Graphics Cards - Page 64

ASRock has launched a pair of new Radeon GPUs as part of its new Creator Series - the ASRock Radeon RX 7900 XTX Creator 24GB and ASRock Radeon RX 7900 XT Creator 20GB graphics cards. The compact two-slot thickness and blower-fan cooling (with bonus Vapor Chamber) indicate that these are designed for "multi-card parallel computing," aka running AI workloads.

The 7900 XT and 7900 XTX are the flagship gaming RDNA 3 GPUs for AMD's Radeon RX 7000 Series. They are 4K-capable cards capable of running the latest PC games. What makes the new Creator Series from ASRock interesting is that it turns them into workstation-first cards. That, and they both come equipped with a single horizontal 12V-2x6 power connector.

The 12V-2x6 power connector is the received and updated version of the original 12V HPWR connector that debuted with the GeForce RTX 40 Series. It was created to minimize issues with cables and connectors overheating. The revised connector can be found in newer GeForce RTX 40 Series GPUs, and its introduction here is a first for a Radeon RX 7000 Series GPU.

At its most basic level, Compute Unified Architecture (CUDA) allows general-purpose processing and other tasks to run on NVIDIA GPUs with extensive language support. Since its inception, CUDA has been a game changer. It has been used to tap into the power offered by GPUs for advanced simulation, physics calculations, scientific modeling, crypto mining, and, yes, machine learning.

The latter, and the current AI boom, have been one of the main reasons NVIDIA GPUs have been the go-to for generative AI workloads - developers and researchers alike have been using the CUDA toolkit for years. Widely viewed as the industry standard, this has left GPUs from competitors like AMD looking elsewhere for its software.

That could change as a new startup called Spectral Compute has unveiled SCALE, a toolkit that allows CUDA applications to be natively compiled for AMD's RDNA GPUs. According to its creators, "SCALE does not require the CUDA program or its build system to be modified."

AMD's next-generation RDNA 4 GPU architecture is still in the ovens, but we're hearing rumors of RDNA 4's new and enhanced Ray Tracing engine for Radeon RX 8000 series graphics cards, and Sony's beefed-up PlayStation 5 Pro console.

In a new post on X, leaker "Kepler" said that some of the new RT features coming with RDNA 4 will "most if not all" be found inside of the PS5 Pro console, not just RDNA 4-powered Radeon RX 8000 series GPUs. Kepler shared some details of the ray tracing enhancements expected with RDNA 4:

AMD will not be beating NVIDIA's next-gen Blackwell-based GeForce RTX 50 series GPUs with the likes of the GeForce RTX 5090 and GeForce RTX 5080 to be next-gen performance powerhouses. AMD will compete under those GPU power levels with its RDNA 4-based Radeon RX 8000 series GPUs, with enhanced ray tracing (RT) performance and power efficiency being the key features included with RDNA 4.

AMD's new Ryzen AI 300 series "Strix Point" APUs have an upgraded integrated RDNA 3.5-based Radeon 880M GPU, which is 15% faster than the RDNA 3-based GPU inside of the Radeon 780M, according to ASUS.

The new Ryzen AI 9 HX 370 and Ryzen AI 9 365 "Strix Point" APUs have new RDNA 3.5-based integrated GPUs. The Ryzen AI 9 HX 370 APU features 16 x RDNA 3.5-based Compute Units (CUs), while the Ryzen AI 9 HX 365 features 12 x CUs.

The new Radeon 880M integrated GPU inside of the new Strix Point APUs is up to 15% faster than the Radeon 780M in gaming performance according to ASUS. ASUS is currently promoting its new Vivobook laptop, which is powered by AMD's new Ryzen AI 300 series "Strix Point" APUs, giving us our first look at what to expect from the upgraded RDNA 3.5 GPU.

GeForce Game Ready 560.70 WHQL driver has been released, adding day-one support for three new games: Dungeonborne, Flintlock: The Siege of Dawn, and Stormgate. In addition to this, it also adds support for 14 new G-SYNC Compatible gaming displays from Acer, ASUS, LG, and more.

Kicking off with game support, Stormgate is a highly anticipated free-to-play real-time strategy game from Frost Giant Studios, a team of veteran developers who worked on iconic RTS titles like Warcraft III and StarCraft II. Built with Unreal Engine 5, multiplayer, and a cinematic campaign, the game will go live on August 13, with those who pre-purchase being able to jump in on July 30.

Next is Dungeonborne, a new first-person PvPvE dungeon crawler with impressive ray-traced visuals launching into Early Access on July 19. GeForce RTX owners will benefit from optimized day-one driver support with the addition of DLSS 2 Super Resolution upscaling and DLSS 3 Frame Generation.

