Graphics Cards - Page 78

It seems there are more and more regular NVIDIA GeForce RTX 4090 graphics cards being disassembled, tweaked, and put back together as an AI accelerator and sold for huge profits in China... where US sanctions have destroyed the AI market in the country, which is why companies are getting around those issues.

I was tagged on Twitter overnight by I_Leak_VN, who had some photos of more GeForce RTX 4090 graphics cards being turned into AI GPUs and sold in China, where you can see more and more boxes -- and I'm sure these are just a handful of these photos -- filled with expensive, unsellable RTX 4090 cards. US sanctions by the Biden administration have all but dragged the AI GPU market to a halt in China, so companies stockpiled RTX 4090s and are now ripping them apart and making AI GPUs out of them.

This is also something thast's not new, with VideoCardz pointing out that blower-style RTX 4090s have been on the market since November 2022, so over a year now. NVIDIA could've changed its thoughts on the design of its flagship RTX 4090, lifting restrictions with AIB partners that stopped them from making blower-style coolers for the RTX 4090.

AMD will have its future-gen Radeon RX 8000 "RDNA 4" series GPUs in the future, with the new "GFX1200" and "GFX1201" as unnamed graphics processors that have been spotted inside of Linux patches for the LLVM project.

The next-gen AMD RDNA 4 GPU architecture isn't the only one getting some patches for LLVM, with Intel's upcoming Panther Lake and Clearwater Forest series also getting some Linux lovin'. The GFX levels let us know which GPU architecture is being used on individual processors and also show us how similar each GPU is, as well as what features to expect from it.

We know that RDNA 1 was GFX101x, RDNA 2 was GFX103x, RDNA 3 is GFX11xx, RDNA 3.5 is GFX115x, while the next-gen RDNA 4 GPUs will be GFX12xx. That's why we're seeing GFX1200 and GFX1201 so early, which should be the not-so-high-end Navi 44 and Navi 48 GPUs, according to leaker RedGamingTech, who also has two possible configurations for the GPUs:

NVIDIA GeForce RTX 4090 pricing has started to rise quite significantly, with prices bursting over $2000 in the US on both Newegg and Amazon. Check it out:

BestBuy doesn't even have any RTX 4090 graphics cards for sale, MicroCenter doesn't have much stock at all -- and seems to be limited to physical stores, not online -- while the cheapest RTX 4090 seems to be Newegg with the MSI RTX 4090 SUPRIM LIQUID X priced at $1999. That's actually not bad for a liquid-cooled RTX 4090. But, from there, you're looking at much higher pricing across the board from retailers that actually have RTX 4090s in stock.

Right now, something like the MSI RTX 4090 SUPRIM LIQUID X costs $2397 on Amazon, while the GIGABYTE RTX 4090 AERO OC costs $2469. If you're looking for the ASUS TUF Gaming RTX 4090 OC Edition, then that'll be $2619.

NVIDIA's flagship GeForce RTX 4090 is starting to rise in price in the US, while in China, the US sanctions have the GPU banned in the country... but what about the RTX 4090s already there? Well, they're being turned into AI accelerators and sold at a profit, of course.

We have news of some Chinese companies buying GeForce RTX series graphics cards and turning them into AI accelerators, given the huge limitations that the Biden administration and the US government have put on AI hardware being exported from the United States and into China. AMD, Intel, and NVIDIA are blocked from selling specific AI hardware to China, while the higher-end RTX 4090 on the gaming GPU side of things is also banned because of its high compute power.

An insider has recently posted on the Chinese Baidu forums, saying that specialized factories are being formed across China that were stockpiling the GeForce RTX 4090 shipments before the bans and now turning them into AI accelerators. We're talking about GeForce RTX 4090 graphics cards from the likes of Palit, ASUS and its ROG Strix cards, and GIGABYTE's custom GAMING OC variants.

Dell has issued an internal Sales Advisory document to its salespeople to not sell hardware with AMD's most powerful RDNA 3 graphics cards to customers in China and other countries outlined by the U.S. government's recent export ban. The idea is to stem the flow of AI hardware to certain countries, and for Dell, this now includes the Radeon RX 7900 XT, Radeon RX 7900 XTX, and workstation options like the Radeon PRO W7900, which is a little strange.

The Export ban covers raw GPU power, which has led to the most powerful gaming GPU being banned in China. The GeForce RTX 4090 no longer appears on NVIDIA's Chinese website because it's so powerful it enters the realm of high-end AI hardware like NVIDIA's H100. Even though the Radeon RX 7900 XTX is powerful, it's no GeForce RTX 4090 - and as far as we can tell, it shouldn't require a special export license and doesn't fall with the current export restrictions.

