Graphics Cards - Page 50

AMD's next-generation UDNA architecture replaces the RDNA architecture, with next-gen Radeon GPUs using the new UDNA GPU architecture with the "big core flagship" is coming back according to the latest leaks.

In a new post on Chiphell, leaker "zhangzhonghao" said that "GPU UDNA is also N3E process" referring to TSMC's new N3E process node, adding that "the big core flagship is back" to fight NVIDIA's upcoming Blackwell-based GeForce RTX 50 series GPUs.

There's nothing else new about the new flagship Radeon GPU, but what we do know is that the UDNA chip will be made on TSMC's new N3E process node. NVIDIA is using ultra-fast GDDR7 memory on its new RTX 50 series GPUs, with AMD's new Radeon RX 9000 series "RDNA 4" GPUs using the current-gen GDDR6 standard. We should expect an upgrade to GDDR7 memory for AMD's next-generation Radeon GPUs... but what speeds, no one knows yet.

NVIDIA's upcoming made-for-China exclusive GeForce RTX 5090D graphics card will have new restrictions in place over the RTX 4090D: stopping AI performance, and gimping cryptomining performance.

The new NVIDIA GeForce RTX 5090D has been discussed on the Chiphell forums, with a "performance cut-off" in AI and crypto-related applications, where after 3 seconds of sustained workload the GPU will automatically lock its performance (through GPU clock speed reduction or throttling performance through on-board software).

The power limits of the RTX 5090D are most likely going to be restricted, making the RTX 5090D a dud for professional workloads. NVIDIA's new GeForce RTX 5090D will also be blocked from any multi-GPU configurations, meaning adding multiple RTX 5090D cards to a system won't yield any performance benefits, unlike the RTX 4090D which could be stacked.

NVIDIA's new GeForce RTX 50 series Founders Edition graphics cards have been through a teardown experience, revealing the exciting new "Double Flow Through" thermal design, check it out:

The first big change to in-house GeForce graphics cards was with the GeForce GTX 10 series, with NVIDIA introducing a blower-style cooler, upgrading with the RTX 20 series to a dual axial design, which was the first time axial fans were used inside of a Founders Edition cooler.

After that, NVIDIA used dual-axial flow-through designs, with one fan on the front and another on the back. NVIDIA increased the size of the GeForce RTX 30 and RTX 40 series Founders Edition coolers, with up to a 3-slot design on the RTX 3090 and RTX 4090 series GPUs. All of that changes with Blackwell, with the new RTX 50 series FE designs using a slick dual-slot cooler for the flagship GeForce RTX 5090 Founders Edition.

Intel launched its first second-generation 'Battlemage' Arc GPU with the Intel Arc B580 on December 13, 2024, and it was a delightful end-of-year surprise. Outperforming the more expensive GeForce RTX 4060 when it came to 1080p and 1440p gaming, the $249.99 price and 12GB of VRAM made it an easy recommendation for those putting together a new budget build.

For more information, check out our reviews of the Intel Arc B580 Limited Edition and ASRock Intel Arc B580 Steel Legend. This week, Intel is launching the GPU's sibling, the lower-tier Intel Arc B570, priced slightly lower at $219.99 USD. With cut-down specs, it's not as exciting as the B580, but early reviews show the 10GB GPU performing on par with the GeForce RTX 4060 in several titles.

Our ASRock Intel Arc B570 Challenger review is coming soon, delayed by CES 2025 dominating January. We're excited to share our thoughts on this new entry-level GPU - so stay tuned.

AMD's next-gen RDNA 4 and Radeon RX 9070 XT launch and reveal have, so far, been confusing for PC gamers. Granted, we've all been citing rumors, leaks, and credible inside sources for months surrounding RDNA 4's big reveal - but all we got at CES 2025 was a little teaser and word that more information was coming sometime in Q1 2025.

Compared to months of leaked specs, performance expectations, and even credible rumors that the Radeon RX 9070 XT and Radeon RX 9070 would launch on January 24, that date is looking increasingly unlikely as the days go by. What we officially know about RDNA 4 is the following: there will be no flagship or halo product this generation; instead, the Radeon RX 9070 XT and Radeon RX 9070 will compete against the GeForce RTX 5070 Ti/4070 Ti and the GeForce RTX 5070/4070.

