Graphics Cards - Page 125

It looks like AMD is close to revealing its AI-based noise suppression technology, "accidentally" publishing a video teaser to their upcoming NVIDIA RTX Audio competitor on YouTube.

Well, before it was pulled it was screenshotted by the internet of course, and now we know a little of what to expect. AMD's upcoming Noise Suppression technology should work similar to how NVIDIA RTX Audio technology works, using the wonders of AI to clean your audio.

During the pandemic, using a headset with a microphone became invaluable, with NVIDIA stepping in almost immediately with an answer in RTX Voice. AMD's upcoming AI-powered noise suppression technology shouldn't be too far away, with the new RDNA 3 GPU architecture coming in the months ahead.

ASRock has made custom AMD Radeon GPUs for years and years now, with some of the very best -- in fact, the very best custom Radeon RX 6950 XT on the market -- but now the Taiwanese giant is making a custom Intel Arc A380 graphics card.

The new ASRock Arc A380 Challenger graphics card has been spotted, joining the custom Arc A380 graphics card race with GUNNIR and its Arc A380 Photon OC 6G graphics card. ASRock is using 'Intel ARC' branding on the top of the card, just like we see on AMD Radeon and NVIDIA GeForce graphics cards from AIB partners.

We have a small, dual-slot, single-fan graphics card with a single 8-pin PCIe power connector for the custom ASRock Arc A380 Challenger graphics card. Inside, the Arc A380 has the full ACM-G11 GPU on the TSMC 6nm process node with 8 Xe-Cores or 1024 ALUs, with a max GPU clock of 2450MHz. 6GB of GDDR6 is here at 15.5Gbps on a 96-bit memory bus that pumps 186GB/sec of memory bandwidth, with a 92W TBP.

I only got up just over a couple of hours ago, and I'm already head-first into the very latest rumors on NVIDIA's next-gen Ada Lovelace GPU architecture and the new GeForce RTX 40 series graphics cards.

In a series of tweets, leaker "kopite7kimi" teased that NVIDIA is reportedly working on "the beast" which rocks the PG137-SKU0 board and AD102-450-A1 GPU. That GPU would feature an insane 18176 CUDA cores, 48GB of GDDR6X memory at 24Gbps, and a huge 800W of total board power.

But then there was the tweet right after that, where the leaker says that NVIDIA will unveil its next-gen GeForce RTX 4090 graphics card. Kopite7kimi tweeted: "about RTX 4090, we will see her soon. Nothing to say". There's nothing more to say here, but we have to know that the GeForce RTX 4090 is very, very close at this point... but "we will see her soon" sounds very exciting.

NVIDIA is closer than it appears to launching its next-gen Ada Lovelace GPU architecture, with a purported "the beast" GPU powered by the AD102 GPU and a new board design: PG137.

The new NVIDIA PG137 board design is different from the PG139 that will make up the GeForce RTX 4090, with leaker "kopite7kimi" teasing "the beast" GPU: PG137-SKU0. This is reportedly powered by the AD102-450-A1 GPU with 18176 CUDA cores, an insane 48GB of GDDR6X memory at 24Gbps, and around 800W of TBP.

NVIDIA's purported "the beast" GPU with the AD102-450-A1 GPU and its huge 800W of power should require dual 16-pin PCIe power connectors. Quite the insane power leap, but if NVIDIA is scared of AMD and its next-gen RDNA 3-based Navi 31 GPU and its chiplet design... NVIDIA is going to drive up to 800W to beat it.

GPU prices have been slowly coming back down from the stratosphere, where we're back to kinda-regular pricing... and now the GeForce RTX 3080 Ti can be had for $900.

The price of NVIDIA GeForce RTX 30 series GPUs has been coming down over the last month, with the flagship GeForce RTX 3090 Ti already started falling from its $2000+ price down to $1500... and now it can be found for around $1450.

NVIDIA's GeForce RTX 3080 Ti can be purchased for $899... down from $930 with a $30 rebate card at Newegg, but down a big chunk from its $1100+ price from just weeks ago. That's for the custom triple-fan MSI Ventus GeForce RTX 3080 Ti graphics card, while the GeForce RTX 3090 can be purchased for $1250. Not too bad at all, and you get the 24GB of GDDR6X (up from the 12GB GDDR6X on the RTX 3080 Ti).

