Graphics Cards - Page 135

Alright... this is a pretty exciting tease in the lead-up to Christmas in the form of NVIDIA's new mysterious "GPU-N" that has been unveiled in a new research paper published by NVIDIA.

The new NVIDIA "GPU-N" was revealed in a recently published research paper titled "GPU Domain Specialization via Composable On-Package Architecture" in which NVIDIA talks about a next-generation GPU design using a "Composable On-Package Architecture (COPA)" as the optimal solution for maximizing low-precision matrix math throughput to increase Deep Learning performance. Simulated performance is teased, which is the most exciting part.

But first, GPU-N specs:

Intel has shown off a pretty useless trailer during The Game Awards, with the new 'Intel Arc Gameplay Trailer' below:

In the video, Intel shows off a few games that aren't that popular... not running at 4K or anything special (Hitman is used as the "next-gen graphics with ray tracing" title, FFS. At the end, we get a glimpse of what would be awesome to see actually happen: an Intel Arc-based desktop gaming PC, and Intel Arc-based gaming laptop.

Intel's teased Arc desktop gaming PC looks damn cool in its chassis with an overlay of your PC readouts (CPU, GPU temps, etc) on the side panel and wicked lighting inside. There's an Alienware-style 16:9 gaming monitor in, well I guess 'Arc style' as well, and to the side what looks to be a 15-inch Intel Arc-based gaming laptop.

We all knew it would be like this, but it seems like the Intel GPU launch is a mess... now that we're weeks away from the end of the year, and thus so very close to CES 2022 and we still know nothing about the launch of Intel's first graphics card in two decades.

Well, we do know a lot of stuff through leaks and not Intel directly... but the launch of Intel's new Arc Alchemist is a mess. Intel has confirmed it will be launching its GPUs in Q1 2022, but other than that there's nothing concrete. But according to Tom @ Moore's Law is Dead, mass production of Intel DG2 GPUs hasn't even started yet... and on top of that, pre-production (qualification) samples aren't shipping to AIB partners until January, over a month away.

Throw in the holidays, and then Chinese New Year...

Something a little surprising has just popped up and flung into the rumor mill: a new series based on the refreshed RDNA 2 architecture known as "Radeon RX 6000S Mobile Graphics".

Yeah, that "S" is really important but for a couple of reasons -- first, AMD has been pumping outs its AMD 4700S Desktop Kit and the more recent and beefier AMD 4800S Desktop Kit -- both with an "S". The new Radeon RX 6800S is expected to use the Navi 22 GPU but on the new TSMC N6 node, down from the N7 node used on the Radeon RX 6800M and its Navi 22 GPU.

TSMC is expecting around 18% higher logic density over the N7 process, with the N6 process fully compatible with the N7 process technology which makes it easier for AMD to transition its GPU onto TSMC's new node. We could expect AMD to unveil its new Radeon RX 6000S series GPUs at CES 2022 alongside its new AMD Ryzen 6000 series "Rembrandt" CPUs in early 2022.

AMD will be introducing its next-gen CDNA 2-based Instinct MI200 series cards in the coming months, with some more news on the Instinct MI210 with a single graphics compute die.

The new AMD Instinct MI210 has been teased with the new CDNA 2 architecture and next-gen "Aldebaran" GPU, which will pack some serious GPU horsepower. The new AMD Instinct MI210 in BabelStream with HIP is around 40% faster than the MI100.

This is according to George Markomanolis, an Engineer working on the upcoming LUMI supercomputer & lead HPC scientist at CSC, who got remote access to the AMD Instinct MI210. The single GCD (graphics compute die) features 104 of the 128 CUs that the full Aldebaran GPU packs with 6656 stream processors, while the higher-end MI250X features 110 CUs per GPU die for a total of 7040 stream processors.

Alphacool has unveiled its new Eiswolf 2 AIO cooling solution for just a handful of very specific graphics cards: NVIDIA's own GeForce RTX 3080 Founders Edition, and MSI's custom Radeon RX 6800, and Radeon RX 6800 XT Gaming X Trio graphics cards.

