Graphics Cards - Page 105

When NVIDIA announced the Ada Lovelace architecture and the GeForce RTX 40 Series, it mentioned a new bit of technology called Shader Execution Reordering (SER) to improve shader scheduling. This alone could substantially improve performance, making it the perfect tool for real-time ray tracing and path tracing in games.

We've seen the results in the impressive Portal with RTX remaster from NVIDIA, with path-tracing set to arrive in Cyberpunk 2077 sometime in the future. SER, like other technologies, is the sort of thing that needs to be tapped into as part of a game's code and engine.

With DLSS 3 and other RTX technologies coinciding with GDC, NVIDIA has announced that the RTX Path Tracing SDK is now available to all developers.

Only yesterday, we reported on a rumor that the upcoming releases of the more mainstream and affordable GeForce RTX 4070 and GeForce RTX 4060 GPUs could see a shift to the more traditional 8-pin power connectors. A shift away from the new 12VHPWR connectors is seen in the current GeForce RTX 40 Series GPUs line-up.

With the lower power requirements and rumors that some GeForce RTX 4070 models are rated at 200W, it's a move that makes sense. And also one that looks true with packaging for an Inno3D GeForce RTX 4070 leaking out that specifically mentions the use of a single 8-pin power connector.

The image appeared on Chinese social media, shared by @9550pro on Twitter, where we also learned that this specific AIB Inno3D GeForce RTX 4070 features a 2-slot design with two fans.

Raja Koduri has been a key architect behind Intel's recent push to enter the discrete GPU market with the Intel Arc line of graphics cards.

Now it looks like Raja's time at Intel has come to an end with the announcement that he's leaving the company at the end of the month to work at "a new software company around generative AI for gaming, media, and entertainment," according to Pat Gelsinger, CEO of Intel.

In response to the announcement, Raja wrote, "Thank you, Pat and Intel, for many cherished memories and incredible learning over the past five years. Will be embarking on a new chapter in my life, doing a software startup." Confirming that he's making a move away from hardware to software.

The Nintendo Switch will get its successor at some point, though it sounds like Nintendo is in no rush to bring the eventual Switch 2 to the market. The Switch is on track to become one of the best-selling consoles ever. Still, its underlying hardware is based on aging NVIDIA Tegra technology featuring Maxwell-based CUDA Cores. And this is why many wonder what a next-gen Switch might look like.

According to a new bit of info from Korean-based insider @OreXda, the NVIDIA Tegra chip for the Nintendo Switch successor will be built using Samsung's 5nm process - 5LPP, or Low Power Plus. Previous rumors suggested that the custom chip in the Switch 2 would be based on NVIDIA's Ampere architecture - which uses Samsung 8nm or TSMC 7nm - so this would be a big step up.

It's been reported that in the past that a "Switch Pro," using a custom version of the NVIDIA Tegra Orin chip, which is considerably more powerful than current Switch hardware, was in development. Though with the arrival of the OLED refresh, and the age of the Switch, it's looking like Nintendo is moving on to the Switch 2 proper.

NVIDIA's CEO isn't planning to give up his job anytime soon, with Jensen Huang making it clear that his ambition is to be chief executive at Team Green for a long, long time yet.

Huang just turned 60 last month, but that hasn't got him thinking about retirement. Far from it, as in an interview with CNBC, he expressed a desire to continue in his current role for another three or four decades.

That's not enough for Huang, mind. Even though he'll be 90 or 100 at that point, he wants to keep going as CEO beyond that, hoping that the technology will be around to let him become some kind of robot - and stay at the helm of NVIDIA for another 40 years, perhaps. Taking Team Green into the next century, of course...

The ASUS GeForce RTX 4080 Noctua Edition graphics card is the latest collaboration between ASUS and Noctua. The graphics card uses impressive Noctua fans and a revamped heatsink design to improve cooling and offer some quiet high-end gaming.

