Graphics Cards - Page 6

MSI appears to have leaked a new configuration of NVIDIA's RTX 5070 laptop GPU, and it's not the first time we've caught a glimpse of this rumored variant that beefs up the VRAM considerably.

VideoCardz noticed the leak in a press release from MSI which introduces new Raider and Crosshair gaming laptops.

In the spec of the MSI Crosshair 16 Max HX (E2WGXK), under the graphics category there's a mention of the RTX 5070 with 12GB (as opposed to the 8GB loadout that the notebook GPU currently runs with).

According to a new report from Warranty Week, NVIDIA's warranty claim expenses for discrete GPUs increased by a whopping 1000% in 2025 compared to 2024. This refers to the money NVIDIA has spent on warranties, with the report noting that NVIDIA paid $81 million in claims in 2024, rising to $894 million in 2025.

Naturally, this increase is also reflected in the warranty claim rate for NVIDIA's GPUs, which rose from 0.17% in the first quarter of 2025 up to 0.90% in the fourth quarter. Warranty Week obtained all of this data from NVIDIA's public records, so we don't have specific information on which GeForce RTX models are affected or whether this includes RTX Ada and Blackwell data center and workstation graphics card products. However, with $894 million spent on warranty claims, it probably includes all of the above.

For consumers and PC gamers, a rise in warranty claims makes sense when you factor in the power issues surrounding the GeForce RTX 40 and RTX 50 Series graphics cards that adopted the new 16-pin 12VHPWR and 12V-2x6 power connectors, which led to several well-publicized cases of cables melting and GPUs getting damaged. Another factor would be the overall price increases associated with several GeForce RTX 50 Series and RTX Blackwell workstation GPUs.

Last month, we reported on a rumor that NVIDIA is planning to launch a new GeForce RTX 5050 with 9GB of VRAM, increasing the overall VRAM capacity by 1GB. Now, that's not a lot, and the main reason for the change is the recent availability of 3GB GDDR7 modules. For a GeForce RTX 5050 9GB GPU, this would reportedly also feature a 96-bit memory bus rather than 128-bit, delivering a 5% increase in memory bandwidth and a 12.5% increase in VRAM capacity.

This month, we've got a fresh batch of 9GB GPU rumors courtesy of the China-based Board Channels forum, which regularly posts new information on unreleased GPUs and inventory updates from NVIDIA's GeForce RTX partners. The translated post notes that in addition to releasing a GeForce RTX 5050 9GB GPU, NVIDIA might also upgrade its other 8GB GPUs with this new 9GB configuration featuring 3GB modules.

So, that would mean that we could see GeForce RTX 5060 9GB and a GeForce RTX 5060 Ti 9GB GPUs as early as the end of May or the beginning of June. As noted by VideoCardz, which broke this story, while the extra 1GB of memory might come in handy in some titles, if NVIDIA were to shift to the smaller 96-bit memory interface, the overall bandwidth on these new GeForce RTX 5060 and RTX 5060 Ti variants would drop, potentially affecting performance.

Capcom's Pragmata is out this week, and based on early reviews from various media outlets, it's another excellent game from the company that recently released the acclaimed Resident Evil Requiem. Pragmata is a sci-fi action-adventure set in the near future, with a realistic visual style that presents a futuristic vision of our moon and Earth, centered on the discovery of a new material called Lunafilament.

In the game, you take control of investigator Hugh and a mysterious young android called Diana as you explore a lunar installation and face off against a rogue AI. Like Resident Evil Requiem, Capcom's Pragmata runs on the studio's proprietary RE Engine. For the game's PC version, that means support for a visually stunning path-tracing mode in addition to ray tracing.

And for GeForce RTX gamers, especially those with GeForce RTX 50 Series graphics cards, there's full support for DLSS, including Super Resolution, Ray Reconstruction, Frame Generation, and Multi Frame Generation. Plus, the brand-new Dynamic Frame Generation. And with that, NVIDIA has announced the new Pragmata GeForce RTX 50 Series Bundle, which is offering a Steam copy of the game with qualifying GeForce RTX 50 Series purchases.

