Graphics Cards - Page 104

With crypto mining using off-the-shelf gaming GPUs now a thing of the past (or maybe it's just hibernating until the next boom), miners are now looking to offload graphics cards onto the used market. And really, if you're looking at buying a used GPU in 2023 - especially something from the Ampere or RDNA 2 generation - there's a good chance it might have been used to mine a cryptocurrency like Ethereum.

And with that stigma based on crypto mining pushing GPU memory and VRAM, it's not surprising that some mining cards are being sold as brand-new. A new investigation from YouTuber Iskandar Souza and computer technician Paulo Gomes looking at the Chinese market has discovered some new tricks people employ to make their GPUs appear brand-new. And it involves painting memory chips to make them look new because the easiest way to spot a used GPU comes from a yellow tint visible on the GPU and memory.

The video showcases the various degradation stages, which are made more prominent on mining cards based on the high temps and memory clock speeds. Resoldering valuable components onto new boards are common, too, even to salvage high-end materials to be re-used. Using paint to make a GPU look fresh is a new one - and you can see the paint being scraped off in the video to reveal the actual component color.

The latest GeForce Game Ready Driver has been released, GeForce Game Ready 528.24 WHQL, bringing DLSS 3 to HITMAN World of Assassination and Marvel's Midnight Suns. It also brings optimizations and day-one support for Dead Space, Deliver Us Mars, and Forspoken. All three include DLSS 2 support.

DLSS 3 includes NVIDIA's new Frame Generation technology alongside DLSS Super Resolution (DLSS 2) and NVIDIA Reflex, with Frame Generation being exclusive to the new GeForce RTX 40 series of GPUs. Using the beefed-up AI hardware inside the new Ada cards, it leverages AI to generate entire frames to bring a massive boost to performance - especially in cases where there are CPU limitations or bottlenecks. In the case of Marvel's Midnight Suns, it will bring more frames to those with that game's ray-traced reflections and ray-traced ambient occlusion enabled.

DLSS 2's addition to Dead Space is great to see and is a remake of one the greatest horror games of all time it's Frostbite-engine-enhanced visuals make it one of the biggest releases this month.

Google trends can often indicate popularity, especially regarding product interest. In a new study by Raul Bilc over at Razzem, compiling 62 graphics cards from NVIDIA, AMD, and Intel, we get a breakdown of Google searches for currently available GPU models. The results are surprising, and not that most searches were for NVIDIA products but that 78% of total searches were for current NVIDIA GPUs.

AMD followed this with 22% of overall searches, and Intel trailed with only 1%.

The winner was the NVIDIA GeForce RTX 3060, which racked up almost 9% of the total. This is in line with the mainstream Ampere card's popularity on Steam, as the most recent hardware survey results list it as the most popular GPU amongst PC gamers. This was followed by the GeForce RTX 3050, NVIDIA's most affordable Ampere generation GPU.

Many people were expecting AMD's flagship Radeon RX 7900 XTX to compete or trade blows with NVIDIA's GeForce RTX 4090 when it came to in-game performance that didn't include ray-tracing. As per AMD's pre-release unveiling and launch day reviews, its performance was closer to NVIDIA's GeForce RTX 4080 than the RTX 4090.

Not only that, but its power draw and efficiency were a step behind the competition and not in line with claims of a 50% increase in performance-per-watt compared to RDNA 2 and its flagship GPU the Radeon RX 6900 XT.

And with that, many began speculating that somewhere along the way, something went wrong. AMD's high-end RDNA 3-based GPUs feature groundbreaking multi-chip designs, which is the next significant leap forward. So teething problems aren't unusual. This is not to say that Radeon RX 7900 XTX performance isn't impressive; per our review of the reference model, it's a powerful triple-digit frame-rate 4K card.

The NVIDIA RTX 6000 Ada workstation GPU has been listed with a price point of USD 6800, representing a 46% increase over the previous Ampere-based NVIDIA RTX A6000. Dropping the 'A' in naming the new graphics card it's aimed at the workstation market and not enthusiast gamers.

Though one assumes you'll be able to use it for gaming even though it uses NVIDIA's Pro driver suite.

The NVIDIA RTX 6000 Ada specs are higher than the flagship consumer-focused GeForce RTX 4090, with its 18,176 CUDA cores presenting an 11% increase over the RTX 4090's 16,384 CUDA Core count. It also features more memory in the form of 48GB GDDR6 over a 384-bit interface - though it's clocked slower than the RTX 4090's 24GB of GDDR6X memory. The difference in specs points to some headroom for an eventual GeForce RTX 4090 Ti.

Recently we reported that a German retailer received a batch of older Radeon RX 6000s GPUs (RX 6800 and RX 6900) that had failed with cracked GPUs that were beyond repair. It was a concerning sample size at the time, with 48 GPUs out of 61 showcasing the same problem.

All cards were received over three weeks, and using the latest AMD graphics drivers from December 2022 was seen as a potential issue or starting point for the investigation. This was born from the fact that all users were using the cards differently, not simply pushing them to their limit when gaming.

It turns out it wasn't a driver issue after all, with KrisFix finding out that many of the cards were sold by a single seller where it's presumed they were being used for crypto mining. Getting a second-hand GPU that was previously used for mining is not uncommon, and when it comes to the health of a GPU in the long term, it's generally understood that mining wouldn't lead to the sort of failure rate seen here. That is, GPUs that are destroyed and cracked.

Recently it was reported that the GeForce RTX 4060 Ti would launch with a TGP power rating of only 160W - which would be 20% less than its Ampere counterpart. We also got a potential glimpse at the more mainstream Ada card's specs - the AD106-350-A1 chip, 4352 CUDA Cores, and 8GB of GDDR6 memory.

Even though being a 160W card would make it super efficient compared to the rest of the Ada line-up. It will still use the new PCIe Gen5 12VHPWR connector, which would point to NVIDIA, making this the norm.

The most significant point of contention with the specs is the news that the memory bus would only be 128-bit, with an effective bandwidth of 18 Gbps/288.0 GB/s. This brings into question the card's capabilities, alongside 8GB of memory, when playing games in 1440p and 4K. That said, the 32MB of reported cache represents a significant increase over the RTX 3060 Ti.

NVIDIA provided an engineering sample of the new GeForce RTX 4090 laptop GPU to Jarrod'sTech to compare it to the last gen laptop flagship, the GeForce RTX 3080 Ti laptop GPU. Although NVIDIA hasn't sponsored the video, it did provide both laptops, and the initial testing results are mainly limited to DLSS 3 performance in games. However, some impressive productivity results were also shown.

The RTX 4090 laptop featured in the video is the impressive Razer Blade 16 for 2023, which includes an Intel Core i9 13950HX CPU and 32GB of DDR5-5600 memory.

With DLSS 3 enabled, running five titles in 1440p - Portal with RTX, Microsoft Flight Simulator, A Plague Tale: Requiem, Warhammer 40,000: Darktide, and Cyberpunk 2077 - the performance increase compared to the GeForce RTX 3080 Ti laptop GPU is more than double. Going up to 2.6X in Portal and Flight Sim.

The NVIDIA GeForce RTX 4090 GPU is a beast of a GPU, the most powerful currently available. But all that performance comes in a sizable physical package, with both Founders Edition and AIB models having chunky 3.5 to 4-slot thicknesses. This can limit the number of cases and builds the card can comfortably slot into, not to mention requiring brackets and additional support depending on the orientation.

The existence of GeForce RTX 4090 cards using old-school blower fan designs for cooling is interesting for several reasons. It's a style of cooling that is no longer used because of the noise it can generate, not to mention that it is less efficient and capable of cooling a card when temperatures go high.

The silver lining is, of course, that the blower design means an overall smaller card, and that's one of the key reasons Redditor BottleneckEvader bought this custom USD 2675 variant of NVIDIA's flagship GeForce RTX 40 Series graphics card. It fits inside their Fractal Torrent Nano case because it's a 2-slot RTX 4090.

Depending on the workload, it looks like AMD's flagship RDNA 3 GPU, the Radeon RX 7900 XTX, can easily push beyond 3.0 GHz with clocks going as high as 3.5 GHz in Blender. These figures come from German outlet ComputerBase, using the Sapphire Radeon RX 7900 XTX Nitro+ model instead of AMD's reference design.

The report notes that the card wasn't entirely stable with these settings, but still, it's an impressive showing for AMD's flagship RDNA 3 card mainly because the official specs list a 2.5GHz boost clock frequency - a full 1 GHz lower than what was achieved here.

It's also worth noting that these results were achieved at 374W of power, which equates to a 13% boost in frequency at around 20% more power usage than the reference model. Sapphire's custom 7900 XTX features improved cooling compared to AMD's design and is considered one of the best partner cards.

One of the most significant controversies surrounding the launch of the new NVIDIA GeForce RTX 4090 was the switch to a new 12VHPWR connector and how in some cases, the cables were melting and burning out.

This sparked a series of investigations from NVIDIA, board partners, and even tech enthusiasts and the media. Ultimately, it was deemed that the new 12VHPWR cable standard can be fragile at the connection point on the GPU side if there are any bends and the cable itself isn't pushed in or firmly secured.

All cards in the NVIDIA GeForce RTX 40 series lineup come with a 12VHPWR connector, so the potential for them to burn up opened the door for alternative solutions. One came from CableMod, a company specializing in high-quality, durable cables. Its take on the 12VHPWR connector was one of the first to hit the scene, designed to resolve the clearance and bending issues with NVIDIA's cable. And as soon as the product went on sale, it immediately sold out.

NVIDIA's GeForce RTX 4060 Ti will reportedly launch with a TGP of only 160W, which would be 20% less than the Amper generation's GeForce RTX 306 Ti. Power efficiency has been one of the most impressive things of the new Ada Lovelace generation, so it's cool to see the power draw in the mid-range coming down.

The still-to-be-revealed GeForce RTX 4070 (the non-Ti model) will also feature a lower power rating than its Amper counterpart, with a TGP of only 200W. This news arrived courtesy of an established GPU leaker of pre-release information - @kopite7kimi.

We also got word that the GeForce RTX 4060 Ti will use the AD106-350-A1 chip, 4352 CUDA Cores, and 8GB of GDDR6 memory. It remains to be seen how well the RTX 4060 Ti performs against the RTX 3060 Ti. Still, as that GPU ended up being one of the most popular of the Ampere generation due to its impressive performance and reasonable price point, it wouldn't be unusual to expect a decent gen-on-gen performance uplift. Plus, all the benefits of the 40 Series, like DLSS 3.

It's only a matter of time before NVIDIA lifts the lid on the next Ada Lovelace desktop GPU, the GeForce RTX 4070, which is rumored to launch later this quarter. But this hasn't stopped the RTX 4070 logo from appearing on marketing materials from GALAX for some cool-looking EX Gamer White and SG White edition graphics cards.

Weirdly the official marketing materials showcase both logos on the same box, as captured by Videocardz.com, where we can see RTX 4070 Ti on the side of the box and RTX 4070 on the front.

Historically Ti and non-Ti models share similar specs and cooling configurations, so this could be a case of someone over at GALAX putting the wrong logo on the box. Mistakes from GALAX like this aren't uncommon either, as the company made the mistake of creating a banner for an RTX 4090 Ti HOF GPU that doesn't exist. Yet.

MSI is releasing some interesting new models for the GeForce RTX 4090 - the GeForce RTX 4090 SUPRIM X CLASSIC 24G and the GeForce RTX 4090 SUPRIM CLASSIC 24G. So then, what makes these cards different from the existing GeForce RTX 4090 SUPRIM models? Well, the 'Classic' here looks to be about using a previous-gen SUPRIM (GeForce RTX 3090 Ti) shroud and cooling.

This isn't bad, as you've got an impressive TRI FROZR 2S Thermal Design with TORX Fan 4.0, a copper baseplate, and decent airflow. Plus, dual BIOS support pushes NVIDIA's flagship Ada card to its limits - thanks to the 480W gaming mode in the SUPRIM X version. The non-X SUPRIM CLASSIC features a lower boost clock of 2535 MHz compared to 2640 MHz of the X.

We reviewed the current SUPRIM X RTX 4090 back when the GPU launched last year, alongside the cooling and build of the GeForce RTX 4090 SUPRIM X CLASSIC 24G in GeForce RTX 3090 Ti form.

The GeForce RTX 4080, like the RTX 4090, is a powerful GPU capable of delivering triple-digit 4K performance with ray-tracing and in-game detail settings cranked up. Like its beefier sibling, the main issue for many is the cost of entry, with pricing for the GeForce RTX 4080 starting at around USD 1200.

Reports show that the upcoming silicon for the still-to-be-announced GeForce RTX 4070 (the non-Ti version) uses two variants - AD104-250 and AD104-251. The latter is built without a "comparator circuit" and fewer components. And with that, we've now got word that the RTX 4080 is following suit. Cards with the brand-new AD103-301 GPU silicon are starting to ship - with Gainward being the first to do so.

The new chip and the current RTX 4080 chips feature the same specs and performance, with the changes and reduction in components looking to reduce the cost of producing a new GeForce RTX 4080. The TDP, frequency, and Cuda core count of 9728 remain the same.

GALAX is prepping to launch two variants of its Hall of Fame (HOF) GPU range covering the most recent addition to the GeForce RTX 40 Series line-up - the RTX 4070 Ti.

The HOF OC Lab PLUS will include a Performance mode with a boost clock speed of up to 2760 MHz, with the HOF OC Lab (no PLUS) going up to 2715 MHz. What's most interesting about the HOF range and these models is the increase in power draw, which can go up to 366W - a massive increase over the 285W found in stock models. They do so via three PCIe 8-pin power connectors and a single 16-pin power adapter. Most RTX 3070 Ti GPUs on the market only support two PCIe 8-pin connectors.

This means that out-of-the-box, they won't be the fasted clocked OC models for the RTX 3070 Ti. Instead, they represent versions of NVIDIA's GPU with the most overclocking potential. There's a special Hyper Boost button on the I/O bracket that will "'improve overclocking capabilities", which we assume increases the max power limit of the card.

AMD is currently dealing with the RMA issues surrounding faulty vapor chamber cooling in reference model Radeon RX 7900 XTX cards, so a report of older Radeon RX 6000s GPUs mysteriously failing is a little concerning. And bad timing.

Of course, the following is anecdotal and comes from a single outlet, KrisFix Germany, but the sample size is pretty significant. Several AMD Radeon RX 6000 series graphics cards (RX 6800 and RX 6900) have been sent to the shop for repair - 61 in total. And out of that batch, 48 have dead GPUs with the same shorted SOC rail and shorted memory rail. All cards were received over three weeks.

KrisFix also posted follow-up pictures to Facebook showcasing some seriously damaged cracked GPUs. He notes that receiving this many fault cards is unusual, especially with all of them showcasing the same DOA problems.

MSI is set to bring its first batch of Radeon RX 7000 series GPUs to market in January, with two models, the Radeon RX 7900 XTX Gaming Trio Classic and Radeon RX 7900 XT Gaming Trio Classic. In terms of cooling, these will not use the AMD reference designs that have recently been making the news due to vapor chamber issues but instead use MSI's recognizable Gaming Trio design.

Both cards use MSI's TRI FROZR 2 thermal system with TORX FAN 4.0 technology, enhanced PCB materials, and an anti-bend backplate. Both will ship with brackets for that extra dose of physical support, 325 x 141 x 56 mm dimensions, and 3 x 8-pin power requirements, and this puts their physical real estate in the 2.5-slot region.

In terms of clock speeds, there doesn't look like there will be any overclocking action out-of-the-box, with the RX 7900 XTX Gaming Trio Classic supporting up to 2500 MHz boost and 2300 MHz game. Though with 3 x 8-pin power connectors, there might be room to overclock manually.

The recent release of the GeForce RTX 4070 Ti happened without a Founders Edition from NVIDIA, on account of the card essentially being the uncancelled GeForce RTX 4080 12GB renamed. So the odds are that NVIDIA didn't have enough time to create new branded coolers and packaging.

According to a new leak showcasing the packaging from the upcoming still-to-be-announced GeForce RTX 4070, it looks like Founders Edition cards are back on the menu, with the RTX 4070s following the design of the RTX 4080 and RTX 4090 FE models.

The GeForce RTX 4070 is expected to be the next card in the GeForce RTX 40 Series to hit shelves, with a rumored Q2 2023 release window. As per another report, the GeForce RTX 4070 and the GeForce RTX 3060 Ti could end up using the same AD104 chip found in the GeForce RTX 4070 Ti - albeit cut down. Compared to the RTX 4070 Ti's 7680 CUDA Cores, the baseline RTX 4070 will potentially feature 5888 CUDA Cores

For many MSI Afterburner is the tool of choice when it comes to overclocking and tweaking their desktop GPU, and has been a mainstay for several years. Although it falls under the MSI banner it works on pretty much any GPU as it "gives you complete control" and "lets you monitor your hardware in real-time". And it's reliable to boot, offering easy-to-use overclocking tools to anyone; which makes the news that its future is uncertain, a concern.

MSI Afterburner developer Alexey Nicolaychuk aka "Unwinder", is based in Russia, and in a recent post over at the Guru3D forums he outlines that funding has been a major issue due to the recent conflict in the region, the ongoing war, and the current tech ban imposed on the country. This has made it just about impossible for MSI to pay Alexey for his work, to the point where he notes that the "MSI afterburner project is probably dead". The silver lining, and that's a pretty grim way to summarize the situation, is that MSI has provided a statement to Wccftech advising that it's still working hard to try and find a solution.

Alexey Nicolaychuk's comments point to funding as the issue, and that work carried out over the past year has been done out of pocket due to the political turmoil in the region. "War and politics are the reasons," Alexey writes. "I didn't mention it in [the] MSI Afterburner development news thread, but the project is semi-abandoned." Adding, "I have a feeling that I'm just beating a dead horse and wast[ing] energy on something that is no longer needed. Anyway, I'll try to continue supporting it myself while I have some free time but will probably need to drop it and switch to something else, allowing me to pay my bills."