

Visually, they look impressive, with a gradient of purples and neons in the holographic finish and RGB graffiti on the side. With a large heatsink and triple fans for cooling, this new lineup covers all models from the GeForce RTX 5060 up to the GeForce RTX 5070 Ti. All options ship with an out-of-the-box OC Mode, with the most powerful model including an iGame GeForce RTX 5070 Ti Ultra OC SFF variant for compact builds.

In addition, COLORFUL has announced the new iGame GeForce RTX 50 Ultra DUO Series, featuring a similarly stylish design with compact dual-fan cooling. COLORFUL is releasing iGame GeForce RTX 5060 Ultra DUO and iGame GeForce RTX 5060 Ti Ultra DUO models, including RTX 5060 Ti variants with 8GB or 16GB of VRAM.

Playing a modern PC game, we have all grown accustomed to the 'shader compilation' process, which can take seconds or several minutes to complete. This process, designed to minimize stutter that occurs when new shaders are compiled in real time, also increases the time it takes to get into the action. And it's not always a one-and-done thing, as changes to settings and game updates can restart the shader compilation process.

At GDC 2026, Microsoft announced that it's bringing Advanced Shader Delivery for Windows to PC gamers, with NVIDIA confirming that it will be available for GeForce RTX gamers later this year. It's an ambitious update that is rolling out via the latest AgilitySDK 1.619 release, where developers can generate a "state object database (SODB) file" and then use an "offline compiler to compile the state objects into a precompiled shader database (PSDB)."

Although that's the technical summary of what's involved, the good news for gamers is that pre-compiled shaders for their specific GPU model will be downloaded as part of installing the full game, so the process will be automated and seamless. And for developers, there's also a new Stats API that can be used for optimization and seeing how the game runs in relation to the "shader cache hit rate."

Thermal Grizzly's WireView Pro II is designed not only to measure the power consumption of modern graphics cards or GPUs that use the new 12V-2x6 connector, like NVIDIA's flagship GeForce RTX 5090, but also to prevent potential damage. With per-pin current monitoring, it can detect and alert users of any potential load issues.

The WireView Pro II is an enhanced version of the company's original design and features a CNC-machined aluminum housing with active cooling via a 30-mm fan and 90-degree cable routing. It sports a 320x170-pixel TFT-IPS color display for real-time monitoring of critical measurements such as total power, highest-current pin, voltage, and hottest sensor. And with safety being a key feature of the WireView Pro II, there are both audible and visual warnings if limits are exceeded.

Plus, there's automatic data logging via the included USB-C-to-USB-2.0 cable, making it a fantastic tool for overclockers and those with high-end systems looking to keep tabs on GPU power usage. And with the rise of white-colored PC builds and components, Thermal Grizzly has announced it's launching a WireView Pro II White Edition, available to pre-order via the company's official storefront.

The annual Game Developers Conference (GDC) is wrapping up. Although the event is focused on game developers and networking for their projects, it's also home to game technology companies and hardware makers who discuss what's on the horizon, what's next, and new technologies. At GDC 2026, Microsoft began its first public deep dive into its next-generation Project Helix console that will play both Xbox and PC games.

For NVIDIA and the GeForce RTX team, we got a release date for DLSS 4.5's impressive Dynamic Frame Generation technology, first looks at a suite of new titles with RTX-powered path tracing like 007 First Light and Control Resonant, new GeForce NOW cloud-testing tools, and ComfyUI AI video generation for developers. Plus, it's bringing Microsoft's Advanced Shader Delivery (ASD) to GeForce RTX users later this year.

For AMD, specifically Radeon, RDNA 4, and the recent launch of FSR Redstone, there was nothing to confirm or announce. Granted, AMD announced that it was developing its own version of AI-powered Multi Frame Generation, but this was for Project Helix and its suite of FSR Diamond technologies - with no word if it's coming to Radeon gamers on PC.

This year, the iconic GeForce 3, introduced in February 2001, is celebrating its 25th anniversary. More than just an interactive update, the GeForce 3 introduced programmable pixel and vertex shaders, paving the way for modern PC and console gaming. Some of the PC games that defined the GPU and that generation include id Software's Doom 3, with its groundbreaking per-pixel lighting, and Bethesda's The Elder Scrolls III: Morrowind, which featured incredible water effects for its time.

And with that, taking a break from preparing for his big NVIDIA GTC 2026 opening keynote, NVIDIA CEO Jensen Huang sat down with senior members of the GeForce team to discuss the importance of the GeForce 3 and its legacy. Not only that, but the GeForce 3 was instrumental in transforming NVIDIA into a computing company and the global leader in AI.

"Every game looked the same, looked heavily filtered, everything was bilinear and trilinear filtered," Jensen Huang says of the era before the GeForce 3 arrived in 2001. "We felt that games weren't like CAD; games are a medium for artistic expression, and if you look at all these different games, we wanted them all to look different."

NVIDIA has confirmed that the RTX PRO 5000 Blackwell will ship with 160 ROPs, not the 176 ROPs previously reported.

It appears the previous reporting was based on unlocked specification entries in the database, rather than NVIDIA's locked specification sheet for the product. The discrepancy between the previously reported figure and the official figure arose after a Reddit user noticed that his 48GB variant displayed 160 ROPs, despite reports citing that it would ship with 176 ROPs. In a statement provided to TechPowerUp, NVIDIA explained that a bug in GPU-Z is the cause of the discrepancy and that the new workstation card officially ships with 160 ROPs.

