Graphics Cards - Page 84

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.

NVIDIA already has the flagship GPU with its GeForce RTX 4090 graphics card, but the GeForce RTX 4090 Ti was in production in some form -- I/O brackets, mammoth coolers -- and now the RTX 4090 Ti is here again.

The not-released GeForce RTX 4090 Ti graphics card prototype teased again features a huge quad-slot design and an interesting PCB orientation: a 90-degree angle into the motherboard. This is the first time we've seen an RTX 4090 Ti prototype with a PCIe interface, with previous RTX 4090 Ti prototypes only featuring the cooler or I/O bracket.

We've got some new shots from Reddit user "gamer2live," which could be the first complete shot of the RTX 4090 Ti prototype we've seen so far, with a riser and cooler installed. We don't know if the PCB is installed, but the PCIe port is there... now we need a shot of the PCB. The card itself has a gold-ish color to its cooler, and the TITAN RTX is handily there for some side-by-side comparison.

One-Netbook has just announced its new OneXGPU, a new external device with an AMD Radeon RX 7600M GPU, M.2 SSD slot, and more. Check it out:

The new OneXGPU features an AMD Radeon RX 7600M XT, which uses the Navi 33 GPU and 8GB of GDDR6 memory, a perfect GPU choice for handheld gaming devices like this.

There's also OCulink support, which is capable of delivering up to 63Gbps of bandwidth... much better than Thunderbolt 3, Thunderbolt 4, and USB4, which are all limited to just 40Gbps.

AMD has just introduced its new Radeon PRO W7700 workstation graphics card, powered by the Navi 32 GPU and 16GB of GDDR6; the new offering will cost $999.

Inside, the new Radeon PRO W7700 workstation GPU features 16GB of GDDR6 with ECC memory on a 256-bit memory bus with the GDDR6 clocked at 18Gbps, we'll see up to 576GB/sec of memory bandwidth. The Navi 32 GPU clocks at up to 2.3GHz and will have 48 Compute Units, which makes for 3072 Stream Processors that will deliver up to 28.27 TFLOPs of single-precision compute performance with a 190W TBP delivered by a single 8-pin PCIe power connector.

AMD uses a tighter dual-slot design for its Radeon PRO W7700 workstation GPU, measuring in at just 26.7cm, too.

AMD will be unveiling its new Instinct MI300X and MI300A accelerators at its upcoming "Advancing AI" event penciled in for December 6, with some truly awesome specs, including up to 192GB of HBM3 memory.

NVIDIA's just-announced H200 Hopper AI GPU features an increased 141GB of HBM3e memory from Micron, up from the 80GB HBM3 memory used on its industry-leading AI performance monster H100 AI GPU. AMD will trump that with a whopping 192GB of HBM3 memory, which should prove to be some interesting performance battles between NVIDIA and AMD, given that AI workloads require lots of fast on-board memory.

AMD will be using a chiplet design for its MI300X and MI300A accelerators, compared to the monolithic H100 and H200 Hopper AI GPUs, using advanced packaging technologies over at TSMC (Taiwan Semiconductor Manufacturing Company). AMD's new Instinct MI300X will feature a blend of 5nm and 6nm IPs with a huge 153 billion transistors.

NVIDIA's new H200 Hopper AI GPU was just announced -- while at the same time as teasing its next-gen B100 Blackwell GPU -- with the new H200 AI GPU coming in 2024 with the world's fastest HBM3e memory from Micron. NVIDIA also announced that its Grace Hopper Superchips would power the new Exaflop Jupiter supercomputer.

The new NVIDIA H200 GPUs feature Micron's latest HBM3e memory, with capacities of up to 141GB per GPU with up to 4.8TB/sec of memory bandwidth. This is 1.8x more memory capacity than the HBM3 memory on H100, and up to 1.4x more HBM memory bandwidth over H100. NVIDIA uses either 4x or 8 x H200 GPUs for its new HGX H200 servers, so you're looking at a huge 1.1TB of HBM3e memory with up to 4.8TB/sec of memory bandwidth. Impressive numbers, NVIDIA.

NVIDIA also promises that its new H200 AI GPUs will be compatible with its existing HGX H100 systems, which makes it easy for customers to upgrade their systems. NVIDIA has partners like ASUS, ASRock Rack, Dell, Eviden, GIGABYTE, Hewlett Packard Enterprise, Ingrasys, Lenovo, QCT, Wiwynn, Supermicro, and Wistron. NVIDIA will have updated solutions when the H200 GPUs are made available in Q2 2024.

NVIDIA's next-generation B100 Blackwell GPU has been teased, and this time it's not coming from a leaker... but from NVIDIA themselves, teasing 4x the AI inference performance over its already industry-leading H100 Hopper AI GPU.

At the recent SuperComputing 2023 Special Address, NVIDIA provided some slides that compare its A100 Ampere AI GPU against their H100 and H200 Hopper AI GPUs, against the next-gen B100 Blackwell AI GPU with some mind-blowing results. NVIDIA showed GPT-3 175B inference performance where 1x is the baseline with A100, leading an 11x increase with the H100, a larger 18x increase with the H200, which will be released in 2024, and then it looks like a gigantic leap with B100 in 2024.

I've edited the photo and used the 11x performance with H100, showing that NVIDIA is teasing at least 3x the performance in GPT-3 175B inference performance with its next-gen B100 Blackwell AI GPU.

Turning the impressive and beastly ROG Strix GeForce RTX 4090 OC Edition GPU into a skateboard might sound like the sort of flex you can't help but get angry about; the good news is that what we're seeing here is simply the spare shroud from a ROG Strix GeForce RTX 4090 installed into a water-cooled system.

This fact alone makes it a pretty fun mod, commenting on the size of modern enthusiast gaming GPUs. With its length of 357.6mm, the ROG Strix GeForce RTX 4090 is one of the largest cards on the market, but even so, it's still half the size of a standard skateboard, which feature lengths of roughly 700mm.

Reddit user Ashley Said What posted a short clip of them using the ROG Strix GeForce RTX 4090 shroud as a skateboard, immediately sparking a whole range of responses - both hilarious ("How many frames are you running at?" and "driver seems to be working fine") to outrage over the extravagance.

MSI's new GeForce RTX 4060 AERO ITX 8G packs all of the power of NVIDIA's mainstream Adal Lovelace GeForce RTX 40 Series GPU into a lightweight and tiny form factor. With dimensions of 167 x 111 x 43 mm and an overall weight of 455 grams, it's one of the smallest GeForce RTX 4060 cards available.

MSI says it will "fit into virtually any system and ensure a smooth gaming experience." With a 40% smaller and 50% lighter build than most GeForce RTX 4060 graphics card options, MSI hasn't stripped back any of the performance.

Available in two variants, the MSI GeForce RTX 4060 AERO ITX 8G features the GPU reference spec of a 2460 MHz boost clock (which you can overclock to 275 MHz via MSI Center), 8GB of GDDR6 memory on a 128-bit bus, and a super-efficient power rating of 115W.

ASUS launched one of the most beautiful graphics cards it's ever with its new ROG Matrix GeForce RTX 4090 24GB graphics card, which costs $3199 as it is... now it's launched in China, and it's selling for $8230.

The new ASUS ROG Matrix GeForce RTX 4090 has an MSRP of $3199, twice the cost of the $1599 reference pricing, meaning you could buy 2 x RTX 4090s for the cost of the ROG Matrix RTX 4090 on its own. But, when you look at the $8230 pricing in China, you could buy 5 x RTX 4090s at $1599 each for that cost. Or, you know, the rest of the PC: CPU, motherboard, RAM, etc as well as an RTX 4090 and a 4K 120Hz OLED display.

Anyway, the ASUS ROG Matrix GeForce RTX 4090 has launched in China with an official MSRP of 23,000 RMB ($3155 USD or so), while at JD, the card is selling for 59,999 RMB ($8230 USD or so).

That's not a clickbait headline... it's a warning to save you some money. If you're itching to upgrade the graphics card inside of your machine or about to buy a pre-built gaming PC... hold off. Why? NVIDIA is going to announce its new SUPER refresh in the first week of January 2024, with "super aggressive" pricing that makes the current RTX 40 series cards a hard buy right now.

NVIDIA's current GeForce RTX 4080, GeForce RTX 4070 Ti, and GeForce RTX 4070 should be avoided unless absolutely necessary, as the RTX 4080 SUPER, RTX 4070 Ti SUPER, and RTX 4070 SUPER will replace them all. The new RTX 40 SUPER series cards will be improvements in not just GPU + VRAM improvements but also pricing.

In his latest video, Tom from Moore's Law is Dead says that he's hearing NVIDIA could be "super aggressive" on pricing to better compete and outright kill some of AMD's current RDNA 3-powered Radeon RX 7000 series GPUs. They're not my words... that's Tom's sources at NVIDIA saying they want to outright kill the Radeon RX 7900 XT once and for all with the new RTX 4070 Ti SUPER 16GB graphics card.

GALAX is celebrating its 20th anniversary in a massive way, with multiple custom GeForce RTX 4090 graphics cards announced, with the introduction today of not just one but two new 20th Anniversary Edition graphics cards.

GALAX announced its GeForce RTX 4090 20th Anniversary Edition not too long ago now, but today brings the new GeForce RTX 4090 20th Anniversary "Carbon Fiber Edition" alongside a regular white commemorative model. The better looking of the two for me is the new RTX 4090 20th Anniversary Carbon Fiber Edition, which has a front panel covered with 'carbon fiber material', which is probably not carbon fiber.

Whatever it is, it looks nice (at least in the pictures) and has a sleek Founders Edition kinda look. In an RGB-blinged-out world, it's nice to see more conservative designs like this in a high-end RTX 4090 card. GALAX is using dual 10cm 20th-anniversary exclusive fans on the front of the card, while there's a smaller 9cm fan on the back.

Team Red has further tweaked the preview driver for its AMD Fluid Motion Frames (AFMF) technology.

VideoCardz spotted the changes to version 23.30.01.03 which comprise of a couple of important fixes relating to stability.

AMD notes that firstly it has improved driver stability when task switching on the PC, and secondly, it has cured a bunch of intermittent crashes that have been blighting the Adrenalin driver.

You wouldn't swap your Radeon RX 7800 XT for its predecessor - that'd be insane, of course - but believe it or not, there is a scenario where the RX 6800 XT beats out the RDNA 3 graphics card, namely in Linux.

Specifically, in the Ubuntu distro, which was the battleground on which Phoronix organized a shootout between AMD, Intel and NVIDIA GPUs.

This was a test of 1080p resolution gaming on Linux, and the GPU of those picked which topped the rankings was AMD's RX 6800 XT - beating the RX 7800 XT into second place. (The 7800 XT was the most powerful graphics card picked on the AMD side of the equation, which is fair enough for Full HD resolution benchmarks).

Micron has just teased its VRAM roadmap covering memory technology for GPUs through to 2028, including GDDR7 and HBM4E memory technologies.

Micron already has its GDDR6X memory running at up to 24Gbps (up from 21Gbps when it was first introduced) with 16Gb dies at 96GB/sec. But in 2024, we're going to see new GDDR7 memory at a much faster 32Gbps bandwidth, and 16-24Gb dies at 128GB/sec, while faster GDDR7 memory will arrive in 2026, with 36Gbps bandwidth and 24Gb+ dies at 144GB/sec.

We should expect NVIDIA's next-gen GeForce RTX 50 series graphics cards to use Micron's new GDDR7 memory, with 32Gbps bandwidth at its disposal. The next-gen GeForce RTX 5090 graphics card should have a 512-bit memory bus with a whopping 1.5TB/sec of memory bandwidth thanks to the much faster GDDR7 memory.