Graphics chip and technology company NVIDIA has announced that they will be showcasing their latest innovations, as well as cutting-edge AI and CG in simulation, collaboration, and design at this the SIGGRAPH 2022 conference located this year in Vancouver, Canada. They have titled their weeklong event "Illuminating the Future of Graphics".

Jensen Huan, NVIDIA's CEO, as well as other special presenters will be giving attendees of the NVIDIA Special Address at SIGGRAPH 2022, a glimpse into the future of AI-infused virtual worlds. There are plans to cover the research and technology that power these worlds, and how they open new frontiers for artistic expression and creativity or perfectly replicate nature's systems.

NVIDIA will also be presenting it's latest academic collaborations in graphics research with 17 technical papers. These research papers span advancements in neural content creation tools, display and human perception, the mathematical foundations of computer graphics, and neural rendering. A few of the research papers have already been published for consumption, and are available online through the links below.

Intel kicked off its weeks-long Intel Xe-HPG Scavenger Hunt event, where gamers could win 100 Premium and 200 Performance Arc Alchemist desktop graphics cards bu finding hidden information that Intel was hiding across its social media pages.

But now that the competition is over, there are 300 Xe HPG Scavenger Hunt winners that get to know, and so do we, which graphics cards that they'll be receiving soon. Intel explained that the "Premium" graphics card for the Grand Prize winners is their new Intel Arc A770 graphics card with 16GB of memory, while the First Prize winners get the "Performance" card in the Intel Arc A750 graphics card with 8GB of memory.

Intel explained that it has "started the ramp toward the full desktop launch" of its Arc A-series GPUs, mentioning its "recent tech press interviews" with Ryan Shrout and Tom Peterson, as well as the "surprise reveal of the Intel Arc Gaming Truck" that is headed towards LANFest Colorado in September.

AMD is still a couple of months away from launching its new RDNA 3-based GPU architecture, with the Navi 31, Navi 32, and Navi 33 to fill out the new Radeon RX 7000 series graphics cards.

The new AMD Navi 31 GPU is expected to debut with a chiplet design, with a GCD (Graphics Core/Compute Die) and MCD (Memory Complex/Compute Dies). You can see in the new Navi 3X renders from Wild_C on Twitter that the Navi 31 GPU is going to be a bigger design than the Navi 32 -- which is also a chiplet design -- while Navi 33 is a regular monolithic GPU design.

As for the die sizes, Wild_C says that we should expect Navi 3X die sizes of (roughly):

Just imagine... you've ordered yourself an expensive, new, super-fast flagship NVIDIA GeForce RTX 3090 Ti graphics card and it arrives in the mail... but there's no graphics card inside.

This is what happened to a Brazilian man who ordered himself a Palit GeForce RTX 3090 Ti GameRock graphics card, but instead found himself with sand instead. In a not-so-funny coincidence, AIB partners did reportedly tell NVIDIA to "pound sand" over Turing and Ampere stocks sitting in warehouses ahead of the Ada Lovelace GPU launch.

Brazilian gamer Mauricio Takeda ordered a Palit GeForce RTX 3090 Ti GameRock graphics card, opening it to find pots of sand. He did notice that the weight of the package was weird once he received it, asking his wife to record it... which is how we have these pictures. It's pretty ridiculous, with Takeda saying he's usually a calm person but wasn't after he saw this.

AMD's next-gen RDNA 3 GPU architecture isn't too far away now, with the flagship Navi 31 GPU rumored to roll out with a 350mm² graphics die... smaller than the current RDNA 2-based Navi 21 GPU.

In a new tweet, leaker "Greymon55" says that the next-gen AMD Navi 31 GCD (Graphics Complex Die) comes in at a small 350mm², making it much smaller than the 520 mm² on the Navi 21 GPU. VideoCardz has a mock up of the Navi 21 GPU next to the Navi 31, where you can see the GCD in the middle with the MCD (Memory Complex Die) flanking the GCD.

AMD is reportedly using 1 x GCD and up to 6 x MCDs through advanced packaging technologies, where we're expecting to see hybrid 5nm + 6nm nodes being used. AMD is reportedly using this hybrid 5nm + 6nm advanced packaging technology on its Navi 31 + Navi 32 GPU designs. Navi 33 is reportedly secure on 6nm, no hybrid 5nm + 6nm GPU there.

I'm sure you can tell that the tech press isn't happy with what Intel has done with what they're calling the "launch" of their new desktop Arc Alchemist GPUs -- first out of the gate is the new Intel Arc A380 -- a total mess of a launch.

