Graphics Cards - Page 96

Here's a thing that we probably all do when it comes to testing out a brand-new GPU - briefly touch the backplate to see how hot it is to the touch. And here's the thing, GPU backplates are designed to get hot; they cover the back of the PCB, and for larger GPUs, they offer heat dissipation and structural support.

Still, you can see the logic behind these strange backplate-cooling solutions sold on AliExpress. If it's PC hardware-related and gets hot, try to cool it down with a few fans. And so, overclocker and tech reviewer der8auer has tested some of these new backplate coolers to see what they do.

As you can probably guess from the headline, the end result is little more than adding extra weight and physical size to a GPU. With AliExpress being home to several tech brands we've never heard of, der8auer begins the testing by taking a closer look at one of the coolers, and immediately the physical makeup is a little cheap. Sharp edges, glue for the fan mounting, zip ties, and adhesive tape for attaching it to a GIGABYTE Radeon RX 6800 XT.

Picking up a new GPU on the second-hand market is not uncommon, even for a product as fresh as AMD's Radeon RX 7900 XTX. That said, this is a strange one, as someone on Reddit posted an image of a second-hand Radeon RX 7900 XTX they purchased with an 'Internal Testing Only' label.

"Is there anything I should be aware of?" Redditor 'clumsyhotdog' wrote in the post. Many immediately pointed out that this could be an early version of the GPU that could feature some hardware differences from the retail model that shipped in December 2022. The user notes that it was purchased with a sizable discount compared to the MSRP of the GPU.

Pre-release models for GPUs are part of the process, with extensive internal testing to ensure drivers and other features are ready for launch. A pre-release Radeon RX 7900 XTX model could require special drivers and have features not present in the final retail release.

With strict export rules imposed by the United States government in 2021 prohibiting the export of US-based technologies (limiting the total performance of supercomputer hardware to China), we now have some specs and details for NVIDIA's mysterious A800 GPU developed for these constraints.

According to a new article at MyDrivers (via Tom's Hardware), the A800 operates at roughly 70% of the speed of NVIDIA's A100 GPU - which complies with the processing power limits in place for exports.

The NVIDIA A100 Tensor Core GPU is three years old, though still powerful in its own right, with 9.7 TFLOPS of FP64 and 19.5 TFLOPS of FP32 performance for the Ampere-based GPU. There's also the BFLOAT16 and FP16 Tensor Core performance of 624 TFLOPS with sparsity for AI workloads.

If you still weren't convinced that a new RDNA 3 graphics card is about to pop onto the scene, here's another sighting of RX 7600 models, this time from ASRock.

As highlighted by hardware leaker Harukaze5719 on Twitter, ASRock has just filed a trio of RX 7600 graphics cards with the EEC (Eurasian Economic Commission).

Typically, such regulatory filings will be put in place just ahead of the launch of a product, but they don't necessarily mean that launch will happen - sometimes, manufacturers make filings of a speculative nature, or planned releases fall through.

There's a consensus that GPUs like the GeForce RTX 3070, or anything with 8GB of VRAM, are entering an "end of life" phase when it comes to playing newly released PC games at 1080p and 1440p with detail settings cranked up.

A few weeks ago, we reported on a new video from YouTuber Paulo Gomes that saw him successfully modding an RTX 3070, upping the VRAM to 16GB - and it worked. For VRAM-intensive titles like the recent Resident Evil 4 Remake, it increased the 1% lows by a whopping factor of 9X, showcasing that VRAM capacity can make a difference.

Today comes a new video from modder Paulo Gomes detailing how he upgraded the memory of AMD Power Color Radeon RX 5600 XT Red Dragon - taking its 6GB capacity up to 12GB. Not only that, but the memory was replaced with faster 16Gbps chips.

The Hong Kong authorities have stopped smugglers from transporting live lobsters into China, with the strange addition that the vehicle was also smuggling a bunch of old NVIDIA Quadro K2200 series based on Maxwell architecture.

Nearly decades-old hardware at this point, with outlet The Register noting that these GPUs can be picked up for as little as USD 160. Still, Hong Kong Customs has valued the 280 kilograms of live lobsters and 70 "high-value computer display cards" at HKD 600,000 (around USD 76,500).

Lobsters and GPUs, it sounds like the destination might have been the world's strangest seafood restaurant.

Although the GeForce RTX 4080 is a powerful 4K-capable graphics card that presents a sizable uplift over the previous generation's GeForce RTX 3080, the MSRP of USD 1199 also presents a sizable uplift when it comes to pricing.

This has been the primary source of criticism for the GeForce RTX 4080 as a value proposition. Even though we're seeing price increases across the entire industry due to economic factors, there's a sense that the GeForce RTX 4080 shouldn't have crossed the USD 1000 barrier.

