Graphics Cards - Page 9

According to a new leak and report over at VideoCardz, citing leaked documents and information from AMD, the company is planning to launch a new entry-level RDNA 4 graphics card, the Radeon RX 9050. According to the information, this would use the full Navi 44 GPU in the Radeon RX 9060 XT, with the main differences being VRAM capacity and lower clock speeds.

The specs for the Radeon RX 9050, which are still to be confirmed, list the GPU as featuring 8GB of GDDR6 memory on a 128-bit bus at 18.0 Gbps. This would deliver memory bandwidth of 288 GB/s, matching the baseline Radeon RX 9060, while coming in lower than the Radeon RX 9060 XT's 320 GB/s.

As for the clock speeds, the 2,048 Stream Processors, which match the Radeon RX 9060 XT, feature a Game Clock of 1,920 MHz and a Boost Clock of 2,600 MHz. This is notably lower than the 3,130 MHz reference Boost Clock speed of the Radeon RX 9060 XT, by 530 MHz, or around 17%. The difference is larger when looking at the Game Clock speeds, suggesting that the Radeon RX 9050 would ship with a lower power draw than the 9060 XT 8GB's TDP rating of 150W.

According to a new report from retailers and insiders (via VideoCardz), PC gamers in China can now pick up an RDNA 4 gaming GPU like the Radeon RX 9060 XT or the Radeon RX 9070 XT for below MSRP, or at the suggested launch price in the region. This is in stark contrast to other global markets, like in the US, where MSRP pricing for current-gen Radeon RX 9000 and GeForce RTX 50 Series graphics cards is a distant memory.

Per US retailer Newegg, the current lowest-priced and discounted new Radeon RX 9070 XT GPU is ASRock's Challenger model, which is selling for $709.99, which is still notably higher than the $599 MSRP for the card. Right now in China, you can grab several Radeon RX 9070 XT models, including premium overclocked variants like ASRock's Taichi edition for around 4999 RMB, which is the flagship RDNA 4 GPU's original MSRP.

If you're wondering if this is a precursor to global prices for PC gaming GPUs dropping in the weeks or months ahead, this situation looks to be regional, or, best-case, a "wait and see." According to the report, the main reason for the price drop is weak demand for Radeon GPUs and a DIY PC market currently under inventory pressure.

Today, AMD unveiled its new Instinct MI350P GPU accelerator, targeted at AI workloads. The company is marketing the MI350P as a cost-effective, drop-in upgrade for companies that want to expand their AI infrastructure. The MI350P is a dual-slot PCIe GPU accelerator that can be easily installed in existing air-cooled servers without extensive reconfiguration.

The key marketing point is definitely its PCIe form factor and relatively affordable price. AMD, in its press release, is directly targeting enterprises that want to adopt AI without having to invest heavily in new infrastructure. The company is offering air-cooled systems with up to eight Instinct MI350P accelerator cards, designed for small and medium-sized AI deployments.

Under the hood, the Instinct MI350P has 128 compute units, equaling 8,192 stream processors, along with 512 Matrix cores. The GPU uses the CDNA 4 architecture built on TSMC's 3nm production process. The core is in a 4-XCD configuration, boosting to 2200 MHz. The single I/O die is based on TSMC's 6nm FinFET process, and the GPU also houses 128MB of LLC in the form of Infinity Cache.

Nearly four years after PCI-SIG announced PCIe 7.0, the consortium has now revealed it is moving towards its successor. PCI-SIG members have gained access to version 0.5 of the PCIe 8.0 standard, described as the first official draft. The release incorporates feedback from the earlier 0.3 draft published in September 2025, while the final PCIe 8.0 specification remains on track for release in 2028.

PCIe 8.0 represents a major leap over existing standards, offering eight times the transfer rate of PCIe 5.0, which remains the fastest PCIe standard available on consumer desktop platforms today. On paper, the new standard delivers a 256 GT/s data rate and 1 TB/s of raw bi-directional bandwidth via a 16-lane configuration. As with every new PCIe generation, PCI-SIG aims to double bandwidth every three years. Here is the breakdown across different lane configurations:

At x1, PCIe 8.0 effectively matches the bandwidth of PCIe 4.0 running at x16 or PCIe 5.0 at x8. At x2, it reaches the full PCIe 5.0 bandwidth, while x4 mode matches PCIe 6.0 capabilities. Hence, technically, PCIe Gen8 NVMe SSDs will be rated for sequential speeds of up to 120,000 MB/s.