As for the other specifications, NVIDIA reveals in its official materials that the new RTX PRO 5000 Blackwell utilizes the latest Blackwell architecture, and sports 14,080 CUDA cores, 48 GB or 72 GB of GDDR7 ECC memory, 1,344 GB/s of bandwidth, a 512-bit memory interface, PCIe 5.0 x 16, and 300W of total power. Furthermore, NVIDIA states that the card comes with a Graphics Processing Cluster of 10, which verifies the 160 ROPs figure, as 1 GPC equals 16 ROPs.

GeForce Game Ready Driver 595.79 WHQL is now available for GeForce gamers, and it brings day-one support for a big PC game release on March 19, 2026. The highly anticipated open-world action-adventure Crimson Desert from Pearly Abyss and legendary game creator Hideo Kojima's latest opus, Death Stranding 2: On the Beach.

Both games are set to debut with full DLSS 4 support with Multi Frame Generation for GeForce RTX 50 Series owners, with access to the new and improved DLSS 4.5 Super Resolution powered by NVIDIA's powerful new Transformer model via the NVIDIA App. And when it comes to the NVIDIA App, it's getting a major 'Beta' update on March 31.

This update will add DLSS Override support for the new DLSS 4.5 Dynamic Multi Frame Generation update, which is designed to intelligently max out your display's refresh rate and includes a new 6X mode. This update also includes an updated DLSS 4.5 Frame Generation for GeForce RTX 40 and RTX 50 Series gamers, improving image quality and performance.

NVIDIA's GeForce RTX 30 Series, built using a custom 8nm process from Samsung, rather than the custom TSMC process for the GeForce RTX 40 and RTX 50 Series, is set to return to production at the Samsung Electronics Pyeongtaek Semiconductor Factory. According to a new report by the Korean publication Hankyung, Samsung is prepping to recommence production of the 8nm GeForce RTX 3060, some two years after the last batch of RTX 3060 chips were manufactured.

The report doesn't set a timeline for production, other than that it'll happen "soon." With the initial media and insider reaction focused on NVIDIA re-introducing the GeForce RTX 3060 as a response to the current memory crisis and chip shortages, the report claims that the RTX 3060's return could be primarily for the Chinese market.

Due to restrictions on high-performance graphics cards sold in China, older GPUs like the GeForce RTX 3060 represent a way out for providing a cost-effective AI and PC gaming solution in the Chinese market. Of course, NVIDIA CEO Jensen Huang did say "it's a good idea to bring back old GPUs" at CES 2026 as a way to address current shortages, so this move could reflect a broader global strategy for 2026 and beyond.

Jon Peddie Research has posted its latest report on the fourth-quarter 2025 global PC graphics card market. Even though AIB (add-in-board) shipments decreased from the previous quarter due to the current memory crisis, they were up 36% year-over-year for the same period. When it comes to market share for graphics cards "largely supported by gamers," NVIDIA's share rose notably.

According to Q4 2025 data, NVIDIA's market share of total graphics card shipments reached a new high of 94%, up 10 percentage points from Q4 2024. As 2025 saw the release of the new GeForce RTX 50 Series from NVIDIA and the Radeon RX 9000 Series from AMD, this data shows NVIDIA pulling ahead, with AMD's GPU shipment market share dropping to 5% from 7% in the previous quarter and 15% in the same period the year before.

Intel's market share for discrete Arc GPUs remained flat at 1%. However, even though there are GPUs out there, Jon Peddie Research notes that the current high memory prices, driven by the AI boom and US government-imposed tariffs, are "killing the AIB market." In addition, more affordable laptops and notebooks with APUs featuring integrated GPUs alongside CPUs are putting additional pressure on the growing GPU market.

Another rumor has surfaced claiming that NVIDIA is planning to bring back the RTX 3060 in order to help shore up the GeForce line-up, which is facing stock difficulties due to the RAM crisis.

VideoCardz reports that there's a rumor from the Board Channels forum over in China which backs up previous speculation the RTX 3060 GPU could be resurrected in the first quarter of 2026.

This new nugget of gossip from the supply chain claims that RTX 3060s from "various brands" will gradually start to appear from between March 10 to 20. In other words, within the next couple of weeks.

NVIDIA has announced a new hotfix for GeForce Display Drivers after it was discovered that its latest 595.71 WHQL Game Ready driver was implementing voltage caps on GeForce RTX 50-series GPUs.

The timeline of events was as follows. On February 26, NVIDIA released the NVIDIA GeForce Game Ready 595.59 WHQL driver, which was promptly followed by the 595.71 driver. Following that release, reports began surfacing of some RTX 50-series owners experiencing new voltage ceilings, resulting in reduced overall game performance. On March 4, NVIDIA announced a new hotfix that is currently ready to download.

Additionally, NVIDIA acknowledged the Game Ready driver for Resident Evil: Requiem contained many bugs players would have noticed within the game, such as white glowing lights/dots appearing when Subsurface Scattering is enabled. There were also several instances of crashing with Bethesda's Starfield. Those problems have now been all fixed in NVIDIA's latest GeForce driver update, which is currently available to download.