Still, AMD's RDNA 4 CES strategy was weird because we went hands-on with several new RDNA 4 GPUs at the show, from GIGABYTE to ASUS to PowerColor and ASRock. There are reports of cards already being sold by retailers in several countries. The good news is that we should be getting an RDNA 4 reveal very soon.

NVIDIA is launching its new GeForce RTX 50 Series of graphics cards this year, starting with the flagship GeForce RTX 5090 and GeForce RTX 5080 at the end of the month. In February, the company is expected to launch the GeForce RTX 5070 and RTX 5070 Ti, which are set to arrive at precisely the right time.

According to a new report via the Taiwanese site ITHome and BoardChannels, the GeForce RTX 4070 is set to be sold out by the end of January - in that NVIDIA's partners will no longer be able to order new RTX 4070 GPUs. This doesn't mean it'll be impossible to find a GeForce RTX 4070 or RTX 4070 SUPER, but prices will probably go up instead of down as stock clears out.

On the plus side, the GeForce RTX 5070 is launching in February with a competitive price point and, according to NVIDIA, will be roughly 20% faster than the GeForce RTX 4070 in terms of raw gaming performance. That figure increases to 2X with DLSS 4 and the new Multi Frame Generation, exclusive to the RTX 50 Series.

Brazilian modders are patiently awaiting NVIDIA's new ultra-enthusiast flagship GeForce RTX 5090 graphics card to be launched, preparing faster 32Gbps GDDR7 modules to drive upwards of 2TB/sec of memory bandwidth thanks to the huge 512-bit memory bus on the RTX 5090.

NVIDIA's new GeForce RTX 5090 will launch with 32GB of GDDR7 memory on a 512-bit memory bus, using 30Gbps GDDRR7 modules, but Brazilian modders are at the ready with huge memory upgades. We're looking at the team soldering faster 32Gbps GDDR7 modules, and higher rcapacities.

The second-fastest Blackwell gaming GPU -- the GeForce RTX 5080 -- will launch with faster 32Gbps GDDR7 modules, but with 16GB capacity. The fastest Blackwell gaming GPU -- the GeForce RTX 5090 -- launches with slightly slower 30Gbps GDDR7 modules, but a larger 32GB capacity. The change from 30Gbps to 32Gbps GDDR7 modules on the RTX 5090 would enable 2TB/sec+ memory bandwidth... and overclocking fun can begin.

Global NVIDIA exports are the latest casualty of the international AI arms race, with the US Department of Commerce announcing heavy restrictions on the international sale of advanced computer chips, specifically GPUs, used for artificial intelligence.

As detailed by Semafor, the move aims to build a "trusted technological ecosystem around the world" and "protect against the national security risks associated with AI". The proposed framework would affect nations worldwide, with certain allied countries and partners exempt from the controls.

The details of the regulations, as reported by WSWS, include a 3-tier system. The first tier, compromising US allies including Australia, New Zealand, South Korea, and Taiwan - would receive no restrictions. The second tier, consisting of approximately 120 countries, will have limits set on the number of AI chips they are able to receive through a license approval system. The third tier, and the main aim of the regulations - would include total bans on exports from China, Iran, Russia, and North Korea.

With the arrival of the GeForce RTX 50 Series, we're starting to see some new GPU designs from NVIDIA's partners - and ZOTAC GAMING is set to introduce its brand-new SOLID GPU design. Available in OC and non-OC variants, the ZOTAC GAMING GeForce RTX 5090 SOLID and ZOTAC GAMING GeForce RTX 5080 SOLID are expected to launch later this month. These will be the company's more affordable MSRP models.

ZOTAC is also releasing an updated ZOTAC GAMING GeForce RTX 5090 AMP Extreme INFINITY GPU alongside a GeForce RTX 5080 variant that includes the company's flagship Infinity Mirror Design for lighting as well as the more blocky and bronze accents seen in the new SOLID design.

ZOTAC notes that both include a 34% larger vapor chamber for enhanced cooling, updated BladeLink fans, reinforced frames, and metal backplates. ZOTAC has also released specs for the cards - with the GeForce RTX 5090 SOLID OC shipping with an out-of-the-box clock speed of 2422 MHz compared to the non-OC variant's 2407 MHz.

GIGABYTE's custom GeForce RTX 5080 AORUS Xtreme Waterforce graphics card has been listed early by a Finnish retailer, costing over $1700.