AMD Software: Adrenalin Edition 26.5.1 is here for all Radeon gamers, and it's a big one for new game support, adding optimizations for five new PC games. This includes support for Capcom's recent sci-fi action-adventure hit Pragmata, which is shaping up to be another Resident Evil Requiem-like success for the company, critically and commercially.

The latest Radeon driver also adds support for the fantasy-based action-RPG Honor of Kings: World, the immersive FPS and highly anticipated sequel Industria 2, the critically acclaimed and vibrant Tides of Tomorrow, and the monster-taming action RPG Mongil: Star Dive. If you're playing or thinking about picking up any of these titles, be sure to grab the latest Radeon driver.

Outside of New Game Support, AMD Software: Adrenalin Edition 26.5.1 also fixes a few key issues for Radeon RX 9000 Series gamers playing Resident Evil Requiem or God of War. The latest driver fixes stuttering issues in the former and visual corruption in the latter. Here's a look at the full Release Notes.

Ten years ago to the day, NVIDIA formally announced the GeForce GTX 10 Series, led by the iconic flagship GeForce GTX 1080. Interestingly, the launch of the GeForce GTX 1080 and the mid-range GeForce GTX 1070 marked the first NVIDIA GPUs to feature custom, premium Founders Edition models. The first FE cards debuted in 2016, with aluminum bodies and vapor-chamber cooling to support the new Pascal architecture built on the cutting-edge, for the time, TSMC 16nm process.

The Pascal and GeForce GTX 10 Series generations are widely considered among the most impressive in GeForce history, with pretty much every model in the lineup delivering a massive generational leap. Not only in performance, but with the adoption of next-gen GDDR5X memory for increased bandwidth, DisplayPort 1.4, HDMI 2.0b, GPU Boost 3.0, efficiency improvements, and several architectural changes to improve GPU stability, overclocking, and rendering technologies.

The flagship GeForce GTX 1080, which would be superseded by the legendary GeForce GTX 1080 Ti in 2017, arrived on May 27, 2016, with 9 TFLOPS of pure performance and 8GB of fast GDDR5X memory to kick-start the modern era of high-resolution, high frame-rate, PC gaming. Standout titles that showcased what the Pascal era brought to the table include id Software's DOOM reboot, commonly referred to as DOOM (2016), and CD Projekt Red's breakout RPG hit, The Witcher 3: Wild Hunt.

AMD has announced its financial results for the first quarter of 2026, and, naturally, Data Center revenue of $5.8 billion accounts for the bulk of the $10.3 billion in total revenue, driven by AMD EPYC processors and AMD Instinct GPU shipments. That said, Client and Gaming segment revenue was right up there with its $3.6 billion, representing a 23% year-over-year increase.

Breaking this down, AMD's Ryzen CPUs generated $2.9 billion in revenue for the Client segment, a solid 26% year-over-year increase, with the Gaming segment's 11% growth to $720 million "driven by solid demand for AMD Radeon GPUs." In fact, it sounds like Radeon GPUs carried the segment, as AMD notes lower semi-custom revenue for the quarter compared to last year, which covers consoles like the PlayStation 5 and Xbox Series X|S.

With Radeon driving Gaming growth for AMD, this is attributed to the well-received Radeon RX 9000 Series of desktop GPUs for PC gaming and local AI, as well as the company's commitment to expanding its neural rendering suite with the arrival of FSR 4.1 upscaling and Ray Regeneration 1.1 for enhancing ray-traced lighting detail.

Intel's long-awaited 'Big Battlemage' GPU finally got released this year, albeit in workstation form, with the new Intel Arc Pro B70 32GB graphics card. It's powered by the BMG-G31 silicon that was on track to power a potential Intel Arc B770 desktop PC gaming graphics card, which is no longer coming. As for what the gaming performance of this GPU might have looked like, we now have a more detailed look thanks to a new deep dive from Expreview.