Thermal Grizzly's new DeltaMate GPU Block is specifically built for the premium ROG Astral GeForce RTX 5080 GPU from ASUS. It features a "massive copper cooler" with a dual-cold-plate design, one for the GPU die and one for the VRM/VRAM. The VRM/VRAM-Coldplate alone weighs over 900 grams and is built with custom CNC machining to enhance heat transfer and thermal performance.

Interestingly, Thermal Grizzly notes that the DeltaMate GPU Block for the Astral RTX 5080 proved more challenging and complex to build than its previous RTX 5090 model, even though the RTX 5080 draws significantly less power. The reason is that the GeForce RTX 5080 has higher power density than the RTX 5090, making it harder to cool. Basically, it's a smaller, more densely packed GPU.

This helps explain why the price of the DeltaMate GPU Block for the Astral RTX 5080 is currently listed at $643.38 USD, given that it was difficult for Thermal Grizzly to source parts in Europe without paying inflated prices. And when you add in the specialized cold plates, anodized aluminum housing, brass terminals, reinforced glass window, and intricate design, that is why it costs roughly a third of the ASUS flagship GeForce RTX 5080 GPU it's designed to cool. The good news on that front is that it's highly effective.

Here's an interesting piece of hardware you may not have known existed: a Dual PCIe adapter card that adds two PCIe 4.0 M.2 SSD slots to the same PCIe slot you use for your GPU. This is for those systems and rigs where a more affordable GPU might not be utilizing all x16 lanes.

What makes this stand out as an oddity is that the riser card will only work on motherboards that support PCIe bifurcation, which splits the x16 PCIe slot into three lanes: x8, x4, and x4. As posted on Reddit by 'thepromiseman,' they paired what appears to be the RIITOP Dual NVMe PCIe Adapter (available on Amazon for $29.99 USD) with a low-profile GIGABYTE graphics card in a Mini-ITX system.

And with that, the combined height of the GPU and adapter equals the standard height of a non-low-profile GPU. The benefit here is that the user was able to add two standard NVMe SSD slots to their compact system that would have been impossible otherwise. The user also notes that, since their motherboard supports PCIe bifurcation, everything worked without any tinkering or troubleshooting.

Although the current memory crisis and supply issues have led to GPU shortages and price increases in many regions, ASUS's new GeForce RTX 5080 looks set to be the company's most affordable RTX 5080 so far. At a glance, the new ASUS PRIME GeForce RTX 5080 EVO and ASUS PRIME GeForce RTX 5080 EVO OC Edition look identical to previous PRIME RTX 5080 models, except for one notable change.

Based on the product pages, it looks like ASUS has shifted from the original ASUS PRIME GeForce RTX 5080's vapor-chamber cooling to a more traditional approach, while maintaining the same specs and overall physical design. These new EVO variants feature the same SFF-Ready 2.5 slot design with three axial-tech fans and a vented backplate for improved airflow.

Although the vapor chamber is gone, the ASUS PRIME GeForce RTX 5080 EVO OC Edition still maintains the out-of-the-box OC Mode Boost Clock speed of 2685 MHz, a notable but relatively small increase from the reference design's 2617 MHz. This means the overall performance should be virtually identical to that of the non-EVO edition, but the overall operating temperature might be higher.

With power delivery for modern GPUs still a concern with the new 12V-2x6 PCIe power connector, we've seen a wide range of protection measures and cable designs appear in recent years. However, ASUS's new ROG Equalizer might be the most notable, as it's essentially a 12V-2x6 PCIe power cable designed to address load-balancing and thermal performance issues for GPUs drawing up to 600W, like the GeForce RTX 5090.

For example, each wire or pin on the ATX 3.1- and PCIe 5.1-compliant ROG Equalizer cable now supports up to 17 A, compared to 9.2 A on standard 12V-2x6 PCIe power cables. And with improved load balancing and a patented design, the ROG Equalizer cable can withstand some of the most extreme power scenarios.

"Under an extreme test scenario, we removed the middle 4 wires from the +12V cable to simulate potential current imbalance conditions that may occur with a standard 12V-2x6 PCIe cable," ASUS claims. "Even under these conditions, the ROG Equalizer maintained a temperature of approximately 73.4°C. In contrast, a standard 12V-2x6 PCIe cable reached around 146°C - far exceeding safe operating limits."