Most modern GPUs, especially GeForce RTX 4080 cards, use three slim 110mm fans - the ASUS GeForce RTX 4080 Noctua Edition includes two Noctua NF-A12x25 PWM 120mm fans with that classic Noctua brown look. According to Noctua, the heatsink for the RTX 4080 Noctua Edition has been redesigned from the ground up compared to the RTX 3070 and RTX 3080 Noctua Edition cards.

In addition to five as opposed to two 8mm heat pipes, a vapor chamber is added for "significantly better cooling performance than the original TUF Gaming GeForce RTX 4080." And it's here where the price point for this custom RTX 4080 comes into play, with the ASUS GeForce RTX 4080 Noctua Edition being sold for USD 1649 versus the TUF RTX 4080's MSRP price point of USD 1199.

According to a new report by Igor's Lab, the upcoming GeForce RTX 4070 and GeForce RTX 4060 GPUs from NVIDIA could see models launched with the traditional 8-pin power connectors. Until now, all GeForce RTX 40 Series graphics cards have used the new 12VHPWR connector - which has seen all cards ship with 8-pin adaptors.

With the GeForce RTX 4070 and RTX 4060 representing more mainstream options for the Ada Lovelace generation, going the 8-pin route would simplify upgrading or swapping out GPUs in many PC builds.

The report indicates that when it comes to the GeForce RTX 4070, OC cards with a power rating of up to 225W will use the 12VHPWR connector or two 8-pin connectors, with the non-OC 200W models set to use a standard 8-pin connector. And with that, NVIDIA is expected still to adopt the 12VHPWR connector for Founders Edition models.

With releases around the corner, it looks like we've got that final confirmation on the memory makeup for the GeForce RTX 4070 and GeForce RTX 4060 - with an accidental leak coming via GIGABYTE.

It looks like version "23.03.02.01" of the company's GIGABYTE Control Center software for managing motherboards and graphics cards adds support for two new GPUs that haven't been announced. The GIGABYTE RTX 4070 AERO OC 12GB (GV-N4070AERO OC-12G), and the GIGABYTE RTX 4060 GAMING OC 8GB (GV-N4060GAMING OC-8GD).

There's been a little bit of a debate and back and forth relating to the VRAM capacities of what many see as the two mid-range options from NVIDIA, and with this, we now know that the baseline GeForce RTX 4070 will feature 12 GB of VRAM just like the GeForce RTX 4070 Ti. This is good news because, much like its predecessor, the RTX 3070, it's set to be a high-performance 1440p option.

MaxSun has lifted the lid on its new MGG "Mega Gamer GPU" range, a lineup of custom GeForce RTX 4070 Ti and RTX 4080 graphics cards that up the number of fans from the usual two or three to an impressive five. And this is backed up by a funky design with an aRGB light strip or LED matrix display.

These custom cards aren't just a case of increasing the number of fans, and at a glance, they do look like they feature the standard triple-fan design seen in custom GPUs built for cooling and overclocking. You've got two additional - and small - fans on either edge of the heatsink, which we assume would make a difference in cooling the GPU hardware, and you might have to look closely to spot them.

The fans are placed on either side of a stylish ARGB strip or matrix LED display, with the overall look of the GPU featuring a very cool all-enclosed design.

The story of the GeForce RTX 4070 Ti has been well documented, with NVIDIA initially announcing the GPU as the GeForce RTX 4080 12GB before "unlaunching" that and releasing it as the RTX 4070 Ti in January 2023.

This meant that it was the first GeForce RTX 40 Series card to launch without a Founder Edition model because of this change. Even so, we expect to see more FE models for new GeForce RTX GPUs - including the baseline RTX 4070 and the RTX 4060 Ti, as these models got equivalent Founders releases for the Ampere RTX 30 Series generation.

New photos shared by @KittyYYuko confirm that a Founder Edition RTX 4070 Ti was either planned or has been in development, though it's not guaranteed to be released. Perhaps enough time has passed that NVIDIA could skip the RTX 4070 Ti Founder Edition and go straight to the RTX 4060 Ti.