The new Eiswolf 2 uses Alphacool's 360mm NexXxoS ST30 full copper radiator, while for the fans they've used 120mm Alphacool Aurora Rise fans. To round up the AIO, Alphacool has used its Eisblock Aurora water cooler, which doesn't just cool the GPU chip but also all of the important components around it -- especially the GDDR6/X memory and VRAM -- and of course, beautiful RGB lighting rounding it off.

Alphacool's own DIY range has all of these parts available separately, but you get them all in the package: the radiator, fan, cooler, DC-LT 2 pump, and more. Alphacool adds: "The combination of all components results in a solution that is as easy to install as an AIO, but offers the performance of a pre-assembled and pre-filled custom loop".

ASUS has just unleashed its new GeForce RTX 2060 12GB Dual EVO Series graphics cards, aimed at both gamers and crypto-miners.

The new ASUS GeForce RTX 2060 12GB Dual EVO Series graphics cards are pretty much like every other custom AIB variant: a simple dual-fan design, and a single 8-pin PCIe power connector. ASUS does have a more premium-looking card than the competitors so far with its new RTX 2060 12GB Dual EVO.

ASUS will have two different variants to choose from: the DUAL GeForce RTX 2060 12GB EVO, and the DUAL GeForce RTX 2060 12GB EVO OC Edition. Both of the cards arrive in a 2.7-slot design, with the dual fans being of the ASUS Axial-tech design, with IP5X dust resistance and larger blades that shift more air around.

NVIDIA has made its refreshed GeForce RTX 2060 12GB official, and now Inno3D has unveiled its new GeForce RTX 2060 12GB Twin X2 OC graphics card.

Inno3D's new custom GeForce RTX 2060 12GB Twin X2 OC graphics card rocks the refreshed Turing TU106-300 GPU, 12GB of GDDR6 memory clocked at 14Gbps on a 192-bit memory bus. Inno3D is using a dual-fan cooling design on its card, which should keep the Turing-based GPU and 12GB of GDDR6 memory cool during gaming (and crypto-mining) loads.

There's a basic backplate on the card, with a single 8-pin PCIe power connector required. The dual-fan cooler comprises of 2 x 9cm scythe fan blades, with 3 copper heat pipes that will keep the card "performing at an optimal level, this is especially important when the factory overclocked TWIN X2 OC is tackling the most demanding of games".

NVIDIA is reportedly launching its new flagship GeForce RTX 3090 Ti graphics card on January 27, 2022, according to our friends at VideoCardz.

In a new post, VC says that NVIDIA's embargo for the GeForce RTX 3090 Ti lifts on January 27, 2022, with the only details on the release being on-shelf dates. According to the latest rumors on the RTX 3090 Ti, it will have a much higher 450W TDP, the same but faster 24GB of GDDR6X memory.

NVIDIA will reportedly be tapping new single-sided GDDR6X memory which means the backplate on the new GeForce RTX 3090 Ti shouldn't be able to cook eggs, sorry guys. Other rumors on the RTX 3090 Ti have it packing the new 16-pin PCIe power connector, ready for PCIe 5.0-based PSUs like the ASUS ROG Thor PSUs.

Palit has just unveiled its new GeForce RTX 2060 12GB Dual Series graphics cards, now that NVIDIA has made its new GeForce RTX 2060 12GB official.

The new Palit GeForce RTX 2060 12GB Dual Series graphics cards include the beefed-up 12GB of GDDR6, up from the 6GB of GDDR6 on the regular GeForce RTX 2060. Remember this is a last-gen Turing-based refresh, with the TU106-300 GPU and the 12GB of GDDR6 ramped up to 14Gbps of bandwidth.

Palit is using a regular dual-fan cooler with two 90mm smart fans, with an "optimized thermal solution to enhance the airflow and heat dissipation efficiency. The model offers cool temperature, minimum noise and maximum stability for gamers and creators to enjoy competitive performance".

You might find it hard to buy a graphics card right now, but that doesn't mean AIB shipments haven't slowed down in the last 12 months... in the biggest GPU shortage in history (and no, it's not because of miners).

According to Jon Peddie Research, total AIB shipments were 25.7% higher in Q3 2021 than they were in Q2 2021 -- with 12.7 million graphics cards shipped, up from 11.47 million cards shipped in the previous quarter. The way AIB (add-in boards) work is that they use discrete GPUs (dGPU) with dedicated memory (a graphics card essentially).