The new RTX 5080 AORUS Xtreme Waterforce is the flagship custom RTX 5080 from GIGABYTE on the retailer's list, with Finnish retailer ProShop, preparing PC gamers with early GeForce RTX 5080 listings. GIGABYTE's upcoming GeForce RTX 5080 WindForce 3 OC graphics card starts at €1229 or US$1258, which includes 20% VAT.

NVIDIA has an MSRP for the GeForce RTX 5080 in the EU region, and without that 20% VAT, the price of the custom RTX 5080 will fall to around $1000. GIGABYTE has higher-end custom RTX 5080s like the Gaming OC, AERO OC, and AORUS branded cards that are more expensive again.

With the GeForce RTX 50 Series announced at CES 2025, NVIDIA's gaming performance focuses on DLSS 4's impressive update and the arrival of DLSS Multi Frame Generation. Multi Frame Generation leverages AI and RTX Blackwell hardware to generate up to three frames compared to the one with DLSS 3. This sees hardware-intensive path-tracing games like Black Myth: Wukong, Alan Wake 2, and Cyberpunk 2077 hit 200+ FPS in 4K on the GeForce RTX 5080 and GeForce RTX 5090.

As we previously reported, a game like Black Myth: Wukong's performance increases by 8X with DLSS 4 on the GeForce RTX 5090 - going from around 30 FPS to 245 FPS. When it comes to a brand-new generation of graphics cards, one of the big questions always surrounds the performance uplift compared to the previous generation; for the GeForce RTX 50 Series, that would be taking DLSS 4's Multi Frame Generation out of the equation and comparing raw performance or like-for-like DLSS settings.

During NVIDIA's Editor's Day, GeForce product manager Justin Walker stepped us through these numbers for the GeForce RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070 using Resident Evil 4 with RT and no DLSS and Horizon Forbidden West with DLSS as examples of the raw or 'rasterized' performance uplift we can expect.

AMD is set to debut its new RDNA 4-powered GPUs sometime soon. It will formally announce its new Radeon RX 9070 and RX 9070 XT combo at CES 2025. However, without pricing, specs, or performance data, it felt more like a tease than a proper all-out reveal.

Even so, we were able to go hands-on with several upcoming models at the show - design-wise, RDNA 4's GeForce RTX 5070 and RTX 5070 Ti competitors look very cool. This week, one of AMD's key Radeon partners, Sapphire, began teasing its Radeon RX 9070 line-up by posting images of what looks to be an updated version of the company's Pulse models.

With triple-fan and dual-fan variants teased, we're likely looking at the new Sapphire PULSE AMD Radeon RX 9070 and Sapphire PULSE AMD Radeon RX 9070 XT.

NVIDIA's RTX 5090 GPU is set to arrive on January 30, alongside the RTX 5080, and we've just caught sight of the first third-party Blackwell flagship graphics card that pushes its power usage to 600W.

If you recall, the RTX 5090 has an official TDP of 575W as a baseline, which is where NVIDIA's own Founders Edition weighs in. As ever, entry-level variants from third-party board makers will demand the same power consumption, but beefier graphics cards can exact more of a toll on your PC's power supply.

Zotac has just listed its RTX 5090 models, and the AMP Extreme Infinity product tips in at 600W, with a recommended power supply of 1000W.

With NVIDIA set to debut its Balckwell-powered GeForce RTX 50 Series, four brand-new models will soon hit the market. The new GeForce RTX 5080 and flagship GeForce RTX 5090 will debut on January 30, with the GeForce RTX 5070 and GeForce RTX 5070 Ti following sometime in February.

All four will tap into the latest AI advances from NVIDIA, including the brand-new DLSS 4 Multi Frame Generation technology that sees the GeForce RTX 5070 match the previous-gen GeForce RTX 4090 performance in specific titles. That said, the raw performance of the GeForce RTX 4090 is still unquestionably impressive, as we saw in our recent review of the PNY GeForce RTX 4090 24GB VERTO.

As a longtime partner of NVIDIA in data centers, workstations, software, and PC gaming, PNY is set to launch its GeForce RTX 50 Series in two flavors: ARGB and non-ARGB.

AMD was meant to unveil its new RDNA 4-powered Radeon RX 9000 series GPUs at CES 2025 last week, but delayed the unveiling... which is now being called a tactical move, to hurt NVIDIA's impending GeForce RTX 50 series GPU launch.

In a new post on Chiphell, user "nApolean" said that this was a planned move by AMD, as they were waiting to see what NVIDIA was going to do with its GeForce RTX 50 series "Blackwell" unveil, and then come back with a bigger, and badder response.