The outlet tested the Intel Arc Pro B70's gaming performance at 1440p, with and without ray tracing, comparing it to the NVIDIA GeForce RTX 5060 Ti 16GB and Intel Arc B580 12GB cards, testing five games in each category. Looking at the overall results, the Intel Arc Pro B70's gaming performance falls short of beating the GeForce RTX 5060 Ti 16GB in both categories, with NVIDIA's GPU being around 5% faster on average when it comes to non-ray tracing or rasterized performance, and around 2% faster when it comes to ray-tracing performance.

That said, the two cards do trade blows, with the Intel Arc Pro B70 delivering faster performance in Cyberpunk 2077 in both scenarios, and faster ray-tracing performance on the Intel card in F1 25 and Kingdom Come: Deliverance II. Here's a closer look at the results.

The latest Steam Hardware and Software Survey results for April 2026 are in, and we've already gone through the changes in what graphics card models PC gamers are currently using. In addition to this. We've also noticed an interesting change in GPU VRAM capacity, with roughly half of PC gamers now playing on cards that feature more than 8GB of memory.

Now, before the whole memory crisis, which has led to GPU price increases and even some scarcity of cards with more than 8GB of memory, the debate over VRAM capacity has garnered significant attention. The reason is simple: as image fidelity in PC games has steadily increased, the memory requirements for playing with high-quality textures and cutting-edge effects like ray tracing have risen.

This is something that we mentioned in our reviews of the GeForce RTX 5060 8GB and GeForce RTX 5060 Ti 8GB GPUs, where it became clear that performance at higher resolutions and higher in-game graphics settings was becoming bottlenecked by VRAM capacity. According to the latest Steam data from Valve, 49.95% of PC gamers on the platform now use GPUs with more than 8GB of VRAM.

We recently covered a rumor that NVIDIA might be bringing back the RTX 3060 12GB, and it looks like more changes are afoot in the graphics card business. The ongoing memory shortages have impacted every aspect of the PC hardware industry, and the GPU business is no exception. Now, it looks like ASUS is pivoting away from its RTX 5070 Ti graphics cards toward its bigger brother, the RTX 5080.

According to industry sources quoted by Channel Gate, ASUS is planning to reallocate its GPU supply, prioritizing the RTX 5080 over the RTX 5070 Ti, thereby limiting the supply of the latter. The move is apparently a consequence of the current market conditions, with the global DRAM shortage in full swing. While NVIDIA remains firm that the supply of all RTX 50-series GPUs is stable, it seems ASUS has its own plans.

The current timeline for implementing this strategy is Q2 2026, but that is, of course, subject to change. The report claims that ASUS will focus on selling only a few mainstream RTX 5070 Ti models, but that the focus will eventually shift to the RTX 5080. The mainstream models in question might refer to low-end or mid-range variants such as the Dual and PRIME, while the high-end ROG Strix models are likely to be limited in production.

The Steam Hardware & Software Survey results for April 2026 are in, and based on recent news, it looks like the GeForce RTX 3060 might remain one of, if not the most popular PC gaming GPUs, for the foreseeable future. According to the latest rumor, the GeForce RTX 3060 12GB graphics card is set to make a return this summer with 'new' retail models from ASUS, MSI, COLORFUL, and others.

Naturally, this is the direct result of the current data center boom taking up GPU and GDDR7 memory capacity, and with the GeForce RTX 3060 sharing a similar market share with the GeForce RTX 4060, the number two GPU in Valve's latest Steam Hardware Survey, this comeback could cement the RTX 3060's dominance in the PC gaming space for a while.

That said, NVIDIA's latest GeForce RTX 50 Series continues to grow in popularity, with both the GeForce RTX 5070 and the GeForce RTX 5060 accounted for in the Top 10. Elsewhere, AMD's latest Radeon RX 9000 Series is no longer 'missing in action,' so to speak, with the Radeon RX 9070 making the list.

We recently covered rumors that the RTX 3060 12GB was set to return to market shelves in June. Now, we have a bit more evidence to back up that rumor: Chinese Board Channels have confirmed that the 5-year-old NVIDIA GPU is indeed returning this summer. The timeline differs from what was previously leaked, but it seems the Ampere GPU will be on store shelves by July at the latest.

