Graphics Cards - Page 32

NVIDIA's PhysX and Flow technologies are fully open-source. The source code for PhysX V5.6.0 and Flow V2.2.0 is now available via GitHub under the BSD-3 license. This is good news because the last time PhysX made headlines was back in February when it was reported that NVIDIA's new GeForce RTX 50 Series no longer supports 32-bit CUDA applications.

This means 32-bit PhysX games like Batman: Arkham Asylum, Arkham City (seen above), and Borderlands 2 run extremely poorly on GeForce RTX 50 Series GPUs with PhysX enabled. As the name suggests, PhysX is all about simulating real-world physics using your GeForce GPU - with everything from body dynamics to deformable objects to fluid simulation and more. NVIDIA Flow is all about large-scale fluid simulation.

With the complete source code, developers and modders will not only be able to leverage GPUs and CUDA to implement, tinker, and experiment with real-time physics, but they could also use the code to mod or update older 32-bit PhysX games so that they can run on modern GeForce RTX 50 Series hardware via a 32-bit ti 64-bit compatibility layer.

According to a new post at Board Channels (via Videocardz), the upcoming GeForce RTX 5060 Ti will follow 'the same pricing strategy' as the GeForce RTX 4060 Ti launch. A card that debuted with a $399 MSRP for the GeForce RTX 4060 Ti 8GB model and then $499 for the 16GB version that dropped weeks later. Several leaks point to the GeForce RTX 5060 Ti launching with both models on day one - which means 'the same pricing strategy' would make the 16GB version more expensive. $100 more.

Many assumptions and several grains of salt need to be considered regarding these pricing rumors for the upcoming GeForce RTX 5060 Ti, a card reportedly launching on April 15. In addition to shortages and demand pushing new and older GPU prices up to the point where MSRP can feel meaningless, there's also the looming chaos on the way with new tariffs being imposed on imports by the Trump administration.

Companies like NVIDIA and AMD also disclose pricing at the last minute to avoid leaks, and based on the current climate, we expect pricing to come at the very last minute for the GeForce RTX 5060 and Radeon RX 9060 Series.

ASUS has increased the headroom available for juicing up its top-end RTX 5080 graphics card, and by a large amount of wattage, although there are very much diminishing returns in terms of overclocking potential when it comes to flooding the GPU with power like this.

The board in question is the ROG Astral RTX 5080, and a new BIOS from ASUS has upped the ceiling for the TGP to 450W. Previously, it was 400W, so this is a jump of 12.5%, as highlighted by Tech Powerup (via VideoCardz).

However, if you're expecting to get 10% faster frame rates by pushing the power limit of the ROG Astral to 450W, then don't be under any illusions - that's not going to happen.

The new GeForce RTX 50 Series and Radeon RX 9070 Series graphics cards are hard to find, and when they are available, they sell out at inflated prices; the current GPU market for PC gaming is pretty dire in the U.S. With limited availability of brand-new models, increased pricing, and the whole tariff fiasco, the chaos extends to older GPUs too.

We could detail issues and delays surrounding the GeForce RTX 50 Series launch and discuss reports indicating that NVIDIA chose to halt the GeForce RTX 40 Series last year as the entire RTX Blackwell line-up was on track for a Q1 2025 launch. However, all you need to know about the current state of PC gaming GPUs can be found in Newegg's best-selling GPU list.

The MSI Ventus GeForce RTX 4060 2X is in the top spot with a price point of $459.98, a 53% increase over the MSRP. To make matters worse, the $460 price is presented as a discount, 9% off the $507 price. A GIGABYTE GeForce RTX 4060 EAGLE model is also available for $454.98.

Cooler Master's novel take on an RTX 5080 graphics card was presented at CES 2025, and we've now seen some further details about this innovation (complete with pics, too, a couple of which you can check out above and below).

You may not have seen the original revelation, but Cooler Master's idea is to present an RTX 5080 with a custom cooling solution that allows for the fans to be swapped out - and other manufacturer's fans can be used, not just those from Cooler Master.

In other words, you can pick a high-quality fan of your preference, one that's nice and quiet, to keep the RTX 5080 cool and running smoothly.