GPU images and GPU-Z screenshots have emerged for the GeForce RTX 4070 10GB that could have been. The unreleased prototype from NVIDIA features 10GB of VRAM on a slower 160-bit memory interface. The 'for reals' GeForce RTX 4070 launched in April 2023 and features 12GB of VRAM on a 192-bit interface, with a total memory bandwidth that is 20% faster than what we see here.

However, the AD104-275 chip in this unreleased GeForce RTX 4070 10GB prototype features 7,168 CUDA Cores - more than what is found in the actual GeForce RTX 4070 (5,888 CUDA Cores) and in line with the GeForce RTX 4070 SUPER that was released earlier this year.

Interestingly, this prototype also features faster Base and Boost Clock speeds compared to the retail GeForce RTX 4070 - 2355/2520MHz versus 1920/2475MHz.

With U.S. export restrictions on GPUs hitting a certain performance level for AI, the flagship GeForce RTX 4090 is no longer officially available in China and has been replaced with the GeForce RTX 4090D. For PC gaming, there's not much separating the two; however, this raises questions about NVIDIA's plans for the upcoming Blackwell generation.

It's all but 'officially confirmed' that the next-gen GeForce RTX 5090 will be considerably more powerful than the RTX 4090 - with the question now being by how much, 30%, 40%, 80%? According to insider hongxing2020 on X, NVIDIA has its next-gen GeForce RTX 5090D ready and will release the GPU sometime in January 2025.

It's unclear what the configuration here will be or how it would differ from the baseline GeForce RTX 5090. On that front, NVIDIA could still be working on finalizing the GeForce RTX 50 Series specs. That said, any RTX 5090D would kind of have to be a PC gaming powerhouse - but without the raw AI performance.

ASUS has updated its entry-level GeForce RTX 40 Series card design with the new ASUS Dual GeForce RTX 4060 V3 8GB. As the naming suggests, this is the third version of the card. As the most efficient GPU in NVIDIA's Ada Lovelace line-up, the redesign sports smaller axial fans in a more compact two-slot build.

The look has also changed compared to the more sci-fi transparent shroud of earlier Dual models that were also 2.5 slots thick. The V3 looks plain and simple, with a slightly curved shape and an all-black finish. The specs and requirements remain unchanged, so performance should be on par with previous Dual designs in a smaller, more compact package. 3072 CUDA Cores, a 2490 MHz OC-mode Boost Clock, and 8GB of GDDR6 memory on a 128-bit interface.

It's a simple design that befits an entry-level mainstream GPU. It's similar to the recent ASUS Dual Radeon RX 6600 V3 8GB. However, it removes the white text and graphics.

NVIDIA's next-gen GeForce RTX 50 series "Blackwell" gaming GPUs have had their TDPs leaked by PSU maker Seasonic, with the monster ultra-enthusiast GeForce RTX 5090 expected to have a 500W TDP.

In the leak, Seasonic details the RTX 5090 with a 500W TDP, and the second-highest Blackwell GPU -- the RTX 5080 -- with a 350W TDP. We also have TDP details on the RTX 5070 with 220W, RTX 5060 with 170W TDP, and the RTX 5050 with a 100W TDP. These are preliminary TDP details, nothing final, just leaks... just so you know.

NVIDIA's new ultra-enthusiast GeForce RTX 5090 with a 500W TDP means it has an 11% higher TDP than the current Ada Lovelace flagship, the RTX 4090, with its 450W TDP. We should expect the new GeForce RTX 5090 to boost even higher in overclocked and custom RTX 5090 form, past 500W.

Been wondering how Intel Arc GPUs are faring these days in terms of game compatibility? Well, you're in good company, as Hardware Unboxed on YouTube certainly has, and they've tested every game they own with an Arc A770 graphics card, in fact.

So, how did this experiment (flagged up by VideoCardz) go? The good news is that Intel's Alchemist flagship GPU managed to run the vast majority of PC games in Hardware Unboxed's library (using that latest graphics driver, of course).

This was an extensive test of 250 games, and only 2% of them failed to work at all, while a further 5% weren't playable (due to performance issues or other wonkiness). 93% were found to be playable, which means running with at least 40 frames per second and no glitches or artefacts and the like.

Intel has released a new Non-WHQL driver for discrete Intel Arc GPUs and built-in Arc Graphics for its Intel Core Ultra CPUs. Intel Arc Game On Graphics Driver version 32.0.101.5762 adds support for a couple of new PC game releases, Once Human and Zenless Zone Zero. The latest Beta driver also significantly increases the performance of Blizzard's Diablo IV.