Per a new post from Chinese Board Channels, the RTX 3060 12GB supply will resume in June. This means that NVIDIA will start supplying the 3060 GPU to board partners in June, allowing them to enter mass production. The add-in board partners are then expected to deliver the finished RTX 3060 12GB cards around July.

Among the partners expected to receive allocations for this re-launch are ASUS, MSI, Colorful, and GALAX (recently incorporated into Palit). All these board partners already have existing designs for the RTX 3060, so it may be easier for them to just re-launch those designs instead of making new ones from the ground up. After all, cost-cutting seems to be the entire point of this re-launch.

Microsoft has released a preview of Shader Model 6.10 in the new AgilitySDK 1.720-preview build. The update brings several changes, including tweaks to shared shader memory management and ray tracing, but the headline addition is a new Matrix feature.

According to the developer blog, Shader Model 6.10 introduces a streamlined matrix algebra API through the linalg::Matrix class, exposing a full set of matrix operations across GPUs from AMD, Intel, and NVIDIA. Modern GPUs already include dedicated hardware for AI workloads, such as NVIDIA's Tensor cores, Intel's XMX units, and similar accelerators on AMD hardware. However, DirectX support for these capabilities has been limited until now.

With this update, developers can access those matrix units through a unified API, rather than relying on vendor-specific implementations. This allows neural rendering operations to be executed across multiple GPUs with a single programming effort, rather than as a proprietary feature that developers must implement separately for each graphics card family. Microsoft is essentially trying to make matrix math a standard part of the DirectX API, making neural rendering a core DirectX feature rather than something bolted on by individual GPU makers.

Yesterday, we got the shocking news that GALAX was exiting the gaming GPU business, with Palit taking over the brand and existing products. The implication, at least on paper, was that this meant that we'd no longer be getting new GALAX GPUs - and that the company was no more. Turns out that what actually has happened is that GALAX, which has been a part of the Palit Group for several years, is now fully under the Palit umbrella.

This isn't to say that GALAX employees in various regions haven't been let go, as Palit's official statement calls it an internal transition. "The recent internal transitions are part of a pre-planned global initiative to integrate brand management for enhanced operational efficiency and cross-departmental synergy," Palit says.

Adding, "GALAX is not ceasing operations. We remain fully committed to the development, production, and support of our high-performance hardware. Our product roadmap continues as planned, and our commitment to delivering cutting-edge technology to gamers and creators remains our top priority."

Corsair has announced and launched a new 12V-2x6 GPU power cable called ThermalProtect that it says features "innovative technology that monitors the temperature of a GPU's power cable in real time to help prevent damage to the GPU." This Over Temperature Protection (OTP) is included in the seemingly normal cable comb that sits 30mm from the connection point.

The cable comb actively measures and monitors cable temperatures and, when an issue is detected, shuts down the GPU's power to prevent potential damage to the card or cable. This is the latest in what feels like a string of new power supply and cable technologies being introduced in response to several cases of cables melting due to incorrect power loads and overheating on modern GPUs in the GeForce RTX 50 Series.

Corsair notes that the ThermalProtect 12V-2x6 GPU power cable, available in Black or White for $24.99 USD, is compatible with any power supply with a native 12V-2x6 connector. Adding that this "broad compatibility with modern systems" helps give PC gamers peace of mind with a simple plug-and-play solution.

GALAX has been a notable GeForce brand for several years, with the company's Hall Of Fame models renowned for pushing performance to the limit and breaking world records in the process. It's a series that we've gone hands-on in the past, and for a little trip down memory lane, be sure to revisit our review of the impressive, for the time, GALAX GeForce GTX 1080 Ti HOF Edition graphics card.

Today, we got some relatively shocking news via Wccftech: the GALAX brand and company are no more, and the entire global team has been dismissed. And with that, in a statement provided to the outlet, Palit has "assumed full control and operations of the GALAX brand." Palit is now "solely responsible for all activities and commitments related to the brand," covering existing retail stock still out there, as well as warranties and RMA services.