Is this Dell misinterpreting the new export rules or a case of them being extra cautious? The list also includes AMD's AI Instinct MI210, MI250, MI250X, and MI300-series datacenter GPUs, which makes sense, but the inclusion of the Radeon RX 7900 XT doesn't.

ASUS recently launched its GeForce RTX 4060 Ti DUAL SSD, which has an M.2 slot on the back, handling either an M.2 SSD or a flagship GeForce RTX 4090 as an external GPU, I guess... awesome to see, though!

The RTX 4060 Ti DUAL SSD graphics card uses only 8 of the 16 PCIe lanes it has connected to the motherboard, with the integrated M.2 slot on the back of the card adding more functionality to a mid-range GPU. ASUS has room for a full-sized M.2 2280 SSD, with the company making some changes to its card from prototype to retail on the RTX 4060 Ti DUAL SSD.

ASUS has teased the GPU pushing over 12GB/sec reads from the integrated M.2 slot, thanks to ASUS using an M.2 2280 PCIe 5.0 SSD slot on the back of the RTX 4060 Ti DUAL SSD.

If there were a collector's item from NVIDIA's current-gen Ada Lovelace GPU family, it would be the ASUS ROG Matrix GeForce RTX 4090 that was personally signed by NVIDIA CEO and founder Jensen Huang and sold in an auction for $16,000.

The super-custom ASUS ROG Matrix GeForce RTX 4090 was put up for auction by ASUS to celebrate the launch of the majestic-looking card that I'd never want to actually use, just display it somewhere, with proceeds going to the Make-a-Wish International charity, helping fulfill wishes for children undergoing medical treatment.

ASUS had its custom ROG Matrix GeForce RTX 4090 signed by NVIDIA CEO Jensen Huang, with overclocker Der8auer spending a hefty $16,000 USD on the card, winning the auction. What did he get? Well, not only does he get the ASUS ROG Matrix GeForce RTX 4090 graphics card that's personally signed by Jensen, but he also gets a ROG Matrix NFT... yeah, NFTs are still a thing it seems.

NVIDIA is supposedly set to discontinue the RTX 4080 and RTX 4070 Ti when its refreshed RTX Super graphics cards are launched, but the vanilla RTX 4070 will stay on shelves.

This is the latest in a flurry of gossip around the revamped Super models NVIDIA is seemingly planning to reveal at CES 2024, and how they might dictate the fate of the existing mid-to-upper-range Lovelace line-up.

This comes from the Board Channels forum over in China (via VideoCardz), the claim being that the RTX 4070 Super is well differentiated enough compared to the RTX 4070 so they can both be kept on sale. (The full-fat AD104 GPU expected to be used in the Super offers a lot more CUDA cores than the considerably cut-down chip in the existing 4070, over 20% more in fact).

The world's fastest gaming GPU is no longer officially available in China, thanks to new US government restrictions that restrict sales of powerful GPU hardware that can be used for AI processing. Based on the new export measures, the GeForce RTX 4090 - a GPU designed for PC gaming - falls under the 'banned' list because it's simply too powerful. Every other gaming GPU, including the GeForce RTX 4080 and Radeon RX 7900 XTX, is okay.

As spotted by Videocardz.com, the GeForce RTX 4090 is no longer listed on NVIDIA's website in China - effectively removing it from the official GeForce RTX 40 Series lineup in the region after the restrictions took effect on November 17. The RTX 4090 has even been removed from all official performance charts, so the GeForce RTX 4080 is the most powerful GeForce RTX 40 Series GPU in China.

That said, you can still pick up an RTX 4090 in the region. Several reports indicate that NVIDIA and its partners sent a surplus of RTX 4090 stock into the region ahead of the restrictions. This isn't great news for the rest of the world because GeForce RTX 4090 prices are at an all-time high.

If you're in the market for a new graphics card right now, you can enjoy some killer Black Friday sales on both NVIDIA GeForce RTX series and AMD Radeon RX series graphics cards.

The ASUS GeForce RTX 3060 12GB Phoenix V2 graphics card has been seen as low as $255 on Amazon, while on the Intel side of things, you've got Sparkle's A770 TITAN OC graphics card for $289. The GeForce RTX 3060 for $255 isn't bad at all, especially if you're playing games on a 1080p display and want to spend under $300. The Intel Arc GPU... I don't think they're ready for primetime just yet, but that's just me.

If you want to go a bit higher-end, you can get a good deal on the GeForce RTX 4060 Ti for around $330, which isn't too bad at all for 1080p, 1440p, and ultrawide gamers with high refresh rates (120Hz, 240Hz, etc). Alternatively, there's the XFX Radeon RX 6800 Speedster SWFT319 graphics card for $370.