Intel Arc graphics card owners can expect up to 30% faster performance at 1080p and up to 35% faster performance at 1440p when playing the game using the Ray Tracing Ultra preset. Diablo IV featured ray-traced shadows and reflections, so there's a performance hit when enabling them, which makes this new driver a game changer for Diablo fans with Intel Arc graphics cards.

The new driver release fixes a few lingering issues; however, there's still a long list of known issues for both Intel Arc GPUs and the Arc Control software. Here are the full release notes.

The Sphere, located in Las Vegas, is an impressive feat of engineering-a massive circular dome that is also the world's largest display. The Sphere's LCDs cover 750,000 square feet inside and outside the entertainment venue. Not only that but with a resolution of 16,000 × 16,000 pixels for the interior display, it's the highest-resolution screen in the world. So then, what sort of hardware do you need just to 'press play' on a video with that resolution?

According to NVIDIA, the 16K x 16K 60 frames per second display for the Sphere's interior is powered by 150 NVIDIA RTX A6000 GPUs - valued at around $1 million. The NVIDIA RTX A6000 is a powerful workstation card with 48GB of GDDR6 memory, 10,752 CUDA Cores, 336 Tensor Cores, and 84 RT Cores. Put 150 of these together, and you've got some serious GPU horsepower, enough that they also power the 1.2 million programmable LED pucks on the Shere's exterior.

In addition to the hardware, the Sphere also uses NVIDIA's network connectivity technology to ensure that all GPUs work together with minimal latency, plus the NVIDIA Rivermax SDK framework for media streaming. However, with this level of hardware and 7.2 TB of VRAM, we'd love to see some PC games running on the Sphere.

AMD has revealed detailed information on its new technology, designed to decrease the storage space that games require. AMD Neural Texture Block Compression, or NTBC for short, compresses high-quality textures without impacting overall performance and maintaining 'reasonable quality.' AMD claims that NTBC can reduce a game's storage footprint by up to 70%, which would shrink the installation size of Call of Duty from 150GB to just 45 GB.

Game sizes are increasing as fidelity increases, and even though 'block compression' is used to reduce the overall size of textures, the fixed compression ratio can only do so much when dealing with hundreds or even thousands of high-quality 4K textures. As the name suggests, AMD Neural Texture Block Compression adds neural network-based image compression into the mix.

AMD's Shin Fujieda and Takahiro Harada have released the full technical paper on Neural Texture Block Compression and how it works.

Black Myth: Wukong is one of the most anticipated game releases of 2024, and it's right around the corner - launching on August 20. Today, NVIDIA has announced a new Black Myth: Wukong GeForce RTX 40 Series Bundle, offering PC gamers a steam copy of the game with eligible GPU and Laptop products.

On the desktop GPU side, the bundle covers everything from the GeForce RTX 4070 and higher: so the RTX 4070, RTX 4070 SUPER, RTX 4070 Ti, RTX 4070 Ti SUPER, RTX 4080, RTX 4080 SUPER, and RTX 4090. Anything that isn't a GeForce RTX 4060 or RTX 4060 Ti - two of the most popular graphics cards for the current generation.

The laptop side of the bundle is the same. Eligible systems are those with an RTX 4070 Laptop GPU or higher. So, what gives? Well, it looks like Black Myth: Wukong's impressive ray-tracing visuals will require at least a GeForce RTX 4070 to hit a playable frame rate.

NVIDIA's next-generation ultra-enthusiast GeForce RTX 5090 graphics card is rumored to have a base clock that will be near 2.9GHz, meaning it will trounce every other graphics card on the market when it arrives later this year.

The new rumor is coming from Chiphell forum user "Panzerlied," who has previously provided leaks on the GeForce RTX 50 series "Blackwell" GPUs. In the post, the user said that the "base frequency of 5090 is almost 2.9G" which has everyone excited.

NVIDIA's current-gen flagship GeForce RTX 4090 has a base GPU clock of 2235MHz, so if the RTX 5090 rolls out with a 2.9GHz+ base clock, we're looking at a 30% improvement in base GPU clocks alone. Add on that NVIDIA will be using next-generation GDDR7 memory that Micron has teased will provide 30% gains in gaming in both rasterization and ray tracing, we're in for quite the performance leap with the RTX 5090.

AMD hasn't officially revealed any information about its next-generation Radeon graphics cards for PC gaming and desktop computers. However, we've heard rumors and rumblings about what to expect for over a year. The most recent gives us a time frame for when to expect the first Radeon RX 8000 Series product to debut - CES 2025 in January, where AMD will formally lift the lid on the flagship Navi 48 GPU.