"Following the closure of the previous organizational structure and the dismissal of its team, all management and operations will now be handled exclusively through Palit's official channels," the statement reads. As for the reason for the closure, an additional notification sent to Palit's customers confirms it's due to "supply constraints of various raw materials caused by the AI era."

NVIDIA has released GeForce Game Ready Driver 596.36 for GeForce RTX gamers, and in terms of new game support, this one is all about Conan Exiles Enhanced. This Unreal Engine 5 update to the long-running survival, crafting, action-RPG includes 9 years of expansions and additional content, with enhanced visuals.

The game is set to launch on May 5, with support for DLSS Multi Frame Generation, DLSS Super Resolution, and NVIDIA Reflex. Rebuilt on Unreal Engine 5, the game takes advantage of Lumen Global Illumination for improved lighting and Nanite Geometry to dramatically increase environmental detail and fidelity. Best of all, it's also a free upgrade for existing players.

In addition to Conan Exiles Enhanced support, the big thing with GeForce Game Ready Driver 596.36 is support for the new GeForce RTX 5070 Laptop GPU with 12GB of VRAM, which is now available. Plus, a few gaming and general bug fixes. Here's a look at the full Release Notes.

Following NVIDIA's announcement of the GeForce RTX 5070 12GB laptop GPU, Framework is adding it as a new Graphics Module for the Laptop 16. However, launching a 12GB VRAM option in the middle of a DRAM crisis is going to sting Framework fans, as this version is apparently much more expensive than its lower-VRAM counterpart.

The Framework Laptop 16 RTX 5070 12GB Graphics Module is now available for pre-order at $1,199. This puts it at a 72% premium, or a whopping $500 extra, for just 4GB of additional VRAM over the existing RTX 5070 8GB model. The two chips are otherwise identical. Both feature 4,608 CUDA cores, boost clocks of up to 2.4 GHz, 384 GB/s of memory bandwidth, and a maximum TGP of 100W. NVIDIA has simply swapped the 2GB memory modules for 3GB chips and made no other changes. So, unless you are running VRAM-intensive tasks, you are unlikely to notice much difference between the two.

On top of that, Framework has also warned that the $699 RTX 5070 8GB module may not stay at its current price for long. In an X reply, the company said the price hikes it is seeing come directly from silicon suppliers, and once its current GDDR7 inventory depletes, the 8GB configuration is highly likely to increase in price as well.

NVIDIA is expanding its Blackwell laptop lineup with the GeForce RTX 5070 12GB Laptop GPU. Confirmed by a quiet driver update blog post, this GPU aims to address longstanding user feedback about the VRAM limitations of NVIDIA's mid-range mobile silicon.

Rather than a direct replacement, the 12GB model will serve as a higher-tier alternative to the existing 8GB model. In other words, laptops with the larger memory capacity model will carry a price premium.

This is not a new GPU core with a wider memory bus. At its core is the same GB206 chip, featuring 4,608 CUDA cores and a 128-bit interface made up of four 32-bit channels. For the 12GB version, NVIDIA is using higher-density GDDR7 modules, with each channel populated by a 3GB chip, for a total of 12GB. By comparison, the original 8GB version uses 2GB modules per channel, for a total of 8GB across the same bus.

Intel's latest driver release, 32.0.101.8517, for Arc Pro GPUs increases the integrated GPU's memory allocation to enable broader LLM inference support. The new driver allows users to allocate up to 93% of their system RAM to the integrated GPU. While the driver currently supports only a select number of SKUs, Intel is paving the way for larger LLM inference workloads without hitting memory capacity bottlenecks.

Traditional memory partitioning usually limits a GPU to 50% of system RAM. AMD's Variable Graphics Memory (VGM) allows high-end configurations, such as the Strix Halo, to allocate 96GB from a 128GB pool to the iGPU. Intel has been more aggressive in this regard. Last year, Intel raised the limit to 87% with its new "Shared GPU Memory Override" for Core Ultra Series 2 processors.

The latest driver release pushes that boundary further to 93% for local AI inference. This only supports integrated Arc Pro GPUs, such as the Arc Pro B390 and Arc Pro B370. While this allocation update is the headline feature for integrated GPUs only, the driver also supports discrete Arc Pro A and B-series cards.