NVIDIA's next-gen GeForce RTX 50 "Blackwell" series gaming GPUs will support the now current DisplayPort 2.1 standard, with NVIDIA quite late to the party considering AMD's current Radeon RX 7000 "RDNA 3" series GPUs have DP2.1 connectivity... and hell, even Intel's current-gen Arc GPU family supports DP2.1.

In some new tweets, leaker "kopite7kimi" replied to someone asking if the new RTX 50 series would support DisplayPort 2.1, to which he replied with a tick, indicating they would. In another tweet, the leaker was asked if he knew the process node that NVIDIA would be using TSMC 3nm.

AMD is currently using the DisplayPort 2.1 UHBR13.5 implementation, which has a maximum bandwidth of 54Gbit/s, but the full UHBR20 implementation of DP2.1 can drive up to a much higher 80Gbit/s of bandwidth. NVIDIA is so late to the party I'd be even more disappointed if they didn't use the higher UHBR20 implementation for DisplayPort 2.1 on the new RTX 50 "Blackwell" series GPUs. NVIDIA's new GeForce RTX 5090, GeForce RTX 5080, GeForce RTX 5070, and everything in between -- Ti, SUPER, Ultra, GX2 -- needs to have DisplayPort 2.1 connectivity this time around, seriously.

While you might be hopeful of scoring a good GPU deal now that the Black Friday price-cutting action is well underway, you can forget it with NVIDIA's Lovelace flagship in the US.

Prices of the RTX 4090 at US retailers are heading up, not being discounted, as a report on Wccftech outlines - backing up recent chatter on the grapevine that the cost of this graphics card was set to rise.

Looking at Newegg for example, there's one model of the RTX 4090 at $1,900, then the next cheapest is around $2,100, and the others on sale are $2,150 or upwards (at the time of writing). On Amazon US, you're also looking at over $2,000, and some other major retailers are simply out of stock (like Best Buy).

AMD is prepping to release a new flagship RDNA 3-powered laptop GPU, the Radeon RX 7900M. Designed for high-end gaming notebooks and laptops, it's set to go head-to-head against NVIDIA's GeForce RTX 40 Series laptop graphics cards. And based on some Geekbench 6 benchmarks (via @BenchLeaks), it could give the GeForce RTX 4090 laptop GPU a run for its money.

However, we're discussing theoretical performance and a single Vulkan API test where the Radeon RX 7900M outperforms the RTX 4090 laptop GPU by around 9%. When stacked up against the RTX 4090 laptop GPU in an OpenCL benchmark, it does fall behind by about 25%.

These benchmarks aren't always a good measure of in-game performance, but the numbers are still impressive. So, chalk this up as good news for those looking for a high-end gaming laptop with Radeon graphics.

NVIDIA has reportedly discontinued the GeForce RTX 4080 and GeForce RTX 4070 Ti graphics cards, ending production of the two Ada Lovelace GPU-powered cards... preparing for the impending GeForce RTX 4080 SUPER and RTX 4070 Ti SUPER graphics cards.

We are going to see NVIDIA unveil its new GeForce RTX 4080 SUPER, RTX 4070 Ti SUPER, and RTX 4070 graphics cards announced during NVIDIA's own "Special Address" at CES 2024 in January. NVIDIA will have staggered releases throughout the second half of January 2024 and into February, of what I'm sure will be in this order: RTX 4080 SUPER, RTX 4070 Ti SUPER, and RTX 4070 SUPER for maximum eyeballs on reviews.

The news is coming from Board Channels, a website that aggregates sources and information from board partners connected to AMD, Intel, and NVIDIA. NVIDIA has reportedly sent its RTX 4080 and RTX 4070 Ti graphics cards in the "discontinuation stage," with both of them stopping mass production, and now just the cards left in warehouses will be sold.

AMD Radeon RX 7900 XTX graphics card prices are starting to leap in China after US sanctions officially banned NVIDIA's faster GeForce RTX 4090 graphics card in China.

Chinese retailers are now shifting the price of the RX 7900 XTX price up, as it's the Team Red alternative to the RTX 4090... well, the closest one, at least. The Navi 31-based Radeon RX 7900 XTX also has 24GB of VRAM, just like the RTX 4090 with its 24GB (although the RX 7900 XT has 24GB of GDDR6 while the RTX 4090 has 24GB of faster GDDR6X memory).

AMD's flagship Radeon RX 7900 XTX can be found for around $950 on Amazon, while during the recent 11/11 deals in China, the Radeon RX 7900 XTX was priced at around 7000 RMB ($1000 USD or so) with prices expected to be up to 500 RMB higher, close to 10% higher. We should expect higher prices on higher-tier graphics cards like the RX 7900 XT, RTX 4070 Ti, and RTX 4080 in China as gamers across the country who now can't buy the RTX 4090 want something with performance at least close to it.

AMD is outmaneuvering NVIDIA to considerable effect in the battle of the current-gen graphics cards according to some fresh sales figures.

Okay, so we must treat them with some caution, but these are the sales numbers from a large German retailer, MindFactory, which are often referred to in trying to gauge the popularity of CPUs and GPUs.

TechEpiphany supplies the figures regularly on X (formerly Twitter), and you can see the number of graphics card units sold during week 45 of 2023 in the above tweet.

NVIDIA is ramping up its GeForce RTX 40 SUPER family of graphics cards hitting the desktop in January 2024, but now we're hearing that there will be NO new gaming laptops with mobile RTX 40 SUPER series GPUs.

Jarrod'sTech tweeted out that we're starting to see the first public confirmations of no new laptop GPUs in early 2024, with Tom over at Moore's Law is Dead confirming this. MLID replied, saying that he can confirm that no "SUPER" laptops are planned and that the next major NVIDIA laptop refresh will "be in 2025 around CES" and released as part of their next-gen GeForce RTX 50 "Blackwell" GPU series.

MLID says that this was confirmed to him "directly from contacts at NVIDIA." We should expect NVIDIA to unveil its next-gen GeForce RTX 50 series "Blackwell" GPU family sometime in Q1 2025, or even sooner at this rate. The GeForce RTX 4080 SUPER, GeForce RTX 4070 Ti SUPER, and GeForce RTX 4070 SUPER will be NVIDIA's only new RTX 40 series GPU releases in 2024, and then all roads will lead to Blackwell for desktop and mobile GPUs for NVIDIA.

AMD's monster Ryzen Threadripper PRO 7995WX processor, with its 96 cores and 192 threads of CPU power, has been overclocked on AIR cooling, breaking multiple Cinebench world records.

AMD did do some in-house overclocking of its new Ryzen Threadripper PRO 7995WX processor during its announcement, but now the 96-core, 192-thread HEDT CPU has been overclocked by "SAMPSON," the same in-house overclocker at AMD that pushed the original 7995WX scores to new records.

The team is back, with SAMPSON using an ASUS WS TRX50 SAGE motherboard and 4 x 64GB DDR5-6400 memory modules from G.SKILL for a total of 256GB of DDR5 memory sitting alongside the monster-core CPU. The Ryzen Threadripper 7995WX processor has 96 cores and 192 threads with a boost clock of 2.5GHz and a max boost clock of 5.15GHz. There's 384MB of cache and a huge 128 PCIe Gen5 lanes at your disposal. SAMPSON used the 7995WX in Cinebench R15, R20, and R23 benchmarks, setting new records across the board.

NVIDIA's next-gen graphics cards (presumably RTX 5000 models) are set to use GDDR7 video RAM and run with a 384-bit memory bus (in the case of the flagship).

That's the latest rumor as aired on X (formerly Twitter) by RedGamingTech (the YouTube leaker) and backed up by Kopite7kimi (a regular contributor to GPU speculation on X).

We can expect bandwidth in the ballpark of 1.5TB/s, RedGamingTech tells us, and Kopite7kimi sees the fact that their sources marry up on this assertion as a 'confirmation' of GDDR7.

NVIDIA DLSS has evolved steadily over the years, and now the tech (alongside other upscaling methods) has become a major part of modern PC gaming - from DLSS 2 Super Resolution to the new DLSS 3 Frame Generation for GeForce RTX 40 Series owners. With DLSS 3, DLAA, and NVIDIA Reflex added to Bethesda's Starfield, we've got word from Team Green that RTX technologies are headed to a few more PC games this week.

For sci-fi, city-builder, and strategy game fans, the Early Access launch of Desynced is getting an update on November 16 that adds DLSS 3 support to the game. Based on NVIDIA's benchmarks using DLSS 'Performance' Mode in 4K, performance improvements are in the region of 2.2X. On the GeForce RTX 4070, the performance increases from 55.5 FPS to 129 FPS; on the GeForce RTX 4080, it increases from 93.7 FPS to 195.9 FPS; and on the GeForce RTX 4090, it increases from 130 FPS to a whopping 257.1 FPS.

With Early Access releases presenting incomplete versions of games, tech like DLSS can improve performance across a wide range of GeForce RTX hardware alongside regular optimizations and improvements.