Graphics Cards - Page 68

The Navi 48 and 44 GPUs have been confirmed by AMD - kind of - to be the only chips that'll make up Team Red's next-gen range of graphics cards with its RDNA 4 architecture.

VideoCardz noticed that Kepler flagged up on X (formerly Twitter) that in an update for its ROCm software, AMD specifically mentions Navi 44 and Navi 48 as gfx1200 and gfx1201 respectively.

There are no other chips in the update, so that all but confirms that Team Red is keeping things down to just Navi 44 and 48, as all rumors have pointed to for some time now.

On average NVIDIA has released new GPU architecture every two years, but due to the overwhelming demand for new chips to power AI models the company is accelerating its manufacturing schedule and therefore the time between generations.

NVIDIA recently reported its first-quarter earnings where the company revealed it made an astounding $14 billion in profit from its dedicated AI chips, which have been in extremely high demand since the AI boom. Companies such as Meta, Amazon, Google, and OpenAI want NVIDIA's advanced GPUs to train and power their new artificial intelligence models. NVIDIA plans on meeting this overwhelming demand by increasing the frequency it releases new GPU architecture generations.

Notably, NVIDIA released new architecture approximately every two years, with Ampere releasing in 2020, Hopper in 2022, and Blackwell in 2024 (rumored to release at end of the year). But NVIDIA's CEO Jensen Huang announced, "after Blackwell, there's another chip. We're on a one-year rhythm." Analyst Ming-Chi Kuo reported earlier in the month that NVIDIA was planning on releasing its next-generation GPU architecture codenamed "Rubin" in 2025, which would likely produce an R100 AI GPU that will be released sometime next year.

GALAX recently launched a single-slot GeForce RTX 4060 Ti 16GB GPU that turned a few heads because it was so small and compact for a modern GPU in 2024. Now, the hardware maker and NVIDIA partner are back with a pair of 'Classic' branded compact GeForce RTX 4070 SUPER and RTX 4070 Ti SUPER graphics cards.

Like the GeForce RTX 4060 Ti 16GB, these GPUs feature a minimal, almost data center-like all-black design with a single blower fan - a throwback to the GPUs of yesteryear. This is probably why they are being called 'Classic' GPUs.

The dual-slot design makes these stand out, with both variants including the same cooling and physical design. Unlike other blower-fan designs, including GALAX's recent RTX 4060 Ti 16GB model, the heatsink looks to be partially exposed - so it's not a fully enclosed design with all air being pushed through the outtake vents on the rear.

3DMark Steel Nomad is here, and it's free for all users. Check out our news story to see the 4K and Light versions of the new benchmark in action running on a GeForce RTX 4080 SUPER. Steel Nomad is the new Time Spy, a non-raytracing benchmark designed to push PC gaming hardware while showcasing modern visual effects and rendering technologies.

UL Solutions, the developers and publishers behind 3DMark, confirm that the new 4K 3DMark Steel Nomad benchmark is "roughly three times heavier for your PC than Time Spy, and almost seven times heavier than Fire Strike." Like 3DMark Time Spy, it's a cross-platform benchmark with support for Windows and Windows-on-Arm using the DirectX 12 and Vulkan APIs, Android using Vulkan, and iOS using Metal.

Alongside the launch of Steel Nomad, UL Solutions announced that since Time Spy debuted in 2016, the benchmark has been run almost 48 million times (runs from results submitted online). We also learn that average GPU performance has quadrupled in the eight years since Time Spy debuted.

3DMark Steel Nomad is the latest benchmark for joining the 3DMark suite. It's free for all current users and owners, including the free 3DMark Demo version. It's a milestone release from UL Solutions, following the popular Time Spy benchmark that launched in 2016.

3DMark Steel Nomad arrives in two versions: the baseline 3DMark Steel Nomad, which runs natively in 4K and is designed for 'high-end gaming PCs', and 3DMark Steel Nomad Light, which is for lightweight gaming PCs. Steel Nomad is a non-ray-tracing benchmark that includes new technologies like volumetric skies and procedural grass. It also includes advanced ambient occlusion and depth-of-field effects. UL Solutions notes that it's "roughly three times heavier for your PC than Time Spy, and almost seven times heavier than Fire Strike."

Watch the video above, captured with a GeForce RTX 4080 SUPER graphics card, to get a sense of the demo. As you'll notice, it's a definite GPU stress test as the high-end GeForce RTX GPU only manages to deliver frame rates in the 65-70 FPS region.