As the next move with RDNA 4 GPUs, AMD is rumored to be launching a GRE spin on its RX 9070 graphics card.

As you might expect, this is supposedly happening in the Asian market, but if an RX 9070 GRE does indeed end up being realized, there's always a decent chance it could eventually be available in the US and other regions. There's past precedent for that, certainly.

The rumor of the RX 9070 GRE comes from IT Home, a Chinese tech site, as flagged up by VideoCardz, but at this point it remains a relatively flimsy assertion.

A faulty GPU chip has reportedly been found in a PowerColor Radeon RX 9070 XT Hellhound, leading to GPU temperatures and hotspots exceeding the 110 degrees Celcius threshold for RDNA 4 GPUs - hitting 113 degrees. The full report can be found at Igor's Lab, where Igor Wallossek conducted extensive tests on the GPU sent to their lab by a reader reporting unusually high temperatures.

Igor notes that the fault lies with the GPU die because even after applying a "high-quality PTM pad and additional thermal putty instead of rigid pads on the memory and voltage converters," the GPU's thermal performance did not significantly improve. Microscopic examination of the GPU die found 1,934 pits across the surface, with the largest exceeding industry standards.

These tiny pits prevent proper heat transfer, leading to temperatures as high as 113 degrees. This alarming situation renders the Radeon RX 9070 XT faulty or damaged; thankfully, AMD has already responded to the report, saying that it believes this is "an isolated incident" and will work with its partners to "understand the issue."

The Steam Hardware & Software Survey results for March 2025 are in, and the very first GeForce RTX 50 Series GPU has made the list of most used graphics cards among PC gamers - and it's the GeForce RTX 5080.

The GeForce RTX 5080 squeaks in at the bottom of the list ahead of the GeForce RTX 5090. As the more 'affordable' high-end 4K RTX Blackwell option launched at the end of January 2025, it makes sense - mainly because it arrived with an MSRP lower than the previous generation's GeForce RTX 4080.

With the mid-range GeForce RTX 5070 Series and Radeon RX 9070 Series being more recent launches, we should expect to see these show up in the results in the coming months. Elsewhere, there's been some reshuffling, with the GeForce RTX 3060 retaking the crown from the GeForce RTX 4060 and the GeForce RTX 4070 dropping out of the Top 10.

You might be looking at the date and thinking that we're close enough to April 1 that this is some sort of tongue-in-cheek story, but it's not. For those that know their NVIDIA lore, as plans were being put in place to start one of the biggest tech companies in the world, co-founder and CEO Jensen Huang's first job was at American diner-style restaurant chain Denny's.

Between long shifts and saving up enough money to start a new business (and probably putting a deposit on his first leather jacket), Jensen Huang would sit at Denny's booth drafting plans for what would end up being NVIDIA. For food, he would wrap Denny's sausages in Denny's pancakes, pour syrup over them, and eat them.

Questionable nutritional value aside, this was (apparently) a key stepping stone toward AI and PC gaming domination with CUDA, RTX, and GeForce. We like to think that seeing the reflection of his surroundings in the syrup made him realize that one day, real-time ray-tracing in games would be possible. Or not. Either way, this combination of Denny's food is now being sold as 'NVIDIA Breakfast Bytes' for a limited time.

NVIDIA's previous-gen GeForce RTX 3090 is being rebranded, rebadged, and ready to scam unsuspecting PC gamers as the more expensive, and higher-end RTX 4090... oh boy.

This new scam involves polishing the GPU die and re-labeling it using a laser, so that it looks like it's got some official markets on the GPU die from NVIDIA. You've got to hand it to these scammers, they're getting better at what they do, but you can still tell: the capacitor layout surrounding the GPU is obvious, as our friends at Uniko's Hardware said in the above post on X.

Uniko's Hardware posted: "Another fake 4090 that used a 3090 die labbed the die then remark it as 4090 AD102, the MLCC surrounding the die are the layout the 3090 uses" adding "or just check the location of the dots at corners".

Hopes that Intel may yet release a higher-end Battlemage graphics card - following the success, and positive reception, of the Arc B580 - appear to be fading fast.