The only option AMD has here is to absolutely nail the lower-end and mid-range markets with its new RDNA 4 GPU, where we could see the flagship Radeon RX 9070 XT graphics card somewhere in the $500 to $700 range. NVIDIA has a starting price of just $549 for its new GeForce RTX 5070 graphics card, which the company says has "RTX 4090" levels of performance (with DLSS, etc enabled).

NVIDIA's next-generation GeForce RTX 50 series "Blackwell" GPU embargo dates have been leaked, with the RTX 5090 embargo lifting on January 24, and the RTX 5080 embargo lifts on January 30.

In a new post by VideoCardz, we're learning that "things went wrong" with the original release date for the GeForce RTX 5080 which was expected on January 21, "forcing board partners to work overtime to prepare for the upcoming launch". VideoCardz reports that NVIDIA was "persuaded by these AIBs to move the embargo date to January 30".

In previous leaks, it was expected that NVIDIA's new GeForce RTX 5080 16GB graphics card would be launching first, while the new ultra-enthusiast GeForce RTX 5090 32GB would be launching after. It appears that things have changed, and now the RTX 5090 reviews go live on January 24, and the RTX 5080 reviews go live on January 30.

MaxSun has just teased its upcoming GeForce RTX 50 series "Blackwell" GPUs under its Mega Gamer family of graphics cards, with a Dot Matrix display.

In a new post by VideoCardz, after unveiling its new iCraft "Cyber Magic Girl Zwei Edition" graphics card -- ridiculous name, but you've gotta love it -- featuring a white theme and colorful graphics on the backplate. It's a 3-slot design with a triple-fan cooler, based on NVIDIA's new GeForce RTX 5080 with 16GB of GDDR7 memory, expected to launch on January 30.

MaxSun has teased its new MGG (Mega Gamer) design that is a little different to its RTX 40 series Mega Gamer graphics card, and should arrive in the form of the RTX 5080 and RTX 5090 Mega Gamer graphics cards. MaxSun's upcoming GeForce RTX 50 series MGG graphics card will have an attachable dot-matrix display, a very unique feature... in a world with full-color displays on graphics cards, but you do you, MaxSun.

Sparkle announced its new Studio-G 850 at CES 2025, with the new Thunderbolt 5-powered external GPU enclosure handling up to 3.5-slot GPUs. Check it out:

The folks over at HardwareLuxx scoped out the new Sparkle Studio-G 850 external GPU enclosure at CES 2025, which features a pre-installed 850W desktop-grade PSU. The Studio-G 850 measures in at 403 x 200 x 237mm, with USB Type-C power delivery of up to 100W, allowing for a single-cable connection to transmit data and charge a laptop at the same time.

Sparkle's new Studio-G 850 external GPU enclosure supports graphics cards that are up to 3.5-slot in design, offering up to 120Gbps of bandwidth courtesy of the super-fast Thunderbolt 5 connectivity. There are 2 x USB-C ports on the back, as well as 3 x USB-A ports and 1GbE so that it can be installed into your home network without a problem.

AMD's RDNA 4 graphics cards have been at the center of an avalanche of rumors since the RX 9070 models were revealed at CES 2025, and we've just had some fresh indications of performance.

Naturally, take these with a whole lot of caution - for all we know they could be somehow wrong, or fabricated - but they come from Chiphell (again), as noticed by VideoCardz (via KimiGuoLM).

In fact, the Chiphell posts in question has been removed now, but not before VideoCardz took some screenshots and reported on all the relevant details.

OC legend "Der8auer" and his team have reverse-engineered NVIDIA's upcoming GeForce RTX 5090 Founders Edition graphics card using 3D tools, gives us a deep dive before the GPU releases on January 30. Check it out:

Der8auer and his team's findings are based on what we've seen from NVIDIA already, as well as additional info gained from sources, showing how the company hid the power connector on the RTX 5090 FE within the cool space, leaving the area around the main PCB totally exposed for additional airflow.

The dual-slot design of the GeForce RTX 5090 Founders Edition graphics card is an achievement in itself, with Der8auer noting that there are different bends in each fin of the heatsink on the RTX 5090 FE. There are 58 fins on each heatsink stack, noting that what is "kind of unusual and absolutely crazy" is every single fin is bent a little different, as you can see in the shot below.