AIBs are used in desktop PCs, workstations, servers, rendering and mining farms (see, but not all GPUs are pushed into mining farms like these companies really want you to believe), and scientific instruments. You as a consumer, or a company, would buy AIBs from a reseller (Amazon) or an OEM (ASUS for example).

NVIDIA has officially launched its GeForce RTX 2060 12GB graphics card, with an upgraded Turing-based TU106-300 GPU and 12GB of GDDR6 memory clocked at 14Gbps.

The new GeForce RTX 2060 12GB is a crazy card for the end of 2021, with pricing starting at the very least $560... and more like $700-$800 without an issue, that's if you can find one. NVIDIA has made the new GeForce RTX 2060 12GB as a Turing-based offering that pushes it into the GeForce RTX 2060 SUPER territory, albeit with lower memory bandwidth.

GPU crypto miners will want to take a look at the new RTX 2060 12GB, as you're going to get around 31MH/s mining ETH with just 83W of power or so. This is up against the 36MH/s on the RTX 2070 SUPER for double the power at 168W, the 22MH/s from the RTX 3060 for 116W, and 28MH/s for 107W on the Radeon RX 6600 XT.

NVIDIA has announced that it is hosting its annual CES 2022 press conference on January 4, 2022 -- the same day as both AMD and Intel.

NVIDIA will kick off its CES 2022 press conference on January 4, 2022 @ 8AM PT where we will "Discover the latest breakthroughs in accelerated computing, from design and simulation to gaming and autonomous machines". In laymen's terms: expect the new flagship GeForce RTX 3090 Ti graphics card, as well as the GeForce RTX 3080 Ti Laptop GPU and GeForce RTX 3070 Ti Laptop GPUs to be unveiled.

We should see NVIDIA introduce a new GeForce RTX 3090 Ti Founders Edition graphics card, which is expected to be an even beefier Ampere GPU with the same, but better 24GB of GDDR6X memory. Not only that, but the RTX 3090 Ti is expected to have a huge 450W TDP and single-sided GDDR6X memory that will help with GDDR6X memory temperatures (instead of having the GDDR6X super-heating the backplate on the RTX 3090).

NVIDIA just launched its new Turing-based GeForce RTX 2060 12GB, a weirdly positioned refreshed GPU that is great for crypto miners but not-so-great for gamers... and now we have news of a card for gamers: a GeForce RTX 3080 with 12GB of memory.

In a new SKU list filing with the Eurasian Economic Commission regulatory office, GIGABYTE has confirmed multiple unannounced and unreleased NVIDIA GeForce RTX 30 series GPUs. First off we have the GeForce RTX 3080 12GB, then the GeForce RTX 3070 Ti 16GB, and finally the AMD Radeon RX 6500 XT 4GB.

NVIDIA launched the then-flagship GeForce RTX 3080 with 10GB of GDDR6X, leaving many baffled -- the super-duper-expensive GeForce RTX 3090 had 24GB of GDDR6X and there was nothing in-between. NVIDIA then launched the GeForce RTX 3080 Ti with 12GB of GDDR6X, but looks to be preparing a new GeForce RTX 3080 with 12GB... and it should be GDDR6X.

Intel's new Arc a380 GPU has been teased in a new tweet, where it will reportedly have performance in the vicinity of NVIDIA's mid-range GeForce GTX 1650 SUPER.

The new tweet is from @TUM_APISAK who says that Intel's new Arc a380 graphics card will have GPU boost clocks of up to 2.45GHz, and 6GB of GDDR6 memory. Intel's new Arc a380 will compete with NVIDIA's purported new mid-range GeForce RTX 3050, and AMD's unreleased Radeon RX 6500 XT.

We should see Intel's upcoming Arc a380 graphics card packing the Alchemist (Xe-HPG DG2) GPU which will have 128 Execution Units (EUs) or 1024 ALUs, all within 8 Xe cores. Intel is fabbing the GPU on TSMC's new 6nm process node, so we should be good for some high GPU clock speeds of up to 2.45GHz, even on a lower-end card like the Intel Arc a380.