GeForce Game Ready Driver 555.85 - WHQL is now available. You can download and install it via the NVIDIA App, the GeForce Experience app, or directly from NVIDIA here. The latest driver supports two big new game releases this month: Senua's Saga: Hellblade II and EA SPORTS F1 24.

Both titles support DLSS 3 for upscaling, Frame Generation, and Reflex, with F1 24 also including ray tracing. F1 24 has stepped up its RT action this year, with the latest entry supporting in-race Ray-Traced Reflections, Ray-Traced Shadows, and the new Ray-Traced Dynamic Diffuse Global Illumination to improve the overall lighting when racing at breakneck speeds.

Running on Unreal Engine 5, Ninja Theory's Senua's Saga: Hellblade II is one of the most visually impressive games we've ever seen. It pushes the limits of fidelity while telling a personal, action-packed, and dramatic story. On an Xbox Series X, it only manages to hit 30 FPS upscaled from around 1080p. With a GeForce RTX 4070 SUPER, you can play in 4K with Max Settings and DLSS 3 at triple that frame rate or around 150 FPS with a GeForce RTX 4090.

Intel has released Intel Graphics Driver 31.0.101.5522 (WHQL Certified) for its Arc graphics cards for desktop and mobile Intel Core Ultra processors. It brings day one support for this week's big PC release, the highly anticipated Senua's Saga: Hellblade II, alongside optimizations for Starfield's big May 2024 update and support for Wuthering Waves and Ubisoft's XDefiant.

Starfield performance for Intel Arc GPUs has been something of a disappointment for months, so it's good to see progress being made here - with Intel noting that desktop Arc owners should see an 8% average FPS uplift at 1080p with Ultra settings and a 7% average FPS uplift at 1440p with High settings. It's not a game-changer, like some previous Arc driver releases that have brought massive 100%+ improvements, but welcome nonetheless.

Unfortunately, the driver's release notes still list several outstanding Known Issues related to games like No Rest of the Wicked, DOOM Eternal, and Avatar: Frontiers of Pandora crashing and software like the popular video more upscale Topaz Video AI having issues exporting videos.

I've gotta say, I would never do this myself... but I really do love seeing modded graphics cards like this. NVIDIA's new GeForce RTX 4070 Ti SUPER graphics card has been modded, with 24Gbps and 26Gbps GDDR6X memory.

A competition was recently held in Brazil between the Paulo Gomes and TecLab teams, with the competition involving swapping out the VRAM on the GeForce RTX 4070 Ti SUPER graphics card with faster memory, and overclocking it to its absolute limits.

NVIDIA ships the GeForce RTX 4070 Ti SUPER graphics card with 21Gbps GDDR6X memory modules, with NVIDIA using Micron 21Gbps GDDR6X dies (161-0479-900) were swapped, and modded, with 24Gbps GDDR6X dies (161-0480-900).

Intel teased some more details about its next-generation Falcon Shores CPU at the ISC 24 event, with some rather huge power consumption numbers revealed.

The next-gen Intel Falcon Shores GPU will use up to 1500W of power, with no air-cooling variant planned. Falcon Shores is the true next-generation GPU architecture to Xe, with Falcon Shores originally planned years ago as an XPU th at featured both an x86-based CPU and GPU cores similar to how AMD's new Instinct MI300A APU.

Intel later downgraded Falcon Shores to be a GPU-only design, so it will now take the best parts of Intel's Gaudi AI accelerators, combined with next-gen Xe graphics architecture for HPC and other compute-heavy workloads. Intel's new Falcon Shores GPU will also be backed up by a huge software ecosystem called oneAPI.

Frostpunk 2 is the highly anticipated sequel to the grim and frosty city-builder set to launch on PC and PC Games Pass on July 25. To celebrate, developer and publisher 11 bit studios have teamed up with our friends at SAPPHIRE Technology to create a very cool custom SAPPHIRE AMD Radeon RX 7700 XT Frostpunk 2 Edition.

This looks to be a special edition of the Sapphire PURE AMD Radeon RX 7700 XT that we reviewed late last year, with a similar frosty-white color scheme to match the game's icy setting with a custom "oil-black metal backplate with rusty accents" to add some steampunk grit. Even the ARGB lighting is given a Frostpunk makeover and set to 'Orange' by default.

The special edition Radeon RX 7700 XT GPU comes with a few Frostpunk-themed extras - three enamel pins, three keyboard keycaps, a keycap puller, and a PC code for the Deluxe Edition of Frostpunk 2.