That said, when the GeForce RTX 3080 launched with an MSRP of USD 699, the mining boom and pandemic demand for all things tech pushed prices for the GPU into the MSRP region of the GeForce RTX 4080 - so there's that. The good news is that most of the new GPUs from NVIDIA and AMD are being sold at prices close to the MSRP, a welcome change from the Ampere generation. And now we've got a GeForce RTX 4080 from PNY that's being sold for almost USD 100 less than the MSRP.

NVIDIA could soon be going head-to-head with AMD as both firms unleash mid-range GPUs at the end of the month, with previous rumors to this effect now backed up by another piece of evidence.

As VideoCardz spotted, one of the better-known hardware leakers on Twitter, Harukaze5719, was eagle-eyed enough to spot GIGABYTE filings with the EEC (Eurasian Economic Commission) for two GPUs - the NVIDIA RTX 4060 Ti and AMD RX 7600.

There's a whole load of entries for the RTX 4060 Ti, and a couple for the RX 7600, with all of these graphics cards shown with 8GB of VRAM as was previously rumored (multiple times).

NVIDIA's RTX 4060 Ti graphics card has been spotted listed at an online retailer, with the Russian outfit spilling a couple of spec details that back up what we've already heard on the grapevine.

As VideoCardz spotted, well-known leaker @momomo_us flagged this up on Twitter, with the Russian retailer listing four Palit models of the RTX 4060 Ti. Those graphics cards show the much-rumored (and maligned) VRAM loadout of 8GB (GDDR6) and that the 4060 Ti will have a 128-bit memory bus.

Naturally, we should still be very cautious about accepting this as the truth about the spec of the RTX 4060 Ti, but as mentioned, it lines up with the previous chatter from the GPU grapevine.

NVIDIA's RTX 3070 has finally sunk below its MSRP in the US and Europe, and those waiting for a more affordable Lovelace GPU - with the RTX 4060 Ti supposedly now on the horizon - might just be tempted to instead pull the trigger on this older mid-range Ampere graphics card.

Would that be a wise decision? We'll come back to that, but first, let's look at the recent price drops NVIDIA's RTX 3070 has witnessed.

As German tech site 3DCenter reports (via VideoCardz), over in Germany at retailer Caseking, there's a Zotac RTX 3070 for €449, which is €50 under the recommended price of €499.

Popular streaming tool OBS Studio just got a new release with version 29.1, bringing AV1 streaming to YouTube over Enhanced RTMP. As per NVIDIA's latest post, this is perfect for the GeForce RTX 40 Series line-up that features real-time AV1 hardware encoding as standard across desktop and laptop variants.

NVIDIA notes that not only does AV1 offer better image quality using the same bitrate as H.264 encoding on previous generation GeForce RTX GPUs, but the new GeForce RTX 40 Series is also 40% more efficient, meaning that you can stream at 4K using roughly half the bitrate to get the same result as using H.264.

NVENC, short for NVIDIA Encoder, puts all hardware encoding onto the GPU versus the CPU. With AV1 designed and becoming the standard for streaming, NVIDIA's latest graphics cards are ideal for content creators on platforms like YouTube and Twitch.

The GeForce RTX 4070 is the latest addition to the 40 Series line-up, representing the first affordable option ahead of the GeForce RTX 4060 Ti and RTX 4060 debuts in the coming months.

Per our reviews for the Founders Edition and MSI Ventus 3X models, we were impressed by the small and lightweight two-slot design - at least compared to beefier GPUs like the GeForce RTX 4080 and RTX 4090.

This makes the launch of the new Colorful iGame GeForce RTX 4070 Neptune OC-V, an all-in-one water-cooled RTX 4070 with a 240mm radiator, a little odd. The GeForce RTX 4070 in all four versions we tested was an exceptionally efficient GPU that rarely hit its 200W TGP, with fan speeds below 40%. And with that, temperatures were still the lowest for the GeForce RTX 40 Series.

Xbox's first-party title Redfall is available now on Xbox Series X|S and PC, and as per our report earlier today, the overall reception has been pretty negative. With performance on PC and Xbox being highlighted by several reviewers as a significant issue, GeForce RTX 40 Series owners might be in the clear thanks to DLSS 3.

NVIDIA has just released GeForce Game Ready Driver 531.79, which adds support for Redfall, including DLSS Super Resolution, DLSS Frame Generation, and NVIDIA Reflex. According to NVIDIA's benchmarks, with the help of DLSS 3, Redfall performance across the entire GeForce RTX line-up hits over 100 frames-per-second in 4K. And with the aid of Reflex, latency is kept low.

The flagship GeForce RTX 4090 sees performance of up to 188 fps at 4K, an increase of over 71% compared to playing the game with DLSS disabled. The GeForce RTX 4080 managed up to 148 fps with DLSS 3, followed by the GeForce RTX 4070 Ti with up to 119 fps and the GeForce RTX 4070 with up to 104 fps.

The RTX 4090 Founders Edition graphics card has been spotted being sold by NVIDIA in Europe at almost 10% below the company's recommended pricing now.

