GPU

ASUS has shared the first images of a new graphics card produced in partnership with Noctua. The RTX 3070 Noctua should not only run quietly but also coolly thanks to a huge heatsink. The card's design may take gamers some getting used to, mind you.

Honor has made good on some earlier teasers with a launch for the new MagicBook 16 series, with an up-to-date Ryzen 7 platform and 144Hz displays. They have been joined with a less expected variant: the V 14. It will now join the first wave of laptops to ship with Windows 11, and will do so with a 2K display and Tiger Lake-H35 processors.

Thanks to a very productive community that is working diligently on an open source driver for the alternative operating system Linux, older AMD graphics cards are about to receive raytracing support in the near future.

According to a leak from a reliable source, AMD plans to release its next RDNA 2-based graphics card in a few weeks. The Radeon RX 6600 will be the cheapest offering in AMD's current product stack, but with only slightly worse specifications than the Radeon RX 6600 XT.

The NVIDIA GeForce RTX 3080 SUPER has appeared in a listing for HP's latest AiO PC. NVIDIA is expected to reveal the RTX 30 SUPER series in early 2022, but HP's references to an RTX 3080 SUPER are likely just an error.

Veteran leaker @kopite7kimi has shared information about Nvidia's upcoming graphics cards, including the GeForce RTX 3090 Super, RTX 3080 Super, RTX 3070 Super and RTX 3060 Super. The SKUs are slated to offer incremental upgrades over their base variants, such as faster VRAM and more CUDA cores.

AotS might not be a popular benchmarking tool due to its convoluted results, yet its reliance on CPU power seems to be great for highlighting Intel's Alder Lake performance. It is still unclear if the i9-12900K benefits from some overclocking treatment in this case, but the Alder Lake processor does seem to be 39% faster than the AMD Ryzen 9 5950X when coupled with an RTX 3080 running at High_1440p settings.

The prices of current AMD and NVIDIA desktop graphics cards are continuing to move away from MSRP, having bottomed out in July. Both the RX 6000 and RTX 30 series are now selling for around 1.7x MSRP, with second-hand prices remaining at inflated levels, too.

In our review of the new Acer TravelMate P6, we observed that this business laptop's Iris Xe Graphics G7 iGPU in the Core i7-1165G7 is among the best implementations compared to the competition. The Acer TravelMate P6 also brings in several other improvements to the overall experience, which have been exhaustively detailed in our full review.

The upcoming AMD RDNA 3 Navi 31-based RX 7900 XT will be based on a 5 nm + 6 nm MCM and could use a 256-bit memory bus but stick to GDDR5 VRAM. Navi 33 (RX 7700 XT) will be a 6 nm monolithic die with 128-bit GDDR6 VRAM and offer more performance than the current RX 6900 XT. The Nvidia Lovelace AD102 RTX 4090 will continue to use 384-bit GDDR6X memory and have GPU clocks greater than 2.2 GHz.

NVIDIA has repurposed its GA104 GPUs for the GeForce RTX 3060, a card that it launched on the GA106. These graphics cards will have nearly half their GPUs disabled to meet the RTX 3060's specifications.

NVIDIA still cannot produce enough RTX 30 series cards to satisfy demand, a year on from the announcement of the RTX 3070, RTX 3080 and RTX 3090. With availability expected to be non-existent throughout 2022, NVIDIA will fill the gap with a revived RTX 2060 that has 12 GB of VRAM.

A leaked Intel slide indicates that the company's upcoming DG2 Arc Alchemist GPU could be targeting the Nvidia GeForce RTX 3070 and AMD Radeon RX 6700 XT at the high end at the US$500 mark while looking to compete with the GTX 1650 Super at the low end in the US$100-199 bracket. The slide does not list any specific SKUs but only mentions the possible TDPs.

Expectations are high for the Lovelace GeForce RTX 40 series of graphics cards from Nvidia, and it seems price predictions are similarly high. There have been a few suggestions that the top-end GeForce RTX 4090 card could end up being stuck with an MSRP of US$2,999, leaving it likely out of reach for the majority of gamers.

NVIDIA may have released the RTX 3080 Ti as a 12 GB card, but it and AIBs were developing 20 GB cards. A leaked BIOS has confirmed a few aspects about the unreleased variant, including its ludicrous 450 W power limit.

Rumour has it that NVIDIA is planning to release two new GeForce RTX generations in 2022, leaving the GeForce RTX 30 series as the only option for a while yet. Reputedly, NVIDIA's RTX 30 SUPER 'Ampere Refresh' will debut first, and will have around nine months on the shelves before RTX 40 'Lovelace' cards arrive. RDNA 3 and RTX 40 cards could arrive at nearly the same time, too.

U.S. and South Korean RISC-V researchers managed to make Nvidia's CUDA code compatible with non-Nvidia hardware. It is quite the achievement considering AMD and Intel have not yet pursued such a solution. The RISC-V-based Vortex GPGPU essentially translates CUDA code into OpenCL instructions through the RISC-V library.

Rumours have started circulating that the Apple M2 will deliver exceptional graphics performance, potentially as high as a former flagship NVIDIA GPU. The reality could not be further from the truth.

The Radeon RX 6600 has appeared online again, over a month after the RX 6600 XT arrived. The leaked marketing assets confirm not only 8 GB of GDDR6 VRAM for the Navi 23 XL card, but also its PCB size relative to the RX 6600 XT.

Even though the INF files do not mention most of the GTX 700 Kepler cards anymore, this version still supports the old GK100 functions and runtimes are still compatible with SM 3.x. Additionally, the preview version integrates new next gen runtime compilers that may end up used by the upcoming RTX 4000 cards.

A few weeks ago, some EVGA GeForce RTX 3090 started dying after running Amazon's new game New World. It turns out that Amazon is not to blame, nor is a faulty fan controller.

Probably a little too late to solve the gaming GPU crisis, but Nvidia's top-of-the-line CMP 170HX crypto mining cards are now available in China. These models come directly from Nvidia and feature a cut-down GA100 die with 4480 CUDA cores complemented by 8 GB of HBM2e VRAM. Interestingly enough, the cooling system has no fans and the card itself cannot be used for gaming or compute workloads. It is also a bit slower than the advertised 200 MH/s speeds.

A Bahrain-based retailer just showed off a massive shipment of AMD RDNA2 graphics cards, including the Radeon RX 6900 XT, RX 6800 XT and RX 6700 XT. The firm exclusively deals with cryptocurrency miners, so most of those cards will likely end up in mining rigs.

While the labels are clearly stating 20 GB, there is nothing else attesting to the actual VRAM capacity of these cards. These are probably some mislabeled models that still pack 12 GB, but just imagine the ridiculous prices asked by retailers if 20 GB models were indeed real.

Nvidia may introduce an RTX 3090 Super later this year to probably counter Intel's upcoming Xe-HPG Alchemist GPU launch in early 2022. The RTX 3090 Super is expected to be based on the GA102 core and will feature 10,752 CUDA cores and a higher 400 W+ TDP. Other specifications are not yet available.

SS appears to be the best of both worlds. On the one hand, it is AI-based and should provide high quality results similar to DLSS, but on the other hand, the tech aims to become completely open source by the second or third iteration.

The Galaxy S22 series will debut Samsung's highly awaited Exynos 2200 and its AMD mRDNA GPU. Early GPU benchmarks of the SoC running on beta drivers have now surfaced, and show a huge advantage over Apple A14 Bionic.

The NVIDIA GeForce RTX 3090 struggles to keep its VRAM temperatures in check anyway, with the GA102-based card shown to exceed temperature thresholds in stress tests. However, NVIDIA left a finger glove inside a Founders Edition card during assembly, much to the surprise of its owner.

Asus has revived the nearly seven-year-old Nvidia GeForce GT 730. Its key specs include 2GB of GDDR5 VRAM, 384 CUDA cores and four HDMI 1.4 ports. Information about its price and availability is unknown for now.

Intel detailed some of the technical aspects of its upcoming Arc Xe-HPG Alchemist DG2 GPU during its Architecture Day 2021 presentation. Arc DG2 is fabbed on the TSMC 6 nm process and is comprised of what Intel calls as Xe-cores, with up to 4 Xe-cores in a render slice which also includes ray tracing cores and fixed function units.

AMD may restrict the RDNA 3 architecture only to top-tier Navi 3x offerings such as Navi 31, Navi 32, and Navi 33, according to a new leak. SKUs from Navi 34 onwards may use a 6 nm refresh of the Navi 24, Navi 23, and Navi 22. Only Navi 31 and Navi 32 are currently expected to feature an MCM design.

Intel's upcoming Arc Xe-HPG DG2 Alchemist desktop GPU could sport a dual-slot, dual-cooler design with dual 9-blade fans going by hints in Intel's teaser video. Arc Alchemist will support all of DirectX12 Ultimate features, and Intel will likely talk about its own AI-based supersampling technology later this week.

NVIDIA's Lite Hash Rate algorithm has been broken again, but not by the company itself this time. Instead, a new version of NBMiner has improved the mining performance of GeForce RTX 30 LHR cards by up to 40%, bringing them closer to their theoretical peak hash rates. NBMiner 39.0 is compatible with NVIDIA's latest graphics drivers on Linux and Windows.

The recently previewed Radeon RX 6900 XT video card made by XFX in collaboration with masters of water cooling solutions EKWB will be added to the existing lineup of Speedster video cards by the Ontario-based company. There is no word regarding its availability and pricing, but it is safe to assume that the upcoming card will be rare and expensive.

According to new reports, NVIDIA's Hopper architecture could tape out at the start of 2022, or late 2021 at an early estimate. Meanwhile, mass production of GeForce RTX 40 series GPUs is expected to start in the middle of next year, so an autumn or winter 2022 release could be on the cards for NVIDIA's next-generation graphics cards.

AMD may have just launched another RDNA 2-based graphics card, but it has already confirmed that it is working on next-generation models. The Navi 3X series is not due until deep into 2022, mind you.

Intel's upcoming Xe-HPG family will now be officially called Intel Arc. Intel Arc is a multi-generational product with the first generation DG2 being codenamed Alchemist and slated for a Q1 2022 launch in both desktop and mobile form factors. Successive Arc generations will include Battlemage, Celestial, and Druid. Intel said that Alchemist will offer native support for all DX12 Ultimate features.

Ethereum miners have been drawn to NVIDIA's GeForce RTX 30 series cards, but AMD now has a compelling option in the shape of the Radeon RX 6600 XT. The Navi 23 card is twice as efficient as a GeForce RTX 3060, thanks to a 32 MH/s hash rate at 55 W.

NVIDIA has unveiled a new workstation graphics card that is not only small, but also economical. As such, even compact computers can benefit from NVIDIA's most advanced technologies, including AI acceleration through dedicated Tensor cores and ray tracing.

Sources familiar with the matter have told a Chinese news outlet that Intel is one of the first customers for TSMC's 3nm node. The Taiwanese chipmaker has earmarked an entirely new factory for the task, which will churn out Intel server processors and graphics cards. Mass production is expected to start sometime in H2, 2022.

The downward trend of NVIDIA GeForce RTX 30 series prices continues, having seemingly peaked sometime in May. By contrast, the cost of AMD Radeon RX 6000 series cards is rising, excluding the RX 6600 XT. Unfortunately, AMD and NVIDIA's desktop graphics cards remain well above MSRP.

A third-party upscaling tool created by a Chinese developer allows any Windows game running at a lower resolution in windowed mode to upscale to native monitor resolution using an early implementation of AMD's FSR technology. Users have been reporting decent levels of upscaling and increased fps though at the cost of increased frametimes, latency, and CPU usage.

Radeon Pro W6000 series MPX modules are now available to order for existing Mac Pro machines. Unsurprisingly, the proprietary GPUs are expensive and are incompatible with some existing Mac Pro configurations.

Despite featuring the Alder Lake CPUs, the NUC 12 Enthusiast models do not support DDR5 RAM and are limited to DDR4-3200. Nevertheless, the upcoming NUCs come with powerful specs like Intel's own DG2 dGPUs with 8 / 12 / 16 GB GDDR6 VRAM, two Thunderbolt 4 connectors, Wi-Fi 6E module, SDXC and UHS-II SD card reader, plus a redesigned chassis.

The latest iteration of the Xenia 15 gaming laptop is powered by the i7-11800H processor coupled with 32 GB of DDR4-3200 RAM, a new 1 TB XPG Gen 4 SSD and either an RTX 3070 or 3060 dGPU from Nvidia. It comes with advanced features like Thunderbolt 4 connectors and a 15.6-inch 165 Hz QHD screen. On the other hand, the new XPG DDR5 modules that can be overclocked to 12600 MT/s now feature as much as 64 GB per single module.

The latest EEC filings from Asus show a lonely RTX 3070 card with the Noctua suffix. Hopefully this experiment proves a success so we can see the Noctua solutions installed on high-end, as well as lower-end solutions in the future.

Redditor @zyck_titan discovered that, despite scalping and astronomical prices, NVIDIA’s Ampere lineup, including cards like the GeForce RTX 3090 have sold ten times the volume of AMD’s RDNA2 cards.

Apple and AMD have announced new Radeon PRO graphics cards for the Mac Pro. Only the Radeon Pro W6800X and Radeon PRO W6900X are available, but the Mac Pro can now be configured with up to four GPUs for a total of 128 GB of GDDR6 VRAM and 120 Compute Units per graphics card.