AMD has released Adrenalin Edition 24.5.1 for Radeon GPU owners. It adds support for three major PC game releases: Ghost of Tsushima Director's Cut, Senua's Saga: Hellblade II, and F1 24.

Ghost of Tsushima and Hellblade II are two of the biggest PC game releases this month, so make sure you update to the new Adrenalin Edition 24.5.1 driver if you plan on picking one or both of these up. The new driver also adds four games to the supported HYPR-Tune titles list, enabling AMD HYPR-RX frame generation, Radeon Super Resolution, and Anti-Lag. These are Ghost of Tsushima, Horizon Forbidden West, Starfield, and The Last of Us Part I.

The new driver also improves AI support and performance with Radeon hardware, with AMD citing "Adobe Creative Cloud, Blender, DaVinci Resolve, Topaz Photo & Video AI, and many more" - though it doesn't provide any specific driver-related performance data.

It sounds like AMD's next-gen Radeon graphics cards powered by the new RDNA 4 architecture are right around the corner. Although rumors and months of speculation have pointed to AMD skipping high-end GPUs that will compete with the GeForce RTX 5080 and RTX 5090 (or the current RTX 4090, for that matter), RDNA 4 could still present an exciting option for the mid-range market.

With a massive improvement in ray-tracing coming and a more efficient GPU, a potential Radeon RX 8800 XT with the right price could be a real competitor to the popular GeForce RTX 4070 SUPER. With AMD's Linux developers adding nearly 20,000 lines of code for 'AMD GFX12 Mesa support' for its drivers, an RDNA 4 reveal and release could be right around the corner.

Linux kernel driver support for RDNA 4 (or GFX12) is a definite sign that things are progressing well for a next-gen Radeon release this year. Depending on the timeline, we could be getting a big reveal at Computex 2024 in the coming weeks.

According to WJM47196, a source of AMD leaks over on the ChipHell forums (via Videocardz), AMD's next-gen RDNA 4 GPU line-up will "merely be a bug fix for RDNA 3," bringing improvements to ray-tracing and incremental updates in other areas. Without a flagship, the top RDNA 4 card is expected to deliver raw performance, albeit more efficient performance, on par with the Radeon RX 7900 XT.

The source claims that RDNA 3, which powers the current Radeon RX 7000 Series, "didn't meet its performance expectations," leading to problems with the core frequency power curve. Throw in the R&D costs associated with developing a multi-chip GPU, a fast on-chip interconnect, and dealing with related issues - RDNA 4 is a stopgap before the main course - RDNA 5.

RDNA 5 "will be a clean sheet design," an overhaul of AMD's Radeon architecture similar to the arrival of Zen and Ryzen. The difference is so big that RDNA 5 could ditch the Radeon branding for something completely new.

MSI and Capcom announced and showcased limited edition Monster Hunter-themed PC gear earlier this year at CES 2024. The line-up of PC hardware included a gaming laptop, GeForce RTX 4060 Ti GPU, a custom gaming controller, case, motherboard, monitor, and liquid cooler. It's quite the line-up, with the MSI GeForce RTX 4060 Ti 8G GAMING SLIM Monster Hunter Edition GPU now available.

The limited edition graphics card is bundled with the custom Monster Hunter controller, both sporting unique designs inspired by the franchise to help celebrate its 20th anniversary. The custom GeForce RTX 4060 Ti 8GB GPU looks very cool with red, gold, and black flourishes to mimic Rathalos - one of the game's giant monsters.

The MSI GeForce RTX 4060 Ti 8G GAMING SLIM Monster Hunter Edition sports a thin 2-slot design and MSI's TRI FROZR 3 cooling system, which includes the company's latest TORX FAN 5.0, core pipe, and airflow setup. However, as a limited edition bundle that also features custom packaging, the price tag is much higher than the GeForce RTX 4060 Ti 8GB's MSRP of $399 USD. It's currently listed for $499.99 USD at Newegg.

MSI has introduced a new GPU with the GeForce RTX 4070 Ti SUPER 16G SHADOW 3X OC, which looks like a darker version of the Ventus 3X reference model we reviewed when the GPU launched in January 2024. The Shadow 3X sports an out-of-the-box boost clock speed of 2640 MHz (which can be pushed to 2655 MHz via MSI Center), so it should perform slightly better than the non-OC Ventus variant.

The big difference here isn't just the color, which takes the silver of the Ventus line-up and switches it out for a darker bronze or brown metallic look. The MSI GeForce RTX 4070 Ti SUPER 16G SHADOW 3X OC is thinner, 43mm versus 50mm, and lighter, 838 grams versus 1107 grams.