VideoCardz reports that this is the case in Germany, and that the price of the RTX 4090 has been dropping below the MSRP of €1949 since February 2023.

To be precise, February witnessed a drop in price of 4.6% in Germany, and then in March, we saw the Founders Edition fall further to 6.7% under the MSRP.

Released in February of 1998, the Voodoo2 from 3dfx was an iconic moment for PC gaming - a discrete graphics card capable of delivering impressive 3D performance at a time when that was still a relatively new thing.

Of course, by 2023 standards, the Voodoo2 is positively archaic and a relic you'd have on display or as part of a dedicated retro gaming rig running something like Windows 95. In case you need a refresher, the maximum supported resolution for the card is 800x600, and the 8MB or 12MB of memory was primarily used to render rudimentary textures.

Still, this 3D printable 3dfx Voodoo2 Armor from creator Jeff Chen is a fantastic option for those remaining Voodoo2 owners as it adds a dose of modern GPU style to the retro graphics card doubling as a bit of protection.

The latest Steam Hardware and Software Survey results for April 2023 are here, and as per the latest figures, the NVIDIA GeForce GTX 1650 is currently the most popular GPU for Steam users. This card from NVIDIA dates back to 2019, released as a budget gaming option built on a 12 nm process featuring a modest 4GB of VRAM.

According to the latest data, the second most popular GPU is the NVIDIA GeForce GTX 1060, a GPU that launched with 6GB of VRAM. The top ten is basically a who's who of mainstream NVIDIA GPUs, including the GeForce RTX 3060 in desktop and laptop variants, the GeForce RTX 2060, and even the more powerful NVIDIA GeForce RTX 3070 and RTX 3060 Ti.

With all of the talk about VRAM capacity of late, it's interesting to note that pretty much all of the most popular GPUs, according to Valve's data, feature 8GB of VRAM or less - though, of course, these are mostly GPUs where you're running games at a lower resolution with lower visual settings.

As is the case with upcoming GPU launches from NVIDIA, we've had a lot of information and details for the upcoming GeForce RTX 4060 Ti appear from multiple sources. The latest (via Videocardz.com) reaffirms a late May launch and release for the mid-range GeForce RTX 40 Series graphics card, with an embargoed NVIDIA GeForce RTX 4060 Series document outlining the current plans and communication being sent out to its partners.

The document notes that cards will ship to partners on May 5 with the "Product Announcement" and "On-Shelf" listed as "End of May" with a specific date still being finalized. Interestingly the same document simply puts "TBD" under the column for the baseline GeForce RTX 4060, which would put the launch of that affordable mainstream option as no sooner than June 2023.

Another interesting takeaway from the document is that the board for the GeForce RTX 4060 Ti is listed as the PG190, the same as the GeForce RTX 3060 Ti. This means that partners can utilize the same coolers, potentially leading to a budget-conscious price point for the first time this generation.

NVIDIA's theoretically incoming RTX 4060 Ti has been spotted again, and this time, we have a photo of the GPU shared on Twitter.

Take a look at the chip in the pic above - clearly marked as the AD-106-350-A1 die, as previously rumored to power the RTX 4060 Ti - which comes from a regular leaker on Twitter, MEGAsizeGPU.

We've just had a sighting of the RTX 4060 Ti in MSI gaming PCs at a Dutch online retailer, too, as you may have caught yesterday.

NVIDIA's RTX 4060 Ti has been spotted in some MSI gaming systems, adding fuel to the fire that this GPU is imminent - and also again showing that, as previously rumored, it'll have 8GB of VRAM.

The MSI MAG Infinite S3 PCs are shown at a Dutch online retailer - presumably by mistake, of course, though the product listings are still available as we write this - and show the RTX 4060 Ti as the graphics card inside the rig. (The PCs are marked as not in stock, of course).

VideoCardz, which spotted this, observes that the Infinite S3 range is a middling offering and typically focuses on xx60 or xx70 Series GPUs from NVIDIA (indeed, new S3 models are just out with the RTX 4070 on board).

Taiwanese hardware manufacturer Sparkle is set to return to the discrete GPU market with new Intel Arc series graphics cards. Sparkle, which used to be an NVIDIA partner back in the day, stopped making GPUs in 2013 after the release of the GeForce GTX 700 series.

There's no word on why Sparkle is returning to the GPU market, but it's always great to see more companies release more models - especially for the relatively new Intel Arc series. Sparkle is releasing three Intel Arc graphics cards, with two Intel Arc A750 models and one Intel Arc A380 model - with Titan, Orc, and Elf naming.

Visually the cards follow the minimal look of the Intel Arc range, though the blue shroud does look really cool. The flagship Intel Arc A750 TITAN OC (it doesn't look like there's an A770 model just yet) features custom cooling and an LED light bar on the card that changes color based on the current temperature.