Although Intel's 'Big Battlemage' graphics card hasn't arrived in PC gaming form, the latest Intel Graphics Driver 32.0.101.8629 (WHQL Certified) has added official 'Gaming Support' for the new Intel Arc Pro B70. This workstation GPU, which recently launched, includes the chip that could have powered an Intel Arc B770 GPU, complementing the mainstream-focused Intel Arc B580 GPU, which is now a couple of years old.

With 32 Xe Cores and 32GB of GDDR6 memory, the new Intel Arc Pro B70 (starting at $949 USD) is an AI-focused GPU designed to offer a more affordable inference solution for AI enthusiasts. However, as its BMG-G31 die also includes ray-tracing capabilities and other gaming-focused technologies, it's also a capable PC gaming option. And with that, it makes sense that the latest 'Game On Graphics Driver' for Arc Graphics has added official Gaming Support for the new Intel Arc Pro B70 and Intel Arc Pro B65 GPUs.

Of course, this isn't to say that the Intel Arc Pro B70 is now a viable option for PC gamers, as its design, cooling, and pricing sit in the AI workstation tier, where hardware costs are significantly higher than those of gaming components. That said, we do have a general idea of how the Intel Arc Pro B70 performs during gaming workloads, so it's a shame that we're not getting an official 'Big Battlemage' gaming option.

As spotted by VideoCardz, it looks like ASUS has just increased the prices of its flagship RDNA 4 offerings in the US: the ASUS Prime Radeon RX 9070 XT OC Edition and the ASUS TUF GAMING Radeon RX 9070 XT OC Edition. Although there has been no official word from ASUS of the increase, the new prices are already up on the official store and have even made their way to retailers like Amazon and B&H.

Based on the double-digit percentage increase, this appears to be a direct response to AMD raising prices on its 16GB GPUs amid the current memory crisis affecting the wider industry. The ASUS Prime Radeon RX 9070 XT has seen its price increase from $799.99 to $939.99, a 17.5% increase, which is reflected on the aggregate site PCPartPicker.

As for the flagship ASUS TUF GAMING Radeon RX 9070 XT, its price has risen from $849.99 to $989.99, a 16.5% increase. Although the $599 MSRP set by AMD for the Radeon RX 9070 XT has never really been a reality for most Radeon RX 9070 XT models, this new price for the TUF GAMING variant represents a massive 65% increase over what some gamers were able to spend to pick up a Radeon RX 9070 XT at launch.

We've all seen the fancy tech that some graphics card makers have implemented to help save owners from melting cable disasters, but there's a new way to protect your GPU apparently - namely to get yourself a tech-savvy cat.

As reported on the PTT Bulletin Board System (BBS) in Taiwan - which is regarded as the equivalent of Reddit for the country - the owner of an RTX 4090 was saved by the call of their cat.

The tale, spotted by VideoCardz, goes like this: the PC enthusiast went to the toilet, only to hear the meowing of the cat, so they rushed back to notice smoke emanating from the computer and a burning smell.

NVIDIA formally announced and unveiled its Pascal architecture during the launch of the Tesla P100 accelerator a decade ago. Ten years to the day, basically, and although it would take a few weeks for the first Pascal-powered desktop gaming GPU to hit the market, the architecture led to the arrival of some of the most iconic PC gaming graphics cards - including the GeForce GTX 1080 Ti and the GeForce GTX 1060.

So iconic that the mainstream GeForce GTX 1060 is still one of the most popular PC gaming GPUs according to the latest data from Steam, which shows us that as of March 2026, there are more 1060s out there than any single Radeon GPU. Although the flagship GeForce GTX 1080 Ti didn't arrive until 2017, it offered such a massive performance increase over anything else that it only recently began to feel obsolete due to its lack of dedicated AI hardware and DLSS support.

The Pascal architecture was a game-changer in the data center space, especially for AI and other applications, but for gamers, its debut arrived with the GeForce GTX 1080 on May 17, 2016, built on Pascal's efficient 16nm FinFET process with 8GB of GDDR5X memory. It was the first GPU to power 4K 120 Hz gaming via DisplayPort 1.4, while also leveraging architectural improvements to elevate VR and 3D gaming. Yes, 2017 was a very different time.