Navi 48 is the codename for the upcoming flagship RDNA 4 GPU, much like how the Navi 31 GPU was the flagship RDNA 3 chip found inside the Radeon RX 7900 XTX and Radeon RX 7900 XT. Based on how discrete GPUs are created, Navi 48 will power at least one or two different products - the upcoming Radeon RX 8800 XT and Radeon RX 8700 XT (product naming hasn't been confirmed or leaked).

Yes, this means there won't be a flagship Radeon RX 8900 XT or 8900 XTX equivalent this time - with AMD skipping a high-end enthusiast flagship model for its next-gen Radeon line-up. So what sort of performance can we expect, what kind of hardware configuration, and when will the mainstream RTX 3060 or RTX 4060 competitor arrive? Let's dig in.

Recently, we reported on GeForce RTX 4070 Ti SUPER GPUs from MSI and Colorful (and a few others) starting to appear with the AD102 silicon, the same GPU chip found in the flagship GeForce RTX 4090. Of course, there's a massive difference between the 18,432 CUDA Core count in the GeForce RTX 4090 and the 8,448 in the GeForce RTX 4070 Ti SUPER.

For these new GeForce RTX 4070 Ti SUPER graphics cards using the AD102 chip, nearly half of the GPU is disabled to meet the spec requirement, pointing to this being NVIDIA using stock that wasn't ready for the RTX 4090 and RTX 4090D. This is just a formality for the everyday consumer, with the RTX 4070 Ti SUPER using a non-standard AD102 GPU instead of AD103. Performance remains the same, albeit with a higher 10W power rating.

It turns out there might be quite a few of these variants coming as ZOTAC GAMING has announced a new GeForce RTX 4070 Ti SUPER SOLID GPU available in OC and non-OC variants with the cutdown AD102 GPU.

ASRock has launched new Intel Arc graphics cards, the Intel Arc A770 Challenger SE 16GB OC and Intel Arc A750 Challenger SE 8GB OC, sporting updated cooling designs and a slightly smaller form factor. These SE or Second Edition GPUs build on the existing Challenger models released by ASRock last year, with the main changes revolving around the cooler and physical build, so performance is expected to remain the same.

Both are compact dual-fan GPUs, and the new dimensions (267mm x 130mm x 48mm) make them roughly 2-3% smaller than the previous variants. The overall look is in line with previous models; however, now, it has added RGB lighting. And much like it does with its Phantom Gaming cards, ASRock includes a physical switch so you can turn the lighting on or off directly from the GPU.

Additional changes include more vents on the metal backplate for better heat dissipation and improved cooling. Elsewhere, specs like clock speeds remain unchanged, though the new display configuration across both is now 3 x DisplayPort 2.0 and 1 x HDMI 2.0b.

Intel's next-generation Arc Battlemage "Xe2" discrete GPUs will reportedly be made on TSMC's newer 4nm process node, according to new leaks.

In a new post from DigiTimes, the outlet reports that TSMC's 4nm process node will be used for Intel's upcoming Arc Battlemage "Xe2" discrete GPUs, graphics cards to compete against AMD's upcoming Radeon RX 8000 series "RDNA 4" graphics cards and NVIDIA's next-gen GeForce RTX 50 series "Blackwell" graphics cards.

The current-gen Arc Alchemist "Xe" GPUs are made on the 6nm process node at TSMC, so when Intel is promising up to 50% more performance, we should expect some good efficiency out Battlemage "Xe2" GPUs, but they won't be competing with the GeForce RTX 5090 or RTX 5080, or even AMD's new higher-end RDNA 4 GPU offerings.

AMD is reportedly gearing up to unveil its next-generation Radeon RX 8000 series "RDNA 4" GPUs in 2025, with new leaks suggesting the higher-end Navi 48 GPU will be unveiled at CES 2025 in January, while the entry-level Navi 44 GPU will be unleashed in Q2 2025.

We have been hearing all about NVIDIA's next-generation GeForce RTX 50 series "Blackwell" GPUs for months and months now, but it seems that AMD won't have anything to fight Blackwell in 2024, and we'll be waiting until 2025, which lines up with previous leaks. In a new post on X, leaker "Kepler_L2" said that "CES is for N48. N44 is probably Q2".

AMD's next-gen Navi 48 GPU should offer performance close to what we already have with the Navi 31 GPU that powers the Radeon RX 7900 XTX and RX 7900 XT graphics cards, but with upgrades like improved ray tracing (RT) performance and more.