Intel is on a roll with its absolute disaster of GPU launch with marketing campaigns shilling videos on LinusTechTips, where it's all smiles... but the real reviewers, they say the Intel Arc A380 discrete GPU is "like living in the middle of a minefield, while playing drunk".

While Intel is bringing suitcases of cash and some Intel Arc A770 graphics cards to YouTubers, traditional tech media that has been around before YouTube was even a word, have been left out of the cheques and cards. Some media worked together to secure an Intel Arc A380 from China, saving on shipping fees. Golem, ComputerBase, and Igor's Lab (all out of Germany).

Intel is throwing some money around... I mean, on a marketing parade for its first GPU since the Intel i740 from the 90s... with its new Arc A770 flagship benched at LinusTechTips.

GPU reviewers aren't getting their hands-on samples, so Intel is doing what Team Red used to do under RTG's broken-apart GPU marketing team from AMD (that are mostly now at Intel, which explains things) but they've got time to stop by GamersNexus, LinusTechTips... and now, a shilly new video.

In the expensive new video, Intel shows off its new flagship Arc A770 desktop GPU that the company says is going to "kill everyone in price to performance" in optimized DX12 titles. Furthermore, and for the largest "Tech YouTuber", this video is embarrassingly slim on actual performance details (frame rates, settings, etc). The video does show Tomb Raider and F1 2021 running DX11 + DX12 APIs, but other than that... Linus must've had a large suitcase to keep those details under wraps.

GUNNIR is the first out of the gate with an Intel Arc A380 graphics card, officially releasing its new GUNNIR Intel Arc A380 Photon 6G OC graphics card... at least in China.

The new custom GUNNIR Intel Arc A380 Photon 6G OC graphics card isn't something that you'll want to run out and buy right now, it's just too late, and too slow against many other competitors from AMD and NVIDIA that you could not just buy right now, but months ago... years ago.

If you're in China, you can secure yourself a custom GUNNIR Intel Arc A380 Photon 6G OC graphics card for 1299 RMB which works out to around $192 USD. Intel isn't sending any of the GUNNIR Intel Arc A380 graphics cards out to reviewers, as it seems ex-reviewer and now Intel GPU marketing guy Ryan Shrout, is about the only one with hands-on with it.

NVIDIA's new GeForce RTX 4090 has been in the headlines this week with teases of 66% more performance in synthetic 3DMark benchmarks, but now we're hearing rumors of the "full-fat AD102 GPU" and some truly brute performance.

In a new tweet, the "hunter of fake leaks" aka @XpeaGPU teased: "Got first in game score. Full AD102 at "high power draw" (whatever does it mean, maybe OC?) in Control Ultra 4K RT + DLSS reaches 160+fps. Don't know which CPU, RAM, GPU clock, but it looks very high to me. What a 3090Ti does in this bench?"

Now, if we compare this against NVIDIA's second-top Ampere-based GeForce RTX 3090, which can run Control at 4K with RT + DLSS at around 72FPS average (thanks Hassan @ Wccftech) which means the new full-fat AD102 GPU with "high power draw" absolutely destroys the GA102-based RTX 3090 with a huge 160FPS+.

NVIDIA's new GeForce RTX 4090 graphics card is getting closer and closer, with the very latest on the next-gen GPU and some 3DMark TimeSpy Extreme benchmark scores.

The new AD102-powered GeForce RTX 4090 reportedly scores over 19,000 points in 3DMark TimeSpy Extreme, making it 62% faster than the GeForce RTX 3090 Ti, and 82% faster than the GeForce RTX 3090. Bloody impressive, eh? For comparison, personally, I overclocked MSI's custom GeForce RTX 3090 Ti SUPRIM X and hit just under 12,000 points in TimeSpy Extreme.

Leaker "kopite7kimi" tweeted rather simply "RTX 4090, TSE > 19000". Remember, this is a synthetic benchmark and not a game so we should expect similar, if not better (and sometimes worse) performance in games from NVIDIA's next-gen GeForce RTX 4090. The last we heard on the GeForce RTX 4090 is that it'll be powered by the AD102 GPU with 450W, while NVIDIA is reportedly saving up another AD102 with 600W TDP that will be released after AMD unleashes its next-gen RDNA 3-based Navi 31 GPU.

Intel has officially unveiled its upcoming Arc A750 desktop GPU, with the Intel Arc A750 Limited Edition graphics card being shown off for real this time... kinda.