Wccftech was on the ball enough to catch an update from a well-known leaker on X which poured cold water on the theories that Intel could still be preparing a beefy Battlemage GPU to take on AMD and NVIDIA.

Specifically, this was the Battlemage BMG-G31 GPU, which has been popping up here and there on the rumor mill for a while - plus the success of the B580 (and B570) might've further buoyed hopes that such a move could be coming from Intel.

We covered Thermal Grizzly's WireView Pro GPU in detail late last year, as it includes internal and external temperature sensors for 12VHPWR 16-pin power connectors. The shift to a new power delivery system and flagship GPUs like the GeForce RTX 4090 and GeForce RTX 5090, which draw up to 450W and 600W on a single cable, led to Thermal Grizzly's developing the device.

An audible alarm not only allows users to monitor a GPU's power draw via the 16-pin power connector but also warns when temperatures reach a certain point to protect the graphics card from damage. Developed in collaboration with Elmor Labs founder and engineer Jon Sandstrom, the new 90-degree WireView Pro GPU offers a different cabled orientation, with the power supply cable router downward for GPUs installed horizontally.

There's also a sensor pin detection system to ensure the power from the power supply is connected correctly. As per the version released last year, an OLED display for real-time monitoring allows users to log and track power usage over time.

NVIDIA has released an "experimental version" of its Project G-Assist System Assistant, which we previewed back at Computex 2024. Available for GeForce RTX desktop owners (GeForce RTX laptop support is coming in a future update) via the NVIDIA App, the AI assistant will simplify optimizing your PC if you have a GPU with at least 12GB of VRAM.

"G-Assist helps users control a wide range of PC settings, from optimizing game and system settings, charting frame rates and other key performance statistics, to controlling select peripheral settings such as lighting - all via basic voice or text commands," NVIDIA explains.

Project G-Assist is powered by a custom Small Language Model (SLM) using a Llama-based Instruct model with 8 billion parameters to interpret natural language. This means you can ask it about your PC's hardware, explain what DLSS or other RTX technologies do, and provide detailed performance. You can even use it to overclock your GPU or change your display settings.

When we first saw PowerColor's impressive Radeon RX 9070 XT Red Devil at CES 2025, we were told it planned to release an all-white version in the months following RDNA 4's launch. The time has come, as the Radeon RX 9070 XT Red Devil Spectral White Edition is set to debut on April 8 in Limited Edition form.

PowerColor's Red Devil cards have always featured all-black designs with red flourishes, signaling their power and premium status among custom-overclocked models. This marks the first all-white edition designed for PC gamers with all-white builds, and as seen in a new video by Overclock3D that previews the card, everything is white.

Everything from the power 8-pin power connectors to the PCB, fin stack, I/O bracket, and heat pipes is all-white - which is very impressive. Most white GPUs rarely go this far in their commitment to an aesthetic.

We've caught an early glimpse of NVIDIA's RTX 5090 laptop GPU in action in a Razer laptop, with a YouTuber providing benchmarks that show the new mobile graphics card promises to be a major leap in performance - certainly in some scenarios.

You can see the results of YouTuber Dave2D running the RTX 5090 through its paces in the incoming Razer Blade 16 in the video above, as noticed by Notebookcheck.net.

In Black Myth: Wukong, Dave2D observes that at 1440p (ray tracing on, top graphics details, DLSS balanced) the RTX 5090 is a healthy 33% quicker than the RTX 4090.

The NVIDIA App now has AI built in to help PC owners fine-tune their games for the best performance, and a bunch of other tricks besides.

If Project G-Assist rings a vague bell, this might be because bringing AI into the mix in this way was once an April Fools joke from NVIDIA (way back in 2017). However, we learned last year that G-Assist was happening for real (the tech was shown off at Computex 2024).

Now, as The Verge spotted, G-Assist is here in the latest version of the NVIDIA App. As the name suggests it's an AI gaming assistant to help optimize your Windows and per-game settings with RTX graphics cards. So, you can request (by typing a query, or issuing a voice command) to optimize a specific game for the best image quality, for example, and the relevant settings will be applied automatically for you.

