Video Cards & GPUs News - Page 142

NVIDIA has just launched the new GeForce RTX 3080 12GB graphics card, with an updated GA102-220 GPU and 12GB of ultra-fast GDDR6X memory.

The new GeForce RTX 3080 12GB offers 2GB more GDDR6X memory over the RTX 3080, which has 10GB of GDDR6X -- but also arrives on a higher 384-bit memory bus -- and offers up to 912GB/sec of memory bandwidth. NVIDIA's original GeForce RTX 3080 had 10GB of GDDR6X, but on a smaller 320-bit memory bus, resulting in 760GB/sec of memory bandwidth.

This means the new GeForce RTX 3080 12GB has the same 12GB of GDDR6X @ 19Gbps with up to 912GB/sec memory bandwidth of the higher-end GeForce RTX 3080 Ti, but with less CUDA cores. Now talking about CUDA cores, let's start with the RTX 3080 which has 8704 CUDA cores, the new RTX 3080 12GB has 8960 CUDA cores, while the higher-end RTX 3080 Ti will reign supreme with 10240 CUDA cores.

NVIDIA has announced out of nowhere its new Deep Learning Dynamic Super Resolution technology, or DLDSR, which is an AI-powered version of Dynamic Super Resolution (DSR) downscaling technology -- something that has been available to GeForce graphics card owners inside of the GeForce drivers.

The new NVIDIA DLDSR technology improves over DSR by using an AI-powered layer, which is powered by NVIDIA Tensor cores inside of GeForce RTX series GPUs. NVIDIA has a new January 14th Game Ready Driver that will update NVIDIA's built-in DSR feature with AI.

DLDSR improves on DSR by adding an AI network that NVIDIA says "requires fewer input pixels, making the image quality of DLDSR 2.25X comparable to that of DSR 4X, but with higher performance. DLDSR works in most games on GeForce RTX GPUs, thanks to their Tensor Cores".

The organization that handles the PCIe standards, PCI-SIG, have announced the release of the next-gen PCIe 6.0 standard which has some truly wicked next-level bandwidth.

PCIe 6.0 will offer up to 256GB/sec through a x16 PCIe 6.0 slot, and with advanced PAM4 technology (Pulse Amplitude Modulation with 4 levels) and many other new features and technologies, the new PCIe 6.0 standard is a huge step up from the already impressive PCIe 5.0 standard (which is huge over PCIe 4.0, and then PCIe 3.0, and so on).

The new super-fast PCIe 6.0 standard will usher in levels we've never seen before, with a x4 PCIe 6.0 M.2 NVMe SSD capable of 64GB/sec -- while if you had an AIC RAID card you could drive that up to an insane 256GB/sec -- untouchable levels with even PCIe 5.0, let alone PCIe 4.0.

NVIDIA's next-gen GeForce RTX 4090 will come and blow our socks off later this year, powered by the new Ada Lovelace GPU architecture, but now there are some massive predictions of the RTX 4090 performance... and wow.

In a new tweet, 3DCenter.org answers itself in the question that the earliest that NVIDIA could test out 12 games on its next-gen GeForce RTX 4090 graphics card would be "sometime in the summer of 2022". The tweet continues, noting that NVIDIA has to do a few things first: AD102 tape out, AD102 validation, real boards in the wild, working clock rates (Ada Lovelace GPU clocks will be much higher than Ampere GPU clocks), and then great drivers.

A video posted in December 2021 from all of the gathered leaks and rumors of the GeForce RTX 4090 has it pegged for some serious 4K gaming powerhouse performance, but we've already heard the rumors of the new RTX 4090 being 100% faster than the RTX 3090... but in a chart, man-oh-man, does it look good. 4K gaming at 400FPS in games like Rainbow Six: Siege... nearly 100FPS at Ultra settings in Cyberpunk 2077... yeah.

NVIDIA has announced that there are 10 games in the pipeline that will soon be receiving either DLSS or Reflex technologies when they launch this year.

The Day Before is a new open-world survival MMO from Fntastic, launching later this year and will pack NVIDIA DLSS support as well as 3 ray-traced effects: ambient occlusion, reflections, and global illumination. NVIDIA's injection of RTXGI will add performant infinite-bounce global illumination light from not only the sun, but from the moon and other natural light sources in the game.

Rainbow Six: Extraction is up next, from Ubisoft and will be launching with both NVIDIA DLSS and NVIDIA Reflex technologies -- but Rainbow Six: Siege runs on the same engine as Extraction -- so it should come as no surprise that the new game supports DLSS and Reflex tech.

COLORFUL has just launched its new GeForce RTX 3050 series graphics cards, with new members of the RTX 3050 series including the RTX 3050 Ultra W, RTX 3050 NB EX, RTX 3050 Ultra W DUO.