Ryan Shrout, ex-PC Perspective, now Intel GPU marketing boss, has benchmarked and shown off the Intel Arc A750 graphics card. Inside, there's the Intel ACM-G10 CPU, which Shrout benchmarked in Cyberpunk 2077 at 2560 x 1440 on "High" settings with 60FPS average -- 17% faster than NVIDIA's GeForce RTX 3060.

Ryan Shrout explained: "Performance results shown here are from a small subset of the games, that work very well with Intel® Arc™ and the Alchemist architecture. I'm not asserting that ALL GAMES will show these results, but it's a view of what Intel Arc A-series cards are capable of with the right software and engineering enablement".

NVIDIA hasn't announced or confirmed its next-gen Ada Lovelace GPU-powered GeForce RTX 40 series graphics cards, but now we know the flagship GeForce RTX 4090 is supported in AIDA64.

The very latest AIDA64 6.70.6033 update includes support for the "GeForce RTX 4090" with the AD102 GPU, as well as the new lower-end GeForce GTX 1320, and AMD's new Navi 23-based Radeon Pro W6600 workstation GPU.

It seems rather early to support a graphics card that won't be coming for months, but NVIDIA will unleash its new AD102-based GeForce RTX 4090 in October in 450W according to the latest information. NVIDIA is reportedly waiting for AMD to release its new RDNA 3-based Navi 31-powered flagship GPU before it unlocks its higher-end AD102-based GeForce RTX 4090 Ti, which should have 600W+ power ready to smash benchmarks and games.

Intel will be launching its new Arc A-Series desktop graphics cards later this year, with many leaks and now official teases and confirmations from the Intel GPU team in the last week... now, we have news on the Intel Arc desktop GPU pricing.

The folks over at Wccftech have now leaked the pricing on the new Intel Arc GPUs will be led by the Arc A770 graphics card that Intel says will be priced at up to $399, which will be for the 16GB GDDR6 variant, leaving an 8GB GDDR6 variant to be even cheaper. Intel has reportedly been showing this chart with their major Taiwanese partners (think the likes of ASUS, GIGABYTE, MSI, etc) according to the leaks.

If the chart is true, Intel tops out at $400 for the Arc A770 graphics cards and calls the $400-$499 range the "enthusiast" GPU market. Under that, we'll have the Intel Arc A750 with 8GB GDDR6 for somewhere in the $280-$340 market or so, while there'll be the Arc A580 with 8GB GDDR6 for $200-$240. Under that there's the Arc A380 with 6GB GDDR6 for under $100, and the Arc A310 for under $100.

Intel let some of its people out of the Arc GPU labs for some marketing campaign runs with YouTubers, with both Gamers Nexus and LinusTechTipsg getting some hands-on with Arc GPUs... but LinusTechTips got something a bit more special.

The new Intel Arc A770 Limited Edition graphics card was unveiled during the last LinusTechTips WAN show, where the plain-ish, but still good-looking Intel Arc A770 Limited Edition graphics card has a sleek dual-slot design, 8+6-pin PCIe power connectors... oh, and some RGB lighting.

Ex-NVIDIA staffer, now Intel Arc GPU team member Tom Peterson confirms that the Intel Arc A770 that is fast enough for you to play games on, and it also overclocks "really well" he says. Inside, it has the full ACM-G10 GPU with 32 Xe-Cores, and it should have 16GB of GDDR6 memory, too.

Intel has ramped up its Arc GPU marketing over the last week, with the company throwing some weight into getting Arc marketing peeps Ryan Shrout (ex-PC Perspective) and Tom Peterson (ex-NVIDIA) onto the sets of YouTubers. But now... a big flex is coming from Intel marketing boss Ryan Shrout.

In a tweet out of nowhere, with absolutely no context and even doing a quick Google, results on it come up scarce -- and I should know, I write GPU content, live, breathe, and report on tech industry rumors for a living -- the Intel Arc A780. Shrout tweeted: "Despite some rumors to the contrary, there is no Intel Arc A780 and there was never planned to be an A780. Let's just settle that debate".

We will be seeing the Intel Arc A770 and Arc A750, both of which we've heard about, reported on, and now the Arc A750 has even been teased in the flesh. But the Arc A780, which was never really a thing, was "never planned". Cool, thanks Intel. There is no debate to settle, but Intel marketing is definitely taking things down a weirder path than AMD and NVIDIA... even after having ex-press that should have some power to do things differently after jumping across the fence from tech press to tech marketing for Intel.