NVIDIA has just released its new GeForce Game Ready 497.09 WHQL drivers, which offer the best day one experience in ICARUS.

ICARUS will feature NVIDIA DLSS and NVIDIA RTX Global Illumination (RTXGI) technologies, with the new GeForce Game Ready 497.09 WHQL drivers able to rock and roll from the start. Not only do we have Game Ready support for ICARUS, but also the just-released Halo Infinite.

On top of that, we have support for NVIDIA's just-announced GeForce RTX 2060 12GB graphics card, which is a refreshed Turing TU106-based GPU with 12GB of GDDR6 memory launching on December 7. There's also support for 6 new G-SYNC Compatible displays from ASUS, Dell, LG, and MSI.

NVIDIA has announced its new GeForce RTX 2060 12GB graphics card, with it appearing on the official NVIDIA website in all its refreshed glory.

The "new" GeForce RTX 2060 12GB will be powered by the Turing TU106 GPU made on the same 12nm process, with GPU clocks at 1470MHz and 1650MHz for base and boost, respectively. The RTX 2060 12GB packs 12GB of GDDR6 memory on a 192-bit memory bus, with up to 336GB/sec of memory bandwidth.

Comparatively, the GeForce RTX 2060 SUPER has 12GB of GDDR6 on a wider 256-bit memory bus and a heftier 448GB/sec of memory bandwidth. NVIDIA is making the new GeForce RTX 2060 12GB graphics card in Founders Edition, too -- so it won't just be overclocked, custom AIB models of the RTX 2060 12GB but a new FE card as well.

SAPPHIRE has just introduced its new GearBox 500 Thunderbolt GPU enclosure, which can transform any PC or Mac with a Thunderbolt 3/4 connection into a desktop-class PC.

Inside, the new external GPU solution from SAPPHIRE packs their own PULSE Radeon RX 6600 XT 8GB graphics card which offers plenty of performance for the desktop, and some sneaky 1080p 60FPS+ gaming on the side.

SAPPHIRE has updated the chassis design inside of the new GearBox 500 Thunderbolt 3 eGFX enclosure, with improved airflow that keeps the GPU inside of it cooler than before. SAPPHIRE says you can run some pretty powerful GPUs inside of the enclosure, from something as the Radeon RX 570 through to the Radeon RX 6700 XT, and from the NVIDIA GeForce GTX 1060 to the GeForce RTX 3060 Ti.

NVIDIA is preparing its partners in Taiwan for a "major" GeForce RTX 40 series "Ada Lovelace" GPU launch in 2022, with next-gen gaming GPUs tapping TSMC's new 5nm process node.

DigiTimes has a new article behind a paywall that RetiredEngineer shed a light on in a tweet: "NVIDIA gaming GPUs getting major refresh next year, Taiwan factories all-out to support 5nm RTX40 generation. NVIDIA's biennial GPU refresh coming in 2022, riding on metaverse and gaming. Following H100, based on Hopper architecture, using TSMC's 5nm + CoWoS, aimed at datacenter/AI, gaming GPU RTX40 series, based on Ada Lovelace architecture, will also tap TSMC's 5nm....".

NVIDIA's next-gen Ada Lovelace GPU architecture is expected to deliver some truly gigantic performance improvements, as well as a shift away from Samsung's 8nm node process and back into the warm fuzzy fabricating arms of TSMC and its new 5nm process node for the next-gen GeForce RTX 40 series GPUs.

Innosilicon has just held its "Fantasy One GPU Product Press Conference" where it unveiled the new Fantasy One GPU family, and a few interesting new graphics cards.

Starting with the Innosilicon Fantasy One GPU itself, which is based on a multi-chip (chiplet) design and flagship dual-GPU graphics card in the form of the new "Type-B" dual-GPU solution uses 2 x Fantasy One GPUs and are connected using Innosilicon's just-announced Innolink interface.

Innosilicon's dual-GPU card has up to 10 TFLOPs of compute performance, and can handle 32 simultaneous 1080p 60FPS streams, or up to 64 streams at 720p 30FPS. The dual-GPU beast packs 32GB of GDDR6X memory, but they're super limited by the 128-bit memory bus on each of the respective Fantasy One GPUs.