Graphics Cards - Page 98

Computex is back in a big way, with COMPUTEX 2023 set to take place at the Taipei Nangang Exhibition Center from May 30 to June 2, featuring over 1,000 exhibitors from 17 countries.

According to the event planners, this year's theme covers "high-performance computing, artificial intelligence application, next-gen connectivity, hyperreality, innovations and startups, and sustainability."

And with high-performance computing and AI in the mix and the return of in-person industry events back on the menu, the "COMPUTEX 2023 Keynote to Reveal Global Key Technological Innovations" is set to kick off the event and will feature NVIDIA CEO and founder Jensen Huang as its headliner.

NVIDIA's RTX 4070 is off the starting blocks, and there's still a good amount of stock of the GPU floating around after launch, with prices sitting at MSRP - and some graphics cards even being discounted already.

It certainly doesn't seem like RTX 4070 stock is flying off the shelves, or at any rate, supply is easily coping with demand.

And indeed there are suggestions that demand may be on the weak side, such as the tweet above from leaker TechEpiphany, who claims that big German retailer MindFactory (MF) only sold around 300 units on launch day, which is indicative that interest is "pretty low" in the GPU.

Intel's latest graphics driver is once again pushing ahead with performance for Arc GPUs, improving frame rates substantially in some cases for a number of popular games.

Note that the freshly released driver (version 31.0.101.4311) is still in beta, and it's for Arc Alchemist graphics cards plus Iris Xe GPUs.

The driver brings in support for a few new games - Boundary, Dead Island 2, Minecraft Legends, Total War: Warhammer III - Mirror of Madness - and takes some big strides forward with frame rates, including some double-digit gains.

With the launch of the GeForce RTX 4070 this week (check out our reviews for multiple models here), we've got a clearer picture of the performance and efficiency gains with NVIDIA's new Ada Lovelace architecture.

Built on a custom TSMC 4N process, it presents a significant leap forward over the GeForce RTX 30 Series Ampere generation and its use of Samsung 8N regarding overall efficiency. Of course, there are advances in the underlying hardware, with next-generation RT Cores and SER (Shader Execution Reordering) and new AI hardware developed for the new DLSS 3 Frame Generation technology.

Back to the efficiency of Ada, the focus of this article - even though the GeForce RTX 4090 is a 450W-rated GPU, it rarely hits this height and scales remarkably well when it comes to gaming at lower resolutions. Based on averages across 15 Games (for a detailed breakdown of the games tested and settings, check out our review of the GeForce RTX 4070 Founders Edition), the GeForce RTX 4090 not only delivers a 70.7% uplift in performance over the 350W-rated GeForce RTX 3090, but it manages that by only using 5.9% more power.

After leaking the other day, today comes the official word from AMD, unