GPU

Games could see up to 28% fps improvements with faster RAM kits, but the average is around 10% at 1080p max settings or 15% for 1080p low settings. The absence of a MUX switch raises the performance gap to 33% in Shadow of the Tomb Raider, so we might want to wait for the more expensive versions that allow us to disable the iGPU.

Nostalgic about GPUs from the past few decades? This museum has everything from the console and arcade boards of the '80s, ISA cards, 3DFX SLI setups, the first true GPU - the GeForce 256 from Nvidia, pre-AMD ATI dual-GPU beasts and even separate sections dedicated to the Chinese e-sport scene sponsored by Colorful iGame or VR and high-end racing simulators.

Dell may be up to its CUDA core cannibalizing shenanigans again. A cursory glance at the Dell G15 listing tells us that its CUDA core count is set at 3,584 instead of 3,840. One could have dismissed it as an error had it not been for Dell's recent Alienware m15 R5 debacle. Update: The GeForce 3060 on the Dell G15 has the correct number of CUDA cores

The revamped W-series now includes the WS76/66 and WE76 high-end models with up to an i9-11980HK CPU, up to 64 GB of DDR4-3200 RAM and a choice between the RTX A5000 or A3000 professional GPUs plus screens up to 4K. These models also include PCIe 4.0 support and a Thunderbolt 4 connector, while the WF76/66 models do not, and also support only up to an i7-11800H CPU with RTX A2000 or T1200 and 1080p screens.

The AMD Radeon Pro W6800 has surfaced on the Geekbench 5 OpenCL database. The Radon Pro W6800 is based on the same Navi 21 GPU as the consumer Radeon RX 6800 but offers 32 GB RAM and, apparently, much higher clocks. The OpenCL results of the Radeon Pro W6800 show significant improvements over the W5700 series and seem to be on par with an NVIDIA Quadro GV100.

Lenovo will be offering new Legion 5 models with Ryzen 5 5600H and Radeon RX 6600M in Germany at prices starting from €1,299. However, this pricing is significantly higher than a Ryzen 7 5800H and RTX 3060 Laptop GPU configuration at €1,249 indicating that Legion 5 models with NVIDIA GPUs could offer a better price-to-performance ratio.

A cryptocurrency miner has confirmed that Nvidia's hash rate limiter works as intended on the GeForce RTX 3080 Ti. Hash rates for Altcoins such as Ethereum Classic, Ergo, and RavenCoin appear to be significantly lower, too.

The slower RAM gimps the gaming performance by up to 20% on the Asus model, while the missing MUX switch is further limiting the potential that should match models equipped with RTX 3080 mobile dGPUs. Similar RAM problems can also be encountered on the Lenovo Legion 5 Pro. However, the Lenovo Legion 7 AMD features faster RAM, so this is not really just another conspiracy against AMD.

The NVIDIA GeForce RTX 3070 Ti has made its debut on benchmarking websites. The RTX 3070 Ti competes with the RTX 2080 Ti and Radeon RX 6800 XT, which are more expensive than NVIDIA's latest effort. The RTX 3070 Ti appears to be worse value for money than the RTX 3070, though.

Several Alienware m15 R5 laptops are apparently reporting just 4,608 CUDA cores for the RTX 3070 Laptop GPU instead of 5,120. Apart from the CUDA cores, even the clocks, TMUs, and pixel fillrates are being reported to be lower than normal leading to performance loss. Currently, a possible workaround is to flash the VBIOS from an m15 R4 that seems to restore the correct values and performance.

Looks like Nvidia is playing the prudent game and is not touching any smartphone chips before completing the ARM acquisition. Its current solution for Android and iOS devices is the GeForce Now cloud gaming service, allowing AMD and Samsung to be first to market with the upcoming Exynos SoC that integrates an RDNA2 GPU later this year.

This move might not surprise too many, but it does show that Nvidia can still be caught off guard by AMD, even when Team Red announced the arrival of the mobile RDNA 2 dGPUs half a year ago. On top of this, AMD might not even come up with a decent supply of RX 6000M dGPUs due to the ongoing chip shortages.

Leaked ThinkPad X1 Extreme Gen 4 specifications indicate quite a massive internal upgrade for Lenovo's multimedia flagship laptop. The specs indicate Intel Tiger Lake-H CPU and, for the first time, NVIDIA Ampere RTX 30 Super Laptop GPU options. The ThinkPad X1 Extreme Gen 4 is also slated to offer a slimmer chassis, 16:10 2.5K and 4K display options, 90 Whr battery, FHD webcam, backlit keyboard with 1.5 mm key travel, and more.

Intel's graphics chief Raja Koduri showed off the Xe-HPG DG2 512 EU GPU on Twitter indicating that now it's software engineering's turn to optimize performance in benchmarks and games. Raja also clarified to a query that Intel is indeed looking into considering AMD's FidelityFX Super Resolution (FSR) technique in the interest of ISVs.

NVIDIA is bringing its DLSS image upscaling technology to Linux users via Proton. This will enable supported games on Linux to leverage the Tensor cores on RTX GPUs for AI-based supersampling. NVIDIA said that DLSS will be enabled for Vulkan games this month followed by DirectX titles in the fall.

This new NUC goes into beast mode with an 8-liter enclosure that can fit full-size GPUs plus a PSU. It also supports 45 W Intel Tiger Lake-H CPUs including the HK overclockable models, but apparently only supports air coolers, so it might get some TGP restrictions. Nevertheless, it also comes with a NUC 11 Extreme Compute Unit that includes features such as PCIe 4.0, Optane Memory, 2x Thunderbolt 4.0 connectors and even a 10 Gb NIC.

What has been only a rumor so far is finally official: AMD CEO Lisa Su has confirmed during her Computex 2021 keynote speech that the next-gen Exynos SoC will feature a custom AMD RDNA 2 architecture with ray tracing and variable rate shading. More details about this upcoming 5G chip will surface later this year.

NVIDIA kicked off its Computex 2021 keynote with a slew of gaming announcements and has officially launched the RTX 3070 Ti and RTX 3080 Ti GPUs for US$599 and US$1,199 MSRP, respectively. The company also showed off a Doom Eternal demo running on the RTX 3080 Ti at 4K with RTX on and confirmed that RTX features are coming to several current and upcoming games in the coming weeks.

AMD offered some details on its much anticipated FidelityFX Super Resolution (FSR) upscaling technique, which is pegged as a competitor to NVIDIA's popular DLSS technology. AMD FSR will be a part of GPUOpen and will offer four quality presets that work with almost all modern GPUs on the market. AMD demoed the technology with Godfall running on the Radeon RX 6800 XT and also on the NVIDIA GeForce GTX 1060.

AMD has unveiled the RDNA 2-based RX 6000M mobile GPUs for gaming laptops to take on the competition from NVIDIA Ampere Laptop GPUs. The RX 6000M lineup includes the RX 6800M, RX 6700M, and the RX 6600M designed to target 100 fps+ gaming at 1080p and 1440p. AMD is also introducing the AMD Advantage Design Framework, which is a collaborative effort with OEMs to introduce performance standardization in AMD gaming laptops.

A Czech webstore now offers what is claimed as the recently-launched Asus ROG Zephyrus M16; then again, the new listing contains specs that are considerable downgrades on what the OEM announced while debuting this machine. For example, this apparent M16 leak involves the Core i7-11800H processor, as opposed to the i9-11900H on official product pages.

NVIDIA is expected to announce the GeForce RTX 3080 Ti on May 31, its latest GA102-based graphics card. A leak has not only confirmed that NVIDIA plans to sell a GeForce RTX 3080 Ti Founders Edition, but it has also provided a hands-on look at its design.

Chinese testers note that the new GA106-302 chips featuring the LHR limiter included with the new RTX 3060 cards provides the same gaming performance as the original versions, but effectively halves the Ethereum mining performance in half. The same is apparently true for the upcoming RTX 3080 Ti cards.

We now know the final specs for the reference RTX 3080 Ti cards, including the base and boost clocks, VRAM specs and TGP. Performance should be very close to an RTX 3090, but the first Ashes of the Singularity tests do not seem to support this.

NVIDIA is seemingly only a few days away from announcing the GeForce RTX 3070 Ti, but marketing renders of two Gigabyte cards have already leaked. The renders confirm that the GA104-based card has GDDR6X VRAM, an improvement on the RTX 3070. The RTX 3070 Ti may nearly match the RTX 3080's power consumption, too.

The new A7 X1 is the first AMD-based mid-range gaming laptop from Gigabyte's new Gaming series. It features the Ryzen 9 5900HX processor with integrated Vega 7 graphics plus an Nvidia RTX 3070 Max-Q 140 W discrete solution. The display is limited to FHD resolution, but gets 144 Hz refresh rate, and there is support for up to 64 GB of DDR4-3200 RAM and 6 TB of storage.

The RTX 3090 Ti could feature a full GA 102 die just like the A6000 professional cards, plus maybe some more VRAM and slightly higher clocks with raised TGP. This configuration would net a performance increase of around 15-25% over the RTX 3090 FE models. It is entirely possible this could also turn out to be an entirely new SKU produced on the 7 nm nodes from TSMC.

The Intel-based Neo M21 models might slightly lag behind those equipped with the Ryzen 7 5800H as far as CPU performance is concerned, but the addition of PCIe 4.0 and Thunderbolt 4 could be decisive for many users. Otherwise, the Neo 15 / 17 models powered by the i7-11800H seem to be mostly identical to the E21 AMD systems released earlier this year.

Information about AMD's upcoming Navi 23-based Radeon RX 6600 and RX 6600 XT has now surfaced. The RX 6600 series will be available in both mobile and desktop offerings and is expected to feature up to a 2,350 MHz boost clock, 8 GB GDDR6 VRAM, and PCIe Gen4 x8 connectivity. The PCIe Gen4 x8 limitation may be a deterrent for those looking to upgrade to Navi 23 on the desktop.

NVIDIA has committed to releasing 'Lite Hash Rate' versions of the GeForce RTX 3060 Ti, RTX 3070 and RTX 3080. The RTX 3090 remains unaffected, but all other desktop RTX 30 series cards shipping from this month onwards will be capped at half their theoretical ETH hash rates.

The Chinese mini PC maker is breaking away from its usual square-shaped builds with modest specs and is presenting its first all-AMD gaming-oriented system featuring an open air case packing a Ryzen 5 5600X processor coupled with 32 GB of DDR4-3200 RAM and 1 TB M.2 SSD storage, plus a Radeon RX 6700 XT GPU.

More GeForce RTX 30 series cards have been certified, this time the RTX 3050 Ti and RTX 3070 Ti. The RTX 3080 Ti has been certified once again too, as have numerous CMP HX mining cards.

Online retailers have been trying to enforce special policies to prevent crypto miners and scalpers from wiping out GPU stocks in no time. GPUs are now usually bundled along CPUs, motherboards or SSDs, but crypto miners are obviously not interested in any of these, so they are expected to dump all non-GPU components at low prices on eBay.

GALAX has confirmed NVIDIA's plans to re-release the GeForce RTX 30 series with revised ETH mining limits. All cards bar the GeForce RTX 3090 have received new GPUs with a '2' suffix, replacing the '0' suffix from the original Ampere GPUs.

MSI has, perhaps unwittingly, listed the RTX 3080 Ti on its website. The blunder follows a crate of MSI GeForce RTX 3080 Ti VENTUS 3X 12G OC cards being spotted last month.

The latest form of NVIDIA's Studio standard for creator-grade laptops combine in-demand specs such as RTX 3000-series graphics, 1440p screens and device heights of as little as 16 millimeters in some cases. The company has also now officially announced the RTX 3050 and 3050 Ti platform for laptops, which are to be found in gaming machines as well as those for professionals soon.

AMD is expected to release the Radeon RX 6600 series before the end of the quarter, with the RX 6600 and RX 6600 XT serving as budget entries in its RDNA 2 line-up. Specifications for both cards have now leaked, including an EEC registration made by ASRock.

Intel's upcoming data center accelerators based on the DG2 gaming GPUs will feature 1-tile, dual-tile and quad-tile versions. The single tile model will only have 384 EUs active and will pack 16 GB of HBM2e VRAM, while the dual-tile variant comes with 480 EUs per tile and ups the available video memory to 32 GB.

The RTX 2080 Ti has been modded to support 22 GB of VRAM. NVIDIA's former flagship card comes with 11 GB of VRAM as standard, but the company left in the ability for it to use double its default capacity.

Highlights for the 2021 Dell XPS 17 9710 models include the new Tiger Lake-H mobile CPUs up to an i9-11980HK, up to 64 GB of DDR4-3200 RAM, up to 4 TB PCIe 4 x4 NVMe SSD storage, Nvidia RTX 3050 / 3060 mobile GPUs, 4x Thunderbolt connectors, and 4K HDR touchscreens with 16:10 aspect ratio.

The upcoming Alienware X17 compact gaming laptops will feature a couple of industry's first solutions, including the Element 31 thermal interface material made of a Gallium Silicone matrix for improved heat transfer and a quad fan cooling system. The press release also mentions 11th gen Intel CPUs and RTX 3000 GPUs, while the teaser images reveal a Thunderbolt 4 connector on the back, plus a 16:9 display.

Additional specifications of Intel's upcoming Xe-HPG DG2 GPU have leaked alongside a board layout diagram. The specs indicate a 100 W TDP for the 512, 384, and 256-EU variants. Xe-HPG DG2 could also feature support for DisplayPort 2.0. Intel is expected to launch Xe-HPG DG2 sometime between October and December this year starting with the low-end SKUs.

Opinion The current chip shortage is preventing eGPUs from becoming mainstream. Once it's over, however, we suspect that many gamers will opt for an affordable eGPU over paying 2x to 3x the price for a full-fledged gaming laptop of similar graphics performance.

Dell has issued numerous patches as part of its latest Security Advisory (DSA-2021-088) in response to findings of a "severe" CVE consisting of five different privilege-exploit bugs that have apparently been in effect for the last 12 years. Fortunately, they have never been acted upon, which, as they could have allowed unauthorized users to write data or manipulate memory.

Initially believed to launch towards the end of 2021, AMD's alternative to the Nvidia DLSS technology could arrive in June, but only for PC gamers. According to Youtuber Coreteks, AMD's FSR will be based on a software-level algorithmic super sampling approach that is easy to implement by game developers and also works on non-AMD hardware like Nvidia GTX and possibly even Intel GPUs.

With Resident Evil Village set to release soon, outlets with review copies have published benchmarks that indicate that the game will scale very well on lower-end hardware. PC Gamer even managed to get Resident Evil Village running on the Ryzen 5 3400G with integrated graphics, albeit at 720p/60.

AMD's bet on higher core counts is paying off when it comes to pure multi-thread performance. Nonetheless, Intel still has the upper hand when it comes to AI integration, single-thread performance, and native Thunderbolt support.

NVIDIA recently unveiled its 2021-2025 roadmap for the datacenter segment. The roadmap indicates that an Ampere successor, likely Lovelace, could arrive next year, with a new GPU architecture beyond that set to arrive before 2024.