Being thinner and lighter but retaining the same look as the Ventus 3X, MSI's new Shadow line could be the series' next evolution. It represents MSI's most affordable range of GeForce RTX graphics cards.

With a planned release for later this year, rumors, speculation, and insider leaks about the upcoming GeForce RTX 50 Series graphics cards are ramping up. The most recent rumors suggest that the GeForce RTX 5080 and RTX 5090 will be released within weeks of each other, much like the initial launch of the Ada Lovelace GeForce RTX 40 Series in 2022.

Built using NVIDIA's Blackwell architecture, the upcoming GeForce RTX 4080 is said to use the new GB203 GPU, while the flagship GeForce RTX 5090 is set to use GB202. Based on leaked or rumored specs, there will be a significant disparity between the two - with GB203 set to include 96 SMs compared to GB202's 192 SMs. Memory bandwidth will be 256-bit on the GeForce RTX 5080 compared to 512-bit on the GeForce RTX 5090.

Although the GeForce RTX 5080 is expected to be more powerful than the current GeForce RTX 4080 SUPER graphics card, the US export ban on powerful GPU hardware could be the reason for the GeForce RTX 5080 launching in such a 'cut down' state.

NVIDIA is reporting launching its next-generation GeForce RTX 5080 graphics card first, and then weeks later, the monster GeForce RTX 5090 will be unleashed.

That's according to leaker "kopite7kimi" who posted on X saying "RTX 5080 should be released first" to which "AGF" replied, saying "to be exact, they should be announced at same time with availability separated by (a) few weeks".

I can't see NVIDIA unveiling its GeForce RTX 5080 first and then the GeForce RTX 5090 second, the launch would be the both of them at the same time with availability splitting over a few weeks making sense. I'm sure that NVIDIA will want to make sure it has plenty of GeForce RTX 5080 graphics cards as they'll sell more units of those cards over the ultra-high-end RTX 5090.

AYANEO has just teased its upcoming external GPU solution: AG01 "Graphics Starship" will be officially unveiled on May 18. First, the tease:

The new AYANEO external GPU dock being called "Graphics Starship" is really cool, as it definitely looks like an old-school 90s-themed sci-fi ship we've seen in the movies. We don't know anything at all about the AYANEO Graphics Starship, other than it's a GPU dock for your graphics card.

We could have USB4, Thunderbolt 4, or OCulink connectivity... but AYANEO does state that it's a "multi-function graphics dock" so we should expect some great functionality out of it. AYANEO says it's an external graphics card dock, but we can't see a PCIe x.0 x16 interface, so we should expect just a few particular graphics cards that will be sold, or will work with, the AYANEO AG01 "Graphics Starship" external GPU dock.

NVIDIA is reportedly experimenting with a huge triple-fan, quad-slot, 600W-capable cooling solution for its upcoming next-generation GeForce RTX 50 series GPUs.

In a new report from Benchlife, the outlet explains: "According to information from a familiar cooling module manufacturer, NVIDIA has already conducted relevant tests and verification on cooling modules for the Blackwell GPU architecture's GeForce RTX 50 series graphics cards, evidently preparing for the launch of the GeForce RTX 50. Although there's no specific timeline, currently, about 4 plans are underway, with power ranging from 250W to 600W".

We did hear all the way back in late 2022 that NVIDIA had prototype versions of the GeForce RTX 4090 -- or RTX 4090 Ti or RTX 4090 SUPER -- with a gigantic triple-fan, quad-slot, 600W cooler. Fast-forward two years, and we're hearing the same thing leading up to the launch of the RTX 50 series.

Two Chinese companies have teamed up for an external GPU dock that supports OCulink -- and even features a 550W PSU -- check it out:

Metaphyuni and Gzosmeta collaborated on what's called the EG01C, which is a new external GPU solution that has a built-in 550W power supply. Regular external GPUs rely on fast USB4 or Thunderbolt 4 connectivity, or they use a standard called OCulink. OCulink has up to 64GB/sec of bandwidth, which is far better for external graphics cards over USB4 and Thunderbolt 4 which have just 40GB/sec of bandwidth at the ready.

Enthusiasts, performance lovers, gamers, and AI researchers are flocking to OCulink because of the increased bandwidth, that is, until Thunderbolt 5 launches... then the external connector game gets real fun. The companies made a newly-designed OCulink connector, which is capable of being plugged in, and unplugged over 10,000 times. On top of that, instead of the regular manual locking mechanism, this external GPU solution features an automatic one that fixes the cable in place. Very nice touch.