Graphics Cards - Page 33

Hardware makers are always trying to find innovative ways to differentiate their products from the other offerings out there, some of which are bold and clever ideas, and others - well, not so much.

So, here's a new one in the world of GPUs: fragrance.

Yes, in a similar idea to the built-in fragrance dispensers that ASUS has been incorporating into some laptop models, and indeed peripherals, Yeston has equipped its new RX 9070 XT Sakura Atlantis graphics card with faint hint of sea breeze.

According to established leaker wxnod, we have a date and time for NVIDIA's GeForce RTX 5060 Ti launch. The post, which went live over the weekend, states that the 8GB and 16GB variants will launch on April 16 at 9 p.m. (which we assume is based in Asia). NVIDIA has yet to announce the GeForce RTX 5060 series; however, this information aligns with the leak/rumor we reported last week that said the RTX 5060 Ti was coming in mid-April.

According to that source, the baseline GeForce RTX 5060's launch has been delayed to mid-May, so the RTX 5060 Ti will arrive first. As seen with the GeForce RTX 30 and ERX 40 Series, it's also expected that NVIDIA will create a Founders Edition model for the GeForce RTX 5060 Ti.

In the past couple of days, we've also seen the full specs for the GeForce RTX 5060 Ti leak out from its 4,608 CUDA Cores to its Boost Clock speed of 2572 MHz. With up to 16GB of GDDR7 memory on a 128-bit bus, the GeForce RTX 5060 Ti will offer significantly more memory bandwidth than the GeForce RTX 4060 Ti. However, as we get down to the mainstream cards for the RTX Blackwell generation, pricing will be a key factor in determining the card's overall value.

In recent weeks, we've received details on NVIDIA's upcoming mid-range GPUs, the GeForce RTX 5060 and RTX 5060 Ti. According to leaks, the GeForce RTX 5060 Ti is set to arrive in two flavors: one with 8GB of GDDR7 VRAM and one with 16G on a 128-bit bus. Like the mid-range RTX 5070 series, it's set to see competition from AMD with the RDNA 4-powered Radeon RX 9060 XT.

According to new product listings from the South Korean Radio Agency (via @harukaze5719 on X), the Radeon RX 9060 XT will also arrive in two flavors: 8GB of GDDR6 VRAM and 16GB - on a 128-bit bus. However, as AMD's new RDNA 4 GPUs all use older GDDR6 technology, the overall bandwidth will be slower than NVIDIA's cards with the latest GDDR7.

The new product listings cover various Radeon RX 9060 XT cards from ASUS. Only the premium variant will arrive with 16GB of VRAM. Both the ASUS Dual and Prime models will ship with 8GB of memory, in OC and non-OC form, with the ASUS TUF Gaming model set to ship with 16GB of memory, also in OC and non-OC form.

NVIDIA's new beefed-up RTX PRO 6000 Blackwell GPU has turned up on US-based IT reseller Connection, with a price of $8435 (if you buy it in bulk) or $8565 (boxed).

The new NVIDIA RTX PRO 6000 Blackwell GPU comes with a huge 96GB of GDDR7 memory, 3x the GDDR7 memory that the new gaming flagship GeForce RTX 5090 has. There are a couple of different RTX PRO 6000 Blackwell GPUs, with the RTX PRO 6000 Blackwell Workstation Edition, and the RTX PRO 6000 Blackwell Max-Q Workstation Edition.

The difference here is that the Max-Q variant is capped at 300W, while retaining 88% of its AI performance according to NVIDIA. NVIDIA's upcoming RTX PRO 5000 Blackwell GPU will launch with 48GB of GDDR7 memory and a price between $4439 and $4569, while the RTX PRO 4500 comes with 32GB, and the RTX PRO 4000 with 24GB.

NVIDIA's previous-gen GeForce RTX 4090 graphics card ships with 24GB of GDDR6X memory, but China being China, and we've got modded RTX 4090s with 48GB of VRAM being sold for $3400.

The custom-modded RTX 4090 has 48GB of GDDR6X memory and a custom AIO cooler, confirmed on Reddit by u/CeFurkan. The AIO cooler on the custom GeForce RTX 4090 has a triple-fan cooler attached to the radiator, with the retail packaging clearly showing "GeForce RTX 4090 48GB".

BYKSI makes the AIO cooler that you see on the modded RTX 4090 48GB graphics card, with the 48GB of memory seen in both Windows Task Manager as well as GPU-Z.

A free NVIDIA RTX 5080 from MSI sounds like a too-good-to-be-true offer, but while it does seem to be genuine - as far as we can tell from the reports - it's certainly one of the strangest promotions we've ever seen, and not one you'll likely to be able to avail yourself of.

For starters, you'll need to live in Taiwan to be eligible, as MSI's offer is exclusive to that country.

If you live there, all you need to do is collect a set of MSI Dragon 'Lucky Around The World' figurines - there are 9 in total (well, actually 10, as there's a bonus 'secret edition' model). Sounds easy? Well, it very much isn't.

With NVIDIA launching four GeForce RTX 50 Series GPUs this year, the company has yet to formally announce its mainstream RTX Blackwell offerings - the GeForce RTX 5060 and GeForce RTX 5060 Ti. Based on leaks from reputable sources, we have quite a few specs on these cards (see below), alongside word that NVIDIA plans to launch both 8GB and 16GB variants for the GeForce RTX 5060 Ti.

When we can expect these cards to hit retail shelves is up in the air - the rumored late-March launch for the GeForce RTX 5060 Ti has come and gone. Although NVIDIA doesn't usually announce new gaming hardware at GDC or its GTC events, many expected an RTX 5060 Series reveal in the past week.

According to information from Board Channels (via Videocardz), a forum where NVIDIA partners offer information on things like supply, the current word is that the GeForce RTX 5060 Ti's launch has been delayed to mid-April, with the GeForce RTX 5060 arriving in mid-May.

The NVIDIA App has been updated to version 11.0.3, which brings several new features to the table, including Optimal Settings support for seven titles. The most notable update is the new ability to customize the DLSS Override settings when enabling DLSS 4, which allows "precise tuning of the new transformer model to optimize performance or enhance image quality." You can set the render resolution percentage to find the right sweet spot between Ultra Performance (33%) and DLAA (100%).

Control over the input resolution is an excellent feature, as DLSS 4's new transformer model significantly improves image quality at all resolutions - adding another level of control beyond choosing Performance (50%), Balanced (58%), or Quality (67%). This NVIDIA App update also adds legacy Control Panel features, namely "Display Scaling Settings," which let you choose from "different scaling modes, including integer scaling" and "Display Color Settings" - another highly requested feature.

Optimal Settings, NVIDIA's custom in-game graphics settings tuned for various GeForce RTX GPUs that offer the best balance of performance and image quality, are now available for Avowed, DYNASTY WARRIORS: ORIGINS, FINAL FANTASY VII REBIRTH, Kingdom Come: Deliverance II, Marvel's Spider-Man 2, Sid Meier's Civilization VII, and SUPERVIVE.

If you didn't think that high-end versions of NVIDIA's RTX 5090 were expensive enough - and they most certainly are - a new version of the ASUS ROG Astral RTX 5090 is going to come with a price tag that'll surely make any wallet quake with fear.

Mainly because this graphics card is decked out in gold (the shroud is golden) and signed by none other than Jensen Huang himself (at GTC 2025).

The ASUS ROG Astral flagship is pricey enough as it is - the liquid cooled edition already weighs in at $3,410 on the ASUS store in the US. So, a golden and signed version of the ROG Astral is going to command a price tag that far exceeds a mere few grand.

AMD recently hosted an AI PC summit in Beijing, with local media reporting that the company said its initial stock off Radeon RX 9070 series "RDNA 4" GPUs had almost sold out, with over 200,000 graphics cards shipped.

The amount of graphics cards shipped doesn't mean graphics cards sold, but AMD did say that the global supply of around 200,000 units of Radeon RX 9070 series graphics cards had almost sold out.

Benchlife was at the event, reporting that AMD said: "Recently, on March 66th, the Radeon RX 9070 XT and Radeon RX 9070, based on the new RDNA 4 GPU architecture, were released. The initial global supply of over 200,000 units of these graphics cards was almost sold out. Following this, on March 12th, the Ryzen 9 9950X3D and Ryzen 9 9900X3D were released, offering gamers and multimedia enthusiasts a more balanced and comphrehensive performance boost".

Another image of what's purportedly a reference RX 9070 board from AMD has surfaced online, though we should obviously be very careful about accepting the pictured GPU at face value.

VideoCardz noticed the image (via Olrak29 on X), which comes from a forum member over at Chiphell, a Chinese source of plenty of hardware leaks, but far from the most reliable one it must be noted.

The picture shows a pretty plain black graphics card with two fans, and it's presumably a prerelease sample of what would be a 'Made by AMD' version of the RX 9070.

AMD's RX 9070 GPU may witness more robust levels of supply after April, according to info that comes direct from a Chinese graphics card maker.

Yeston, who you've likely heard of, is a sizeable force in the GPU market over in Asia, and the firm made the revelation on X, as you can see above.

The wording leaves a slight question mark for us, though, as the arrival of a more stable level of supply could happen 'after' April begins or 'after' April is over. We assume it's the latter, as that'd make more sense, but to us, it could be read as the former, too.

Texture compression has been used in PC gaming for decades, ever since the first texture was applied to a basic 3D character model or environment. In 2025, with the rise of 4K gaming across the PC and console space, these high-resolution textures often contain several layers. The dinosaur example below shows eight 4096 x 4096 PBR textures with 13 channels of surface detail covering roughness, ambient occlusion, and other properties.

Add this all up, and you've got 200 million texture values per texture that would need to be stored in memory (the VRAM on a GPU) for the single in-game asset. Regarding gaming in super-high resolution with ultra settings in 2025, this is one of the reasons why the general rule of thumb is - the more VRAM, the better. Even so, new AI-powered methods of texture compression will significantly improve performance.

At GDC 2025, Intel's Anis Benyoub is joining Microsoft on stage as it showcases 'Cooperative Vectors on DirectX,' a new feature that will lead to a 10X improvement in inference performance for Neural Block Texture Compression on Intel Arc GPUs. Joe Rozek from AMD and Alexey Panteleev from NVIDIA will also be there to discuss how Cooperative Vectors will be used with Radeon and GeForce RTX hardware.

AMD supposedly operates a two-tier system for supplying chips to its graphics card making partners, or that's the belief of the rumor mill.

As VideoCardz reports, this comes from the Board Channels forum over in China, which is always a source we regard with some caution, but admittedly has proved correct with speculation in the past.

The assertion here is that AMD's top-tier board makers - known as 'core' brands - are ASUS, PowerColor and Sapphire, as well as Vastarmor and XFX.

The upcoming GeForce RTX 5060 specs leaked from industry insider @kopite7kimi a week ago. With 3840 CUDA Cores, a 25% increase over the GeForce RTX 4060, it should offer a decent performance increase over one of the most popular GPUs among PC gamers. The leak confirmed 8GB of GDDR7 VRAM for the GPU, offering a 50% bandwidth increase with the same capacity.

Recently, a retail listing from an unknown Chinese retailer (via Videocardz) listed higher-than-expected pricing for the upcoming GeForce RTX 5060 and RTX 5060 Ti. The interesting thing about the listing is that it shows the GeForce RTX 5060 having 12GB of VRAM - the first time we've seen this configuration online.

Based on the source's unknown quantity, it's probably safe to say that the GeForce RTX 5060 will launch with 8GB of fast GDDR7 memory. This has been widely reported for a while now. Ultimately, this has raised our hopes of a new mainstream GPU from NVIDIA with 12GB of VRAM especially when we're reminded that the popular GeForce RTX 3060 launched with 12GB of VRAM back in 2021.

It's not often you see a massive graphics card partner exit from the GPU market, with the last big exit in recent memory being EVGA in 2022. Now MSI can be added to that list, but just for Radeon RX 9000 series graphics cards.

In a recent statement to Tom's Hardware, MSI has confirmed it will not be manufacturing any graphics card based on AMD's RDNA 4 architecture, which means the company will be skipping out on the entire line-up of RX 9000-series graphics cards.

The partnership between MSI and AMD isn't completely dissolved, as the company will keep the door open for future collaborations, but nothing will be manufactured for this current generation. For comparison, MSI manufactured 45 models of AMD's RDNA 2 generation or the RX 6000-series graphics card. As for the RDNA 3, or the RX 7000-series, MSI only made four models, which equates to a massive reduction of 91% between the two aforementioned generations.

NVIDIA's just-released GeForce RTX 5070 is the most popular graphics card on Amazon right now, dethroning AMD's also just-released Radeon RX 9070 XT.

AMD rapidly rose to the top of Amazon's most-wanted GPU list with its new RDNA 4-powered Radeon RX 9070 series GPUs, but now that NVIDIA's new mid-range GeForce RTX 5070 is here, the RTX 5070 is now the top dog (at least on Amazon). NVIDIA launched its mid-range RTX 5070 the day before the RX 9070 XT, but it took a week before it could topple the flagship RDNA 4 card on Amazon's most-wanted GPU list.

At the very tippy top of Amazon's best-selling graphics cards is the custom ASUS TUF Gaming GeForce RTX 5070 12GB OC graphics card, MSI is second with its previous-gen RTX 3060, while third place (at least at the time of writing) is GIGABYTE's custom Radeon RX 7600 XT GAMING OC 16G graphics card.

Take your pick of any recent GPU launch, and you'll probably find that the majority of cards sold were premium partner cards with enhanced cooling and overclocked settings. This has been a thing in the GPU space for years, where PC gamers spend a little extra for that premium build or stylish model with fancy cooling and LED lighting. The only problem is that premium GeForce RTX 5080 and Radeon RX 9070 XT cards are currently being sold at prices that are 40% higher than the MSRP set by NVIDIA and AMD.

This massive discrepancy is a recent phenomenon and has been a hot topic among the enthusiast PC gaming community for several reasons. One, if you're unable to buy, say, buy a GeForce RTX 5070 or RTX 5070 Ti at the MSRP set by NVIDIA ($549 and $749 respectively), then the next model up might carry a 20-30% higher price which diminishes a chunk of the card's value. In the worst examples it also means that you could end up paying more for a GeForce RTX 5070 than the MSRP of the more powerful GeForce RTX 5070 Ti.

During a recent media pre-brief with NVIDIA, covering the company's latest GeForce RTX and AI announcements as GDC kicks off, we asked about MSRP pricing as a current issue and hot topic.

"Demand for the RTX 50 Series has been off the charts," NVIDIA's Jason Paul said during a recent media briefing event for the company's GeForce RTX and AI announcements for GDC.

With the Game Developers Conference kicking off next week in San Francisco, NVIDIA is announcing several exciting things for developers and gamers alike, from RTX Neural Shaders coming to DirectX to Half-Life 2: An RTX Remix Project getting a playable demo that anyone with a GeForce RTX GPU can experience. There's a lot to get through so stay tuned to TweakTown for all the news.

With the GeForce RTX 50 Series launch, four new models - the GeForce RTX 5070, RTX 5070 Ti, RTX 5080, and RTX 5090 - arrived within six weeks. One key problem with the launch, from the perspective of PC gamers looking to upgrade or put together a new build, is availability.

Remember the crypto-mining insanity, where trying to buy graphics cards was nigh impossible? Well, NVIDIA's new GeForce RTX 5090 is being sold on Chinese black markets for $5000... with some listings on eBay seeing the RTX 5090 selling for between $5000 and an out-of-this-world $10,000... yeah, $10K for the RTX 5090.

The Information reports: "DeepSeek's skyrocketing popularity has boosted demand for NVIDIA's gaming chips in the underground market as tech firms realize they can run DeepSeek's models at a fraction of the price of AI chips, according to five chip smugglers".

It looks like Chinese companies have discovered that it's far cheaper running DeepSeek AI models on NVIDIA's consumer-focused GeForce RTX series GPUs rather than more expensive -- and hard, or impossible to get because of US export restrictions -- like the H20, H100, etc. China only has access to "stripped-down" variants of its GPUs like the GeForce RTX 5090D, and that demand is so strong that these models are now being sold at "black market" prices with the RTX 5090 (non-D) selling for up to $5000 or more.