Graphics Cards - Page 60

AMD has been kicking goal after goal with its CPU business, but the GPU side of the company has been struggling for many, many years now... and now, there's a new issue: oversupply of RDNA 3 graphics cards, so RDNA 4 is being pushed back to Q1 2025 according to new rumors.

In a new video from leaker Moore's Law is Dead, we're hearing that AMD is pushing RDNA 4 to Q1 2025 as there is an "oversupply problem" and that "RDNA 4 is MUCH better" in price/performance, and with too many Radeon RX 7000 series "RDNA 3" graphics cards on the market, it means RDNA 4 is being pushed.

A couple of sources talked to MLID, confirming a "HUGE" oversupply problem with RDNA 3 and that low-to-mid RDNA 3 graphics cards are selling "OK" but the higher-end Navi 31-based GPUs aren't "moving at all right now".

For those looking forward to discovering more about the next generation of GeForce RTX 50 Series cards, it's been a wild week of leaks from trusted insiders and sources. NVIDIA leaker @kopeite7kimi on X provided detailed specs for the GeForce RTX 5090 and the GeForce RTX 5080, which are expected to get a full reveal and launch early next year.

After rumors pointing to these new flagship GPU specs getting 'locked in' sometime in September 2024, the probability of this information being true is high. As reported earlier, the GeForce RTX 5090 will be an absolute beast - 21,760 CUDA Cores, 32GB of GDDR7 memory on a 512-bit bus, and a confirmed 600W power draw or rating.

The Blackwell flagship is also built using the GB202-300-A1 chip. On the other hand, the GB203-400-A1 GeForce RTX 5080 will be a little more down-to-earth, with 10,752 CUDA Cores, 16GB of GDDR7 on a 256-bit bus, and a 400W power rating.

NVIDIA is expected to debut its next-gen GeForce RTX 5080 with 16GB of ultra-fast, next-gen GDDR7 memory... but now we're hearing fresh rumors that an RTX 5080 will launch with 24GB after the first 16GB model launches. Alrighty!

The new flagship GeForce RTX 5090 will ship with a huge 32GB of GDDR7 memory, which is 8GB more than the 24GB of GDDR6X memory on the RTX 4090. The GeForce RTX 4080 launched with 16GB of GDDR6X memory, and the suped-up RTX 4080 SUPER also continued on with 16GB of GDDR6X... now the RTX 5080 is launching with 16GB and I'm sure gamers are going to say "but, there's no VRAM capacity upgrade!"

Well, that's where this new RTX 5080 with 24GB GDDR7 memory steps in, which would be fantastic to see as there's going to be a huge gap between the RTX 5090 with 32GB and the RTX 5080 with 16GB. The RTX 5080 with 24GB would sit perfectly in the middle, and I'm sure would rock higher GPU clocks, the additional 8GB more GDDR7 memory, and of course: a higher price tag.

NVIDIA's next-gen GeForce RTX 5090 and GeForce RTX 5080 graphics cards are rumored for a (huge) debut at CES 2025, with some major new rumors filling out the picture of what to expect from the Blackwell gaming GPUs.

We had some specs on the GeForce RTX 5090 and GeForce RTX 5080 teased yesterday with rumors, but new rumors are suggesting that we'll have a CES 2025 launch... and some more hardware details. NVIDIA is expected to debut a brand new PCB design and new Founders Edition coolers which run on a single 12V-2x6 power connector, squashing rumors of dual PCIe power connectors on the RTX 5090.

The upgraded PCB will house 16 GDDR7 memory modules, each with 3GB of memory capacity and running at speeds of somewhere between 28-32Gbps, which will result in the highest memory bandwidth on a consumer graphics card in history: 32GB of GDDR7 on a 512-bit memory interface -- if run at 32Gbps GDDR7 spec -- will provide an insane 2.0TB/sec of memory bandwidth. Wow, wow, wow. Do want.

Intel Arc Battlemage graphics cards for desktop PCs are expected to arrive in early 2025 and offer a compelling option for the budget-conscious and those looking for a powerful mid-range GPU. At least, that's the idea or hope.

With the new Xe2-LPG 'Battlemage' graphics architecture appearing in new Luna Lake Core Ultra processors for mobile devices, the only question is how desktop performance stacks up.

Wccftech recently spotted a Geekbench OpenCL benchmark result with a GPU featuring one of the Intel Arc Battlemage device IDs spotted in Linux Kernel drivers and shipping manifests. Multiple Battlemage cards are on the way, and this benchmark result covers the Bettlemage 'G21' variant.

One of the biggest PC game launches in the past week has been the long-awaited debut of Sony and PlayStation's God of War Ragnarok. The game arrives with optimizations and support for technologies like DLSS 3 and FSR 3.1. Another big PC release is Frostpunk 2, another sequel and a visually stunning strategy game supporting AMD and NVIDIA upscaling and frame generation technologies.

If you own a Radeon GPU and are playing either of these games, you'll want to download and manually install the latest AMD Software: Adrenalin Edition 24.20.11.01 preview driver. This driver adds New Game support for God of War Ragnarok, Frostpunk 2, and The Sims 4 DirectX 11 Update.

The driver also adds HYPR-Tune support to automatically enable FSR 3 and the Anti-Lag 2 latency reduction tech in games like Black Myth: Wukong, Creatures of Ava, and God of War Ragnarok. The new preview driver also fixes issues with Warhammer 40,000: Space Marine 2 crashing for Radeon gamers. Here's a look at the full release notes.

MSI has just teased its new GeForce RTX 4070 Ti SUPER GAMING Slim "STALKER 2" Edition graphics card, check it out:

MSI has been working directly with STALKER 2 developer GSC Games on designing its GAMING Slim "STALKER 2" Edition graphics card, joining the ranks of MSI's other custom designs including Black Myth: Wukong and World of Warcraft. The new MSI GeForce RTX 4070 Ti SUPER GAMING Slim "STALKER 2" Edition graphics card features a beautiful custom design, new stickers on the fans, and an upgraded backplate.

Inside, we should expect the AD103-275 GPU with 8448 CUDA cores, 16GB of GDDR6X memory, and a 285W TDP. The custom variants from MSI don't usually include "X" or "OC" specifications, so there is now factory overclocking here... but I'm sure you'll squeeze some more performance out of the GeForce RTX 4070 Ti SUPER GAMING Slim "STALKER 2" Edition graphics card.

We've got some new leaks regarding NVIDIA's new GeForce RTX 5090 and RTX 5080 graphics cards -- you can read about the monster RTX 5090 here -- but the new RTX 5080 is poised to have 10K+ CUDA cores and 16GB of ultra-fast GDDR7 memory.

In a new post on X, leaker "kopite7kimi" teased that NVIDIA's second-fast Blackwell gaming GPU, the GeForce RTX 5080, will have the GB203-400-A1 GPU with 10,752 CUDA cores (51% less than the 20K+ on the RTX 5090). We can expect 16GB of GDDR7 memory on a 256-bit memory bus, where if NVIDIA uses 28Gbps GDDR7 we'll see 896GB/sec memory bandwidth, or with 32Gbps GDDR7 we'll get 1.0TB/sec memory bandwidth out of the RTX 5080.

NVIDIA's upcoming GeForce RTX 5080 is rumored with a 400W TDP, which I'm sure is for the Founders Edition, so we'll see custom AIB models of the RTX 5080 exceeding that (and with higher GPU clock speeds, improved thermal designs, etc.

NVIDIA's next-generation GeForce RTX 5090 graphics card has some juicy leaked specs: with a monster of a difference between the RTX 5090 and RTX 5080 graphics cards. The new RTX 5090 has 20K+ CUDA cores, 32GB GDDR7 memory, and will chow down on up to 600W of power.

In a new post on X, leaker "kopite7kimi" teased that NVIDIA's new flagship GeForce RTX 5090 graphics card will feature a GB202-300-A1 GPU based on the new Blackwell architecture, an insane 21,760 CUDA cores, 32GB of next-gen GDDR7 memory on a 512-bit memory bus, and will consume up to 600W of power.

This means that the new GeForce RTX 5090 will have an insane 1.79TB/sec or 2.0TB/sec of memory bandwidth, depending on what bandwidth the GDDR7 memory is: 28Gbps GDDR7 on a 512-bit memory bus will result in 1.79TB/sec of memory bandwidth, while the faster 32Gbps GDDR7 on a 512-bit memory bus will push a higher 2.0TB/sec of memory bandwidth. Bonkers, totally bonkers.

A new report from Jon Peddie Research indicates that GPU shipments from AIB partners, which include cards from companies like ASUS, MSI, and GIGABYTE, for Q2 2024 increased compared to the previous quarter and the same quarter last year. This is surprising because Q2 is generally quieter, and sales dip compared to Q1.

GPU shipment increased by 9.4% compared to the previous quarter, dramatically higher than the 10-year average of -7.1% for Q2. This figure represents around 9.5 million units, with NVIDIA and AMD benefiting from the increased interest in GPUs.

NVIDIA's quarter-to-quarter shipments increased by 9.7%, and a massive 61.9% compared to last year. AMD's quarter-to-quarter shipments increased by 9% while increasing by 3% compared to the previous year. As expected, NVIDIA continues to dominate the market with a whopping 88% market share.

The PC hardware market is experiencing what can only be considered a tumultuous time, which is resulting in price increases and a lack of consumer trust.

To put it in a nutshell, consumers are battling the woes and victories of the big three of PC hardware - NVIDIA, AMD, and Intel. Firstly, Intel suffered a massive brand hit as consumers lost trust in the company's products following the debacle of 13th and 14th-generation CPUs and their power-related problems. On the other side of the CPU fence, AMD's latest Ryzen 9000 series CPUs were only just released and have fallen on deaf ears due to their disappointing gaming performance uplift compared to the uplifts seen in previous generations, particularly the jump from Zen 2 to Zen 3.

As for NVIDIA, the company is an example of how too much winning can eventually hurt consumers, as NVIDIA's incredible graphics card performance has led the company to absolutely dominate the discrete GPU space with more than 80% market share as of 2022. This domination and AMD's official withdrawal from producing high-end GPUs have resulted in NVIDIA being able to charge what it likes for its products, especially at the higher end of its product stack (RTX 4070 Super and above).

NVIDIA's RTX 4090 graphics card is running low on stock, and prices are rising quite rapidly, at least over in a couple of European countries - but this may well soon happen in the US, as we heard in a previous prediction.

3D Center (via VideoCardz) spotted that in Germany and Austria the RTX 4090 is selling out and some big retailers just have the one product left which is being sold at a predictably inflated price.

All this is happening following very recent rumors that NVIDIA is about to pull the plug on RTX 4090 production (and the RTX 4090D for that matter, the variant for the Chinese market).

The Cybertruck Mini-PC is a beast-looking system but doesn't have much GPU grunt if you're an enthusiast-level gamer, which is where Dawid Does Tech Stuff steps in, modding the Xyber PC with a flagship NVIDIA GeForce RTX 4090 graphics card. Check it out:

One of the main issues is that the Cybertruck Mini-PC doesn't feature an easily accessible secondary M.2 slot, as it's positioned directly under where the wheel is, which stops devices being used in it (even if you drill into it). The Cybertruck Mini-PC isn't an easy thing to take apart, so modding it is a hefty job.

Dawid Does Tech Stuff used a PCIe adapter for the M.2 Wi-Fi slot, with teh M.2 PCIe adapter limited to two-lane operation at up to 32Gbps, slower than Thunderbolt 3 and 4 with up to 40Gbps, and much slower than OCulink with up to 64Gbps of bandwidth. The limitation in M.2 bandwidth leads to the GPU -- in this case, the monster RTX 4090 -- from running at full power levels.

According to rumors and information from several sources, NVIDIA plans to launch its next-gen GeForce RTX 50 Series with the new flagship GeForce RTX 5090 and GeForce RTX 5080 early next year - Q1 2025.

With that, companies are lining up new hardware, including Power Supplies or PSUs, designed for the next generation of PC gaming and desktop AI. For example, MSI's impressive new 1600W MEG Ai1600T PCIE5 and the 1250W MPG A1250 (also available in 1000W and 850W), which we got to see during a recent tour of MSI's impressive factory in Shenzhen, China.

Interestingly, these new PSUs feature two 12V-2x6 power connectors (the updated 16-pin 12VHPWR that can deliver up to 600W on a single cable). MSI's reason for the additional connector is to support the next generation of graphics cards. Could we see the GeForce RTX 5090 launch with two power connectors? Possibly.

Recently, NVIDIA quietly announced that it was changing the design of the popular GeForce RTX 4070 graphics cards, switching from GDDR6X memory to the more readily available GDDR6 memory. The change was made due to a GDDR6X shortage, with NVIDIA noting that the update or revision would deliver 'similar performance' to the existing GDDR6X models.

Without a price change, there has been some backlash to the decision - according to vocal community members, NVIDIA should have announced a price drop for what is perceived as a weaker card being manufactured with more affordable components.

With GeForce RTX 4070 cards with the revised GDDR6 memory making their way to retail shelves from various NIVIDA partners, there is one model that features a price point lower than its GDDR6X counterpart. The ZOTAC GAMING GeForce RTX 4070 Twin Edge, which is currently available for $10 less than the GDDR6X version at US retailer Newegg for $549.99 instead of $559.99.

In a recent interview with Tom's Hardware, AMD's Jack Huynh, the company's Computing and Graphics Business Group leader, confirmed that the next version of AMD's FidelityFX Super Resolution or FSR 4 will embrace and use AI. This applies to FSR 4's Frame Generation and Super Resolution components, and the reasoning isn't so much to follow NVIDIA's lead but to improve battery life on handhelds.

Citing devices like the ASUS ROG Ally and Lenovo Legion GO, Jack Huynh states that battery life is the number one priority for AMD's graphics teams and Radeon group. "I need to play a Black Myth: Wukong for three hours, not 60 minutes," Huynh explains. "This is where frame generation and interpolation [come in], so this is the FSR4 that we're adding."

AMD's FidelityFX Super Resolution, or FSR, is the company's response to NVIDIA DLSS. However, unlike Team Green's approach, its algorithms and technology are currently not AI-based. The result is lower image quality and stability than DLSS, with the trade-off being that FSR can run on any GPU.

AMD is expected to unveil its new Strix Halo APU at CES 2025 early next year, but the new Strix Halo APU as well as the GFX1151 integrated GPU are now supported by ROCm.

The new Strix Halo APU is expected to carry on what AMD did with the recent launch of its Ryzen AI 300 series "Strix Point" APUs, but they'll get a huge upgrade in the integrated RDNA 3.5-powered GPU. Strix Halo is expected to ship with a monster 40 Compute Units of RDNA 3.5 power in the form of the "GFX1151" GPU, up from the 16 CUs on the Strix Point APU inside of laptops launching now.

The leak was spotted by X user David Huang, posting some pics showing that Strix Halo and the GFX1151 GPU have got official support for ROCm. ROCm is the Radeon Open Compute platform, an open-source platform for GPU computing that is a handy tool for developers. AMD's upcoming "GFX1151" is the codename of the integrated graphics inside of Strix Halo, based on the newer RDNA 3.5 GPU architecture.

Our friends at Wccftech have posted a review of the GALAX GeForce RTX 4070 OC, one of the new RTX 4070 variants that sports 12GB of GDDR6 memory instead of the more advanced GDDR6X. As one of the most popular GPUs for PC gaming, NVIDIA has reportedly switched to GDDR6 for the RTX 4070 due to memory shortages and supply constraints.

When this news broke, and NVIDIA confirmed that GeForce RTX 4070 GPUs with GDDR6 memory would perform similarly to existing models, a mini-controversy broke out. With no price drop, model number change, or specific labeling, many began to believe that people couldn't tell if they were buying the 'faster' or 'slower' RTX 4070 graphics card.

Granted, this is only a single review, but with Wccftech testing the GALAX GeForce RTX 4070 OC and comparing performance between it and the GDDR6X variant, the conclusion is that the new GDDR6 models offer the "same performance."

AMD has released Technical Preview drivers for its driver-based AMD Fluid Motion Frames 2 or AFMF 2 frame generation technology. Without waiting for per-game implementation like FSR 3 or DLSS 3, AFMF 2 sits within AMD's Radeon drivers or Adrenaline Edition software and can be enabled in thousands of PC games.

It's an exciting technology, and even though the image quality isn't on par with something like DLSS 3, it can be a game-changer for a range of Radeon GPUs. Especially in improved AFMF 2 for,

Today, we can add the integrated RDNA 3.5 graphics in notebooks with AMD Ryzen AI 300 Series APUs to the AMD Fluid Motion Frames 2 compatible chips supported in AMD's Technical Preview drivers for AFMF 2. The new mobile chips also use the Variable Graphics Memory (VGM) tech as part of HYPR-RX to further boost performance.

We've covered AMD's Radeon RX 7800M GPU for mobile devices and gaming laptops in recent weeks due to its appearance in the upcoming ONEXGPU 2 external GPU from One-Netbook. The Radeon RX 7800M is a brand-new RDNA 3 Navi 32 GPU from Team Red for notebooks and portable devices below the flagship Radeon RX 7900M and the mainstream Radeon RX 7600M.

Early reports show the performance is impressive and roughly on par with the Radeon RX 7700 XT desktop GPU. As far as we can tell, it will compete with NVIDIA's popular GeForce RTX 4070 Laptop GPU.

Today, we didn't get an official Radeon RX 7800M announcement from AMD, as much as confirmation that it exists. Instead of the usual press release with information about the GPU and its performance relative to previous generations and benchmarks comparing it to the competition, AMD has simply made the GPU spec page for the Radeon RX 7800M live.