Mainstream GeForce RTX 5060 Series leaks and rumors are starting to heat up. In the past week or so, detailed specs for both the GeForce RTX 5060 and GeForce RTX 5060 Ti have emerged alongside the current planned release dates. The GeForce RTX 5060 Ti will launch on April 16, with the RTX 5060 following in mid-May.

NVIDIA has yet to announce the mainstream GeForce RTX 5060 Series, and based on the latest rumor from Hong Kong outlet HKEPC (via Videocardz), we won't get any official information until the 11th hour. According to the source, NVIDIA is currently planning to announce the GeForce RTX 5060 and GeForce RTX 5060 Ti (which will launch in 8Gb and 16GB variants) on April 15, one day before the GeForce RTX 5060 Ti hits shelves.

The rumor also notes that reviews for the RTX 5060 Ti will go live on April 16 for the GeForce RTX 5060 Ti, covering both MSRP and non-MSRP models. There's apparently a stipulation that NVIDIA's partners (which includes the likes of ASUS, MSI, GIGABYTE, etc.) will have to have at least one MSRP model available on day one for the RTX 5060 and RTX 5060 Ti GPUs.

How does $430 USD sound for a GeForce RTX 4060? What about $1,270 for a Sapphire Nitro+ Radeon RX 7900 XTX or $760 for a Sapphire Pulse Radeon RX 7800 XT? Or, if you're looking for something a little more current, like the brand-new GeForce RTX 5080 - how does $1,540 sound? Yes, the current state of the gaming GPU market is a little depressing as these prices are based on the current best-selling GPUs at the online retailer Newegg.

With NVIDIA launching four of its new GeForce RTX 50 Series GPUs so far in 2025 and AMD rolling out the well-received Radeon RX 9070 and Radeon RX 9070 XT combo, we've not only seen these new cards sell out instantly - most models are premium designs with price points significantly higher than the MSRP. The GIGABYTE Gaming GeForce RTX 5080, available for $1,539.99 at Newegg, is being sold with a 54% markup over the $999 MSRP set by NVIDIA.

It's a similar situation at Micro Center, where select stores are currently selling GeForce RTX 5070s starting from $700 and GeForce RTX 5080s starting from $1400. The bad news is that the inflated prices also affect previous-generation cards like the GeForce RTX 4060.

AMD launched its first pair of RDNA 4 GPUs earlier this month with the Radeon RX 9070 and Radeon RX 9070 XT. However, unlike previous Radeon GPU generations, the company didn't release a reference design, meaning that all cards made available to the public and media were partner models like various RDNA 4 GPUs we've reviewed from GIGABYTE, ASUS, Sapphire, ASRock, and others.

Promo material for RDNA 4 showcases what looks like a reference design with the disclaimer that it was "for presentation only and not available for sale. " It turns out that AMD did create a number of these reference models, including the one for the Radeon RX 9070 pictured here.

A user on the Chinese-based Chiphell forums (via Videocardz) recently showcased the model we see here. The dual-fan design is reminiscent of AMD's reference designs for the RDNA 3-powered Radeon RX 7000 Series. Various applications and benchmarking software also recognize the compact 2.5-slot thick GPU, including 3DMark TimeSpy and Furmark, which points to it being a pre-release sample created by AMD for OEM use.

MSI has just teased a new GeForce RTX 5080 graphics card with an upgraded 24GB of GDDR7, up from the 16GB on the regular RTX 5080, on its own website.

The new GeForce RTX 5080 24GB graphics card is listed on MSI's own website for VGA compatibility for its MAG X870 Tomahawk Wi-Fi motherboard: the RTX 5080 but with 24GB of GDDR7 instead of 16GB. This could be an easy mistake, a typo... but it's not the first time MSI has let slip about an RTX 5080 24GB variant.

In a promotional video, MSI showed off its new GeForce RTX 5080 Vanguard graphics card with "24GB GDDR7" on the side of the retail packaging... yet it was never seen again, and every RTX 5080 has launched with 16GB of GDDR7. MSI also lists the RTX 5090, RTX 5070, and RTX 5070 with their correct VRAM amounts, but the RTX 5080 listing has 24GB with no 16GB variant mentioned, leaning into that belief that this is a typo.