With the recent launch of the Intel Arc Pro B70, the company finally released its 'Big Battlemage' graphics card for consumers, albeit in a form aimed at the AI crown and workstations. Powered by the BMG-G31 graphics processor, the Arc Pro B70 features 32 Xe Cores (with Ray Tracing Units), 256 XMX engines, and 32GB of GDDR6 memory on a 256-bit bus with 608 GB/s of memory bandwidth.

Although it's not a gaming-focused model, we know the DirectX 12 GPU can deliver gaming performance and would make for an impressive Intel Arc B770 if the price is right. That said, with Intel's reference Arc Pro B70 design priced at $949 USD, the card's AI performance and 32GB of VRAM have already made it popular among AI enthusiasts.

Per Newegg's listing for the Intel Arc Pro B70, it's currently the number one best seller in the Workstation Graphics Cards category. It's a card built for AI inference, and with its attractive price point and performance, it looks like its launch has been a success.

AMD debuted its latest generation of desktop graphics cards for PC gaming last year, with the mid-range RDNA 4 combo that is the Radeon RX 9070 and the Radeon RX 9070 XT. In addition to using a more power-efficient 4nm process and an overhauled architecture that brought notable improvements in ray-tracing performance, the Radeon RX 9000 Series also introduced AMD's new AI-powered FSR 4 upscaling, which delivers image quality finally comparable to NVIDIA DLSS.

With these mid-range cards presented as alternatives to the GeForce RTX 5070 and the GeForce RTX 5070 Ti, the value proposition has always been tied to their respective prices. Although GPU prices in the US have made MSRP a thing of the past since the memory crisis kicked into gear, it looks like PC gamers in Europe can pick up a Radeon RX 9070 or Radeon RX 9070 XT at a discount. This includes models sold at a price notably below the region's MSRP.

However, these deals appear to be limited to one country, Germany, where the ASUS PRIME Radeon RX 9070 OC GPU, the non-XT model, is reportedly available for €539, a significant discount from the launch MSRP of €629, including sales tax. Likewise, the ASRock Radeon RX 9070 XT Challenger GPU is reportedly being sold for €640, another notable discount from its €689 launch MSRP.

The Steam Hardware & Software Survey results for March 2026 are in, and there's some definite course correction after February's wild results. Last month, several GeForce RTX 50 and 40 Series cards shot up the market share ranking, with the GeForce RTX 5070 taking the top spot as the world's most popular gaming GPU. Now, based on the results, we were skeptical of their validity, as it represented one of the biggest single-month shifts to date.

The good news is that the March 2026 results look to be more in line with recent trends, which means the GeForce RTX 3060 has reclaimed the top spot as the most popular gaming GPU among PC gamers, followed by the GeForce RTX 4060. Now, the GeForce RTX 5070 is still proving to be the most popular GPU option in the current generation of GeForce RTX and Radeon RX cards, where it sits in the number five spot.

In the span of a year, it's become more popular than the GeForce RTX 3070, RTX 4070, and even the mainstream-focused GeForce RTX 5060. And speaking of Radeon RX cards, the first RDNA 4 model has made its way into the Steam Hardware & Software Survey results, with the Radeon RX 9070 appearing way down the list with 0.16% market share.

NVIDIA is reportedly running into manufacturing challenges with its next-generation Rubin Ultra GPU and is now considering a significant architectural revision. The standard Rubin, built for large-scale AI training, is on track to begin mass shipments this summer. Even before that rollout, reports about the more advanced Rubin Ultra are starting to surface, and its current roadmap may be hitting some real technological walls.

To understand the problem, it helps to know what Rubin Ultra was originally planned to be. NVIDIA introduced a dual-die architecture with the standard Rubin, meaning two silicon chips packaged together into one unit. Rubin Ultra was meant to take that further with a four-die setup, essentially doubling the base design into a much larger package. However, those ambitions may have pushed TSMC's advanced packaging technology past its practical limits.

Reports from Taiwan's Commercial Times suggest that NVIDIA may scale back the Rubin Ultra to a dual-die design, similar to the standard Rubin. The original design reportedly included 16 HBM4 memory stacks, around 1 TB of memory capacity, and CoWoS-L packaging.

NVIDIA's upcoming DLSS 4.5 update will introduce a new feature called Dynamic Multi-Frame Generation (DMFG), but what is interesting is that it's already available via over-the-air (OTA) updates in the NVIDIA App.

It appears that new DLSS files and components have already been delivered through updates, but aren't officially "turned on" by NVIDIA. However, users have found workarounds to enable the feature, such as turning on DLSS Overdrive or tweaking settings with tools like NVIDIA Profile Inspector.

For those who don't know, current DLSS Frame Generation inserts AI-generated "fake" frames between real rendered frames to boost FPS. Dynamic Multi-Frame Generation changes that approach, as it dynamically adjusts the number of frames generated in real time (4x, 6x, etc.). Dynamically adjusting the number of generated frames will, in theory, improve image quality as the number of lower-quality AI-generated frames on-screen is reduced, providing a more balanced mix of natively rendered and AI-generated frames.

Intel has finally released its fully unlocked 'Big Battlemage' graphics card; however, it's arrived in Intel Arc Pro B70 form and not a card aimed at the PC gaming market. This means the existing Intel Arc B580 remains the company's flagship PC gaming GPU, an affordable mainstream offering with performance comparable to the GeForce RTX 4060.

However, with the Intel Arc Pro B70 now out in the wild, we've got our first look at the Intel Arc Pro B70's gaming performance compared to the Intel Arc Pro B60 courtesy of Level1Techs. Even though these are two workstation cards, the results are a good indicator of what an Intel Arc B770 versus Intel Arc B580 gaming showdown might have looked like - with 'Big Battlemage' proving to be around 45% faster.

These results cover three games testing at 1080p and 1440p: Cyberpunk 2077, Shadow of the Tomb Raider, and Monster Hunter Wilds. With hardware-intensive Cyberpunk 2077's performance jumping from 58 FPS to 83 FPS at 1440p on Ultra settings, Intel's 'Big Battlemage' could have been a real disruptor in the affordable mid-range space.

The GeForce RTX 30 Series is widely considered one of NVIDIA's best in the RTX era, delivering notable performance improvements over previous generations and introducing the game-changing DLSS 2.0 technology. However, if there was one flaw or recurring complaint with the GeForce RTX 30 Series, it was that the VRAM and GDDR6X memory ran very hot, with temperatures on high-end models like the GeForce RTX 3080 reaching as high as 100 degrees Celsius.

In YouTube modder and hardware enthusiast TrashBench's latest video, he attempts to solve this problem by taking an Arctic Liquid Freezer WS360-4710 workstation-grade AIO cooler and connecting that to the GPU and memory on a GeForce RTX 3080. The fact that it's a workstation-grade cooler is important because it includes a larger cold plate.

Securing and mounting the Arctic Liquid Freezer WS360 AIO cooler on a GeForce RTX 3080 required a lot of trial and error and a few iterations. However, with its all-black design, the result is pretty impressive, to the point where it resembles a custom-built AIO solution for the GeForce RTX 3080. What followed next was even more impressive, with GPU temperatures dropping substantially and VRAM or memory temperatures dropping by a massive 54 degrees Celsius.

NVIDIA is expected to maintain its 2-year launch cadence for GeForce graphics cards, as leaks about the next-generation RTX 60 series are starting to surface. The GeForce RTX 60-series will be powered by NVIDIA's "Rubin" architecture, which has already debuted on their AI and datacenter GPUs.

A new leak has surfaced from YouTuber RedGamingTech, combining a few interesting pieces of information about the upcoming NVIDIA GPUs. The first key point is that NVIDIA will use a variant of the current TSMC 3nm FinFET node to produce the RTX 60-series GPUs. Of course, this puts an end to the speculation of a new sub-2nm node debuting with the RTX 60-series.

Of course, NVIDIA will likely collaborate with TSMC to derive a custom node based on the 3nm FinFET that suits NVIDIA's needs, as they have done in the past. Furthermore, it has been leaked that the GPU chips will follow the "GR20X" numbering scheme. This is something that wouldn't come as a surprise if you have been paying attention to NVIDIA's GPU nomenclature over the past few years. Reportedly, the top-end GPU will be the GR202, followed by the GR203 and GR205 GPUs.