Graphics Cards - Page 55

Samsung will be showcasing its next-gen 42.5Gbps GDDR7 memory modules at ISSCC between February 16-20, 2025, ready for what could materialize into NVIDIA's GeForce RTX 50 SUPER series GPUs.

How do we know this? Because the ISSCC website lists an event from Samsung "Nonvolatile Memory and DRAM" event that is held on February 19, 2025. Samsung will be showing off its flagship GDDR7 memory, which is capable of running at a blistering 42.5Gbps... this is huge, as the first fleet of RRTX 50 series GPUs should feature 28-32Gbps GDDR7 memory.

NVIDIA's initial fleet of GeForce RTX 50 series GPUs will use 28Gbps or 32Gbps GDDR7 memory modules depending on the SKU, so Samsung's ultra-fast 42.5Gbps GDDR7 modules could be used on beefed-up cards like the RTX 5090 SUPER and RTX 5080 SUPER graphics cards.

AMD's SVP and GM of Computing and Graphics, Jack Huynh, has announced that the company's big CES 2025 press event and keynote will take place on Monday, January 6, at 11 a.m. PT. The event will run for 45 minutes, and AMD will unveil its "next generation of innovation across gaming, AI PC, and commercial."

Gaming before AI? Yes, it sounds like the company's upcoming RDNA 4 graphics cards, the mid-range Radeon RX 8800 XT and the Radeon RX 8700 XT, will be a big focus at the event. Jack Huynh teases "incredible special guests" while adding, "You won't want to miss it!" Based on previous Radeon-focused keynotes, we fully expect to see live or recorded game demos alongside performance benchmarks and new features. Hopefully, AMD's rumored AI-powered FSR 4 will be ready to show.

For those keeping track, AMD's CES event will happen before NVIDIA's evening GeForce RTX 50 reveal. AMD hasn't teased any specific products for CES 2025, but the sheer number of leaks and rumors we've seen all but make RDNA 4 a certainty. The company is also expected to unveil its next-gen Strix Halo and Kracken Point APUs, which will be gaming powerhouses in their own right.

With Intel announcing its first Battlemage GPU, the new Intel Arc B580 and B570, the company has also announced that it's updating its software tools and support with the latest Intel Graphics Software. With expanded display settings tools and more control over things like FPS limiters and latency reduction, a robust suite of overclocking controls is even introduced.

Yes, this means the Intel Arc B580, with its 12GB of VRAM, 2670 MHz Graphics Clock, and base 190W power draw, can all be tweaked to boost performance. The new Intel Graphics Software will replace the existing Arc Control software on the PC. It will be the all-in-one tool for GPU and PC game customization, performance monitoring and tuning, and driver releases going forward.

When it comes to overclocking, during a preview of the new software, we were shown how in-depth and easy to use it is - with control over GPU tuning, voltage limits, VRAM memory speed overclocking, fan curve, and thermal tuning.

Intel launched its first generation of Arc graphics cards, the Intel Arc A750, in March 2022. At the time, it received mixed reviews due to driver and compatibility issues, where performance varied wildly from game to game. It also didn't play nice with specific system configurations, leading many to label Arc as 'wait and see.'

Fast-forward to 2024, and Intel's Arc Graphics represents GPUs for desktop systems, laptops, and gaming handhelds. However, it feels like a mere blip on the radar when compared to GeForce or Radeon.

Still, Intel has released over 50 new driver updates, added support for over 120 new game launches, and expanded Arc's game coverage by 2.5X. Older PC games, including several of today's biggest competitive titles, have received massive performance improvements across all Arc Graphics products.

Intel has formally revealed its first Battlemage GPUs, the Intel Arc B580 and the Intel Arc B570. The Arc B580 will be the first to launch; it arrives on December 13, 2024, with a $249 USD price point. The Intel Arc B570 will launch on January 16, 2025, with a $219 USD price point. Both feature the BMG-G21 chip based on Intel's Xe2 architecture.

With affordable pricing, Intel's Arc B-Series is squarely aimed at the mainstream gaming market, where its direct competition will be NVIDIA's GeForce RTX 4060 and AMD's Radeon RX 7600. The Intel Arc B580 sports 12GB of VRAM, and the company is going so far as to position it as a 1440p 'Ultra Settings' gaming graphics card - with a 60 FPS baseline being the target.

Performance is always at the top of PC gamers' lists when a new GPU is announced, and Intel's second-generation Arc graphics card is no different. The company showcased performance benchmarks as part of the announcement, comparing it to NVIDIA's GeForce RTX 4060 and the first-gen Intel Arc A750.

Intel's first next-gen Battlemage desktop graphics cards will be here in a little over a week - with the Intel Arc B580 debuting on December 13, 2024, starting at $249 USD. With the company's new Xe2 architecture and BMG-G21 chip, it'll be joined by the cutdown Intel Arc B570 on January 16, 2025, starting at $219 USD.

In addition to an Intel Arc B580 Limited Edition reference card that Intel says is whisper quiet thanks to its revamped design and increased airflow, B-Series 'Battlemage' partners include Acer, ASRock, Gunnir, Maxsun, Onix, and Sparkle. Multiple B-Series B580 and B570 cards are coming, with Intel telling media that they're already being shipped to retailers.

Although Intel's first Battlemage GPU is not the beefier B750 (Intel wasn't ready to talk about its B-Series roadmap), the B580 is still being positioned as a 1440p Ultra gaming card that can handle modern blockbuster titles with or without ray-tracing. And it's ready to give budget-conscious gamers a different option.

NVIDIA never released the GeForce RTX 4090 Ti, but a quad-slot prototype GPU was found in the dumpster last week... and now that card is being sent to GamersNexus for review.

The experimental RTX 4090 variant has a unique design, and engineering tweaks that would've been nice to see materialize, with the RTX 4090 Ti prototype sharing the core specifications of the retail RTX 4090 Founders Edition. Inside, it features the PG147 SKU 330 board architecture, with power specifications at 450W typical draw, and 600W peak draw.

But, there are some slight changes: the BIOS version is 95.02.01.22.01 with operating frequencies of 2115MHz base, and 2335MHz boost, which means the GPU clock speeds are around 5% lower than the RTX 4090 Founders Edition graphics card.

AMD's flagship Radeon RX 8800 XT graphics card will be the highest-end RDNA 4 SKU, with rumors of its performance and power consumption leaking with some rather surprising results.

The new Radeon RX 8800 XT "RDNA 4" graphics card is the last RDNA 4 card we'll see before UDNA is introduced for the next-gen GPUs from AMD in the future. In a new post on the Chiphell forums, we're learning that AIBs will officially start mass production of the Radeon RX 8800 XT in mid-December, noting there are "still some minor bugs".

As for the power consumption, AMD's new RDNA 4-powered Radeon RX 8800 XT graphics card will consume 25% less power than the RDNA 3-powered flagship RX 7900 XTX. In some quick-and-dirty testing (I'm sure) in Resident Evil 4 Biohazard, ray tracing performance on the RX 8800 XT is reportedly 45% faster than the 7900 XTX.

You're doubtless familiar with Valve's hardware survey for Steam, and amongst the stats provided there is a ranking of the most popular graphics cards - with Intel's Arc brand finally making it onto that list.

The revelation came via a post on X from Haze that VideoCardz spotted, and we should clarify that the new entry in Steam's GPU rankings (for November) is for 'Intel Arc Graphics' which is the collective listing for all Alchemist laptop GPUs. These now have a share of 0.18% of all the graphics cards used on Steam.

These integrated graphics solutions don't register as separate model names, they're all lumped under that one umbrella label, which obviously helps bolster their chances of appearing in the rankings.

After new retail listings for Intel Arc B580 'Battlemage' GPUs from ASRock leaked, alongside details on a Limited Edition reference model from Intel, we've got word on a second next-gen Arc GPU coming. The Intel Arc B570 will be a lower-tier option with 10GB of VRAM and slightly lower specs than the B580.

The information arrives via Videocardz, which posted a full spec sheet for the ASRock Intel Arc B570 Challenger 10GB OC variant. According to the details, spec sheets, and the ASRock Intel Arc B580 Steel Legend 12GB OC, the lower-tier B570 will feature 18 Xe2 Cores compared to the 20 in the B580.

It also confirms that the B570's 10GB of VRAM will feature a 160-bit memory bus, so its overall bandwidth will be notably lower than that of the B580's 12GB of VRAM on a 192-bit bus. With Intel confirming that it's set to formally announce its next-gen Arc desktop graphics cards this week, here's a look at the leaked specs for these two GPUs ahead of the official reveal.

"We've got some big graphics news," the post from the Intel Gaming social media channel confirms, pointing to December 3 at 9 am ET and the Intel Gaming YouTube channel as the destination for its next-gen Battlemage graphics card or cards.

The teaser video doesn't confirm or explicitly state that it will be for the new Intel Arc B580 (and B570) 'Battlemage' GPUs that have leaked, but it does clearly show the outline of a new card that looks like the first-gen Limited Edition models from Intel. So, with retail listings, benchmark results, spec sheets, and reference models all showing up online in the past week or so - we'll finally get our first look at Intel's second-gen Arc graphics for desktop rigs this week.

How this presentation will play out is anyone's guess. Still, with the Intel Arc B580's $250 or so price point - which will also feature 12GB of VRAM - Intel will undoubtedly need to present its overall value as a budget or affordable winner.

Rumors suggest Intel will release its first desktop Battlemage GPU product, the Intel Arc B580, around December 12. We've already seen retail leaks of models online from Intel's partners at ASRock, with new reporting via Videocardz seemingly confirming a reference 'Limited Edition' model of the GPU coming from Team Blue in a couple of weeks.

However, based on the specs and naming, it does look like Intel's second-gen Arc desktop GPU line-up is launching with its entry-level or budget variant first. That's another thing; we also know from the retail listings that the B580 is priced at around $250, making it cheaper than the GeForce RTX 4060 and Radeon RX 7600. Throw in 12GB of VRAM, and it could be a budget winner.

As for those rumors that Battlemage would be Intel's last hurrah in the desktop GPU space, that doesn't look like it's the case. According to @jaykihn0 on X (via NotebookCheck), Intel is developing desktop Arc Celestial GPUs for release in 2025 or 2026, which will be followed up with Arc Druid.

Earlier this week, we reported on retail listings for new Intel Arc B580 GPUs from ASRock being spotted, pointing to an imminent release of the very first Battlemage desktop graphics cards. According to rumors, the entry-level follow-up to the Intel Arc A580 could be here as early as December, with new retail listing leaks (via @momomo_us on X) confirming a price point of around $250 USD.

This price is for the Intel Arc B580 Limited Edition GPU from Intel, which will probably feature a design similar to the reference models we saw for first-gen Arc GPUs. A $250 price point would make the Intel Arc B580 an affordable new GPU with an impressive 12GB of VRAM and a cheaper card than the GeForce RTX 4060 and Radeon RX 7600.

However, $250 is notably higher than the Intel Arc A580's $179 launch price, so the overall value will naturally come down to performance and how it handles 1080p gaming across competitive and cinematic AAA titles. Also, we finally have additional specs for the card so we can see how it stacks up against the previous-gen A580.

NVIDIA's new GeForce RTX 5090 and its Chinese variant in the newly-revealed GeForce RTX 5090D, won't have any differences when it comes to GPU configurations (core count) and clock speeds.

The new GeForce RTX 5090D graphics card was revealed yesterday, but now leaker "panzerlied" has said "there is no hardware difference between the 5090 and 5090D". He was asked about the RTX 5090 vs RTX 5090D in gaming performance, with the leaker adding "I think there is no difference in frequency".

NVIDIA originally launched the GeForce RTX 4090D in a reactionary way, offering a tweaked RTX 4090 to get around the export ban, as its TPP rating exceeded the limits set by US regulators. NVIDIA simply sliced and diced some of the specifications of the original RTX 4090 in order to get the RTX 4090D into the market early, without looking at other ways to tweak the TPP figure.

NVIDIA is reportedly working on a custom GeForce RTX 5090D graphics card for the Chinese market, with new leaks showing the RTX 5090D label for the retail packaging of the new cards.

In a new post on X leaker MEGAsizeGPU showed off the logo to the GeForce RTX 5090D, with VideoCardz noting that if we're wondering why the "D" is there (the D is for Dragon) because 2025 is the Year of the Snake. But, in the Chinese zodiac, the Year of the Dragon ends on January 28... meaning we could see the RTX 5090D released before the end of January 2025.

China won't be the only country getting the cut-down GeForce RTX 5090D, with other countries including Country Groups D1, D4, and D5 (including, but not limited to Saudi Arabia, the United Arab Emirates, Vietnam, but exclusing Israel). These restrictions have workarounds, like AIBs including ZOTAC moving their headquarters out of Hong Kong, now operating out of Singapore, with graphics assembly production lines setup in Indonesia.

NVIDIA is reportedly using Samsung's new GDDR7 memory chips for its next-gen GeForce RTX 50 series "Blackwell" gaming GPUs.

Samsung recently showed off its new HBM3E AI memory and new GDDR7 memory at the recent electronica 2024 event, but now Korean outlet Greened.kr is reporting that Samsung's latest GDDR7 memory chips will be used by NVIDIA's new GeForce RTX 50 series GPUs in early 2025.

Samsung was one of the top three DRAM manufacturers in the race to supply GDDR7 memory chips to NVIDIA for its next-gen gaming GPUs, with fellow South Korean memory rival SK hynix (which dominates the AI memory business with NVIDIA and its HBM3, HBM3E memory chips) and US-based Micron.

There are leaks, and then there are leaks. Product shots and store listings for two Intel Arc B580 graphics cards from ASRock indicate an imminent release for at least one next-gen Intel Arc 'Battlemage' desktop GPU - the B580.

The retail listings for the ASRock Intel Arc B580 Steel Legend and ASRock Intel Arc B580 Challenger cards, both listed as OC models, also confirm some of the specs for Intel's next-gen Battlemage B580 card.

As the successor to the entry-level Intel Arc A580 that launched with a wallet-friendly price point of $179, the B580 is shaping up to be something that could shake up the budget GPU market for desktops with 12GB of VRAM.

AMD is expected to have a "full unveiling" of its next-gen RDNA 4 GPU architecture at CES 2025, with "at least" one new Radeon RX 8000 series graphics card hitting the market in late January.

We know that AMD's upcoming RDNA 4 GPU architecture is the last one before the company unifies its architectures into the UDNA architecture, expected to power next-gen GPU designs inside of graphics cards and APUs of the future. In January 2025 at CES, leaker Moore's Law is Dead says his sources say we can expect RDNA 4 at the show, and at least one SKU launching weeks later.

A bigger note here is that this source said they couldn't confirm if AMD will have its next-gen FSR 4 upscaling technology ready for RDNA 4 and CES 2025. NVIDIA will be stomping in with its next-gen GeForce RTX 50 series "Blackwell" GPUs led by the ultra-enthusiast flagship GeForce RTX 5090, and I'm sure, DLSS 4 upscaling to boot.

With CES 2025 around the corner, NVIDIA's upcoming Blackwell-powered GeForce RTX 50 Series line-up is starting to take shape. According to sources, the company will launch as many as four SKUs within the first few months of 2025, beginning with the flagship GeForce RTX 5090 and RTX 5080 in January.

In addition, NVIDIA is reportedly planning to quickly follow up its flagship duo with the GeForce RTX 5070 and RTX 5070 Ti, with specs for both of these cards painting a clearer picture of how the Blackwell generation is going to stack up against the current Ada Lovelace GeForce RTX 40 Series.

Looking at the specs of the GeForce RTX 5070, which will feature a CUDA Core count of 6400, 12GB of GDDR7 memory, and a TGP of 250W, the mid-range GPU only features an 8.7% increase in its core count compared to the GeForce RTX 4070, with 10.7% fewer cores than the GeForce RTX 4070 SUPER refresh.

NVIDIA is expected to unveil its next-generation GeForce RTX 5090 and RTX 5080 graphics cards at CES 2025, officially launching and being excitedly installed into their desktop gaming PCs in late January 2025.

In a new video from leaker Moore's Law is Dead, we're learning that an NVIDIA retail partner source tell him that they were told that the unveiling "will be at CES" with a launch "likely to follow weeks (not months) later". Another source corroborated that, saying that the RTX 5090 and RTX 5080 would be announced at CES 2025.

The biggest takeaway for me here is that the second source told MLID that RTX 5090 pricing "might be priced on the lower end of the leaked spectrum". NVIDIA's ultra-enthusiast GeForce RTX 5090 in previous price leaks at somewhere between $1999 and $2499, so the "lower end of the spectrum" there suggests $1999 (or so) pricing.