The flagship COLORFUL GeForce RTX 3050 Ultra W rocks a beautiful design, using white, purple, blue, and a touch of pink that culminates in a great effect across the card. The card is cooled by a triple-fan cooler that makes for a larger, but still mid-range GPU yet it looks great while doing so. There's also the dual-fan version in the RTX 3050 Ultra W DUO, if you don't quite want that huge card.

The triple-fan cooler on the COLORFUL GeForce RTX 3050 Ultra W uses 2 x 90mm and 1 x 80mm cooling fans to keep things cool, with a high-performance heat sink has two 6mm-diameter and two 8mm-diameter heat pipes connected to the fins using integrated reflow soldering.

Alienware unveiled its interesting Concept Nyx at CES 2022 this week, but it also showed off its Concept Polaris which turns portable gaming laptops into enthusiast-class beasts.

The new Alienware Concept Polaris is a fancy-looking external GPU solution that can handle huge graphics cards -- up to 16 inches, compared to the original Alienware Graphics Amplifier which took 10.5-inch graphics cards -- as well as rocking built-in liquid cooling.

Alienware Legend 2.0 ID theme is featured throughout, Alien FX loop lighting, and the usual cable management being on point is here with the Alienware Concept Polaris. There's a removable internal PSU, which reroutes the power through to dual 330W adapters, while a higher-end model with dual 425W adapters is also in the pipeline.

NVIDIA has made its GeForce RTX 3090 Ti graphics card official, packing the full GA102 GPU and its huge 10752 CUDA cores as well as even more suped-up GDDR6X memory.

The newly beefed-up GeForce RTX 3090 Ti has the same 24GB of GDDR6X as the regular RTX 3090 non-Ti, but we're now going to see faster 21Gbps bandwidth (up from 19.5Gbps on the RTX 3090). But, the RTX 3090 could have its 24GB GDDR6X memory overclocked to 21Gbps (and beyond) so it'll be interesting to see if the 21Gbps GDDR6X can be pushed north of 22-23Gbps.

NVIDIA is rolling out its new GeForce RTX 3090 Ti with a refreshed and new 16-pin PCIe power connector, something that will help drive the card and its 450W TDP. The GPU boost clock of the RTX 3090 Ti and its GA102 GPU is 1860MHz, but I'm sure that we'll see custom RTX 3090 Ti graphics cards push past that -- while the NVIDIA GeForce RTX 3090 Ti Founders Edition will end up one of the best again -- I'm sure of it.

AMD is about to introduce its new Radeon RX 6000 XT series mobile GPUs, with the introduction of the flagship Radeon RX 6850M XT graphics card that rocks 12GB of GDDR6 memory.

The new Radeon RX 6850M XT graphics card uses Samsung's faster 12GB GDDR6 @ 18Gbps, faster than the 16GB GDDR6 @ 16Gbps on the desktop Radeon RX 6800 XT, and Radeon RX 6900 XT graphics cards. AMD still taps TSMC's regular 7nm process node for its new Radeon RX 6850M XT (and the RX 6650M XT + RX 6650M) while the RX 6850M has 40 Compute Units and a Game Clock of 2463MHz.

AMD has 12GB of GDDR6 memory at the faster 18Gbps bandwidth, on a 192-bit memory interface and 96MB Infinity Cache -- all with a TGP of 165W. There's also the introduction of 2 x Navi 24 GPUs using the TSMC 6nm process node, and will power the Radeon RX 6500M and Radeon RX 6300M mobile GPUs (with 4GB and 2GB of GDDR6 memory, the RX 6300M has 2GB of GDDR6 at a higher 18Gbps bandwidth).

NVIDIA will be officially confirming and detailing its GeForce RTX 3080 Ti Laptop GPU at CES 2022 tomorrow, but before that, we've got a new leak that pretty much acts as a spoiler to NVIDIA's presser.

The new GeForce RTX 3080 Ti Laptop GPU will reportedly tap the new GA103S GPU, with 7424 CUDA cores that makes it just 256 CUDA cores under the full rumored GA103 GPU. What we do have is Samsung 16GB GDDR6 memory here, clocked at 16Gbps and on a 256-bit memory interface that provides up to 512GB/sec of memory bandwidth.

VideoCardz has been told that the screenshot you see above with the GPU boost clock at 1395MHz isn't the best we'll see -- in their words -- this GPU in question "is not the best configuration that will be available". NVIDIA's mobile-focused GeForce RTX 3080 Ti Mobile GPU has a TGP of 175W -- including Dynamic Boost -- which is 10W higher than the RTX 3080 non-Ti.