We recently reported that NVIDIA was closing in on 400 games and apps supporting GeForce RTX technologies, with one of the key bits of RTX tech being the AI-powered DLSS - or Deep Learning Super Sampling. With the arrival of the GeForce RTX 40 Series and the RTX 4070 and RTX 4060 coming soon, we've also seen the arrival of DLSS 3 and Frame Generation.

Here's a look at DLSS 3 in the popular Forza Horizon 5.

Here specialized AI hardware in the GeForce RTX 40 Series graphics cards is tapped to create entire frames -like motion plus-like tech in the TV space, except that it works. And doesn't ruin the gaming experience but pushes performance to new heights. Like the baseline DLSS Super Resolution tech, it's continuing to advance and evolve, with NVIDIA already releasing improved versions of what we saw when the GeForce RTX 4090 launched.

It looks like gaming bundles are back on the menu, which is fantastic because it's always great to get a new game when you purchase a brand-new GPU. It used to be the norm back in the day and seemed to disappear until recently, so it's great to see them return.

As per its namesake, NVIDIA's new Redfall Bite Back Edition GeForce RTX 40 Series Bundle will give those that purchase select GeForce RTX 4070 Ti, RTX 4080, and RTX 4090 graphics cards, desktop PCs, or GeForce RTX 4080 and 4090 Laptop GPUs a digital copy of Redfall Bite Back Edition.

Redfall is a new open-world co-op vampire-hunting action game from Arkane Austin, the studio behind the excellent Prey and Dishonored. Redfall is "loaded with NVIDIA tech," as the game supports the AI-powered NVIDIA DLSS 3 rendering and NVIDIA Reflex, with ray-tracing support arriving in a post-launch update.

The PowerColor Liquid Devil Radeon RX 7900 XTX is described as the "newest and most advanced graphics card" in the PowerColor Radeon range - and it's not hard to see why. It features a custom-designed full-cover, pre-installed EK water block that is EK-Matrix7 compatible-a standard developed by EK to ensure perfect alignment and compliance for its range of products.

The PowerColor Liquid Devil Radeon RX 7900 XTX was officially revealed last month, and now with the cards available and in stock in "limited quantities" (via the EKWB store) we also have the price for this beast. As expected, this is pretty much the most expensive Radeon RX 7900 XTX GPU currently being sold, with a price point of USD 1499 and AUD 2099. This puts its price close to the GeForce RTX 4090's MSRP of USD 1599.

It does look fantastic, though, and unlocks the full potential of AMD's flagship RDNA 3 offering. In addition to the nickel-plated copper base and Acrylic, it features a brand new 14-layer high TG PCB, a 17-phase VRM design, and an additional 2+2 phase for the 24GB VRAM. Inside you'll find monitoring and reporting for each phase's electric current and power consumption impressive. The card uses 3x8-pin power connectors that EK notes will be able to support "the heaviest overclocks."

NVIDIA's most recent WHQL-certified GeForce Game Ready Driver 531.18 featured a bug that would cause a 10% spike in CPU usage when closing games, which we reported here. NVIDIA was made aware of the issue fairly quickly, in addition to adding "Higher CPU usage from NVIDIA Container might be observed after exiting a game" to the known issues list.

NVIDIA also released a hotfix within a few days, which was great to see - but as GeForce Game Ready Driver 531.26 was a beta release, it wasn't pushed out to the wider GeForce audience as part of GeForce Experience or the usual official driver download locations. That's changed as NVIDIA has pushed out the official NVIDIA GeForce Game Ready Driver 531.29 WHQL drivers for all GeForce GPU owners.

It resolves the CPU usage issue but also resolves stability issues with Adobe applications. That said, one new issue related to HDR looks to have been added to the list, where turning the effect on and off can cause issues like crashing when a non-native resolution is being used. Plus, it looks like game-specific issues that cropped up with GeForce Game Ready Driver 531.18 for Hogwarts Legacy and Assassin's Creed Origins (which are fairly old now) are still being worked on.

NVIDIA's RTX 4070 graphics card is about to be launched, but if the latest rumor about the GPU is correct, the price will be far higher than expected.

According to a regular provider of GPU leaks on YouTube, Moore's Law is Dead (MLID), NVIDIA might pitch the suggested price (MSRP) of the RTX 4070 at an eye-watering US$750.

Hang on a minute, we hear you cry - the RTX 4070 Ti has an MSRP of US$799, doesn't it? Why, yes, it does, and therefore this purported recommended price for the vanilla RTX 4070 seems way too close for comfort, to put it mildly.

3DMark Time Spy is one of the go-to benchmarks for testing out a new GPU, a simple DirectX 12 test to see just how much extra performance you can get when making the jump to something like AMD's flagship Radeon RX 7900 XTX graphics card.

Well, it looks like the latest Radeon drivers are causing a pretty massive drop in performance for some rocking a new 7900 XTX. A Time Spy result of 12,861 versus 24,587 was reported in one case (seen above, via Osvaldo Doederlein) - with the only difference being the latest Radeon driver from AMD.

The good news is that the bug seems to be rare, with UL Benchmarks noting that only 3% of users seem to be affected. The 3DMark developer was unable to reproduce the issue internally, but AMD had better luck in its investigation.

When NVIDIA launched its new Ada-generation of GPUs with the GeForce RTX 4090 and GeForce RTX 4080 combo, it introduced a new way to connect desktop graphics to a power supply. The 12VHPWR connector, presented as a new standard by Intel, allows more power to be delivered using a single connection on the GPU side.

For many enthusiasts, us included, it was a brand-new thing with NVIDIA adopting the 12VHPWR connector for the entire GeForce RTX 40 Series range. Once the GeForce RTX 4090 made its way into consumers' hands, though, reports of faulty and melted cables began to emerge.

This became one of the most talked about aspects of the GeForce RTX 40 Series launch and led to several external and internal investigations to find the root cause. Some speculated that the 450W power draw of the RTX 4090 was simply too much for a single connector in terms of heat.

Here's an odd one fresh from the rumor mill: GIGABYTE might be working on a new graphics card, except it'll be using last-gen technology, namely RDNA 2 instead of cutting-edge RDNA 3.

The theory advanced by VideoCardz, based on photos of this purportedly incoming model, is that GIGABYTE has an RX 6800 XT in the pipeline, one which is super-sized (and guzzles more power for that matter).

The Radeon RX 6800 XT Gaming OC PRO model is supposedly the best part of 5cm longer than the standard 6800 XT, and VideoCardz observes that it uses the same cooler as the RX 6950 XT Gaming model.

Intel's Arc Alchemist desktop graphics cards are getting cheaper and cheaper across the globe, with one model now hitting US$120 at a major US retailer.

That would be the ASRock Challenger Arc A380 (6GB) graphics card, a lower-end (A3 series) model. As mentioned, this card has now gone as low as US$120 at Newegg, compared to the recommended price of US$140.

That represents a price drop of almost 15%, and at the budget end of the GPU market, savings are always welcome. An extra US$20 can mean a nice upgrade for another component elsewhere in a cheap PC build.

In a new interview with Japanese publication ITMedia (translated via Google), AMD EVP Rick Bergman and AMD Senior VP David Wang open up about why AMD didn't launch an RDNA 3 GPU to compete directly with NVIDIA's flagship GeForce RTX 4090.

The duo notes that it would have been possible to release a Radeon RX 7000 GPU with specs and performance that could compete with NVIDIA on the high-end, but this would push power to the 600W region. Not to mention raise the price to similar RTX 4090-like levels of USD 1600.

"We have chosen not to adopt such a strategy after considering whether it will be accepted by ordinary PC gaming fans," AMD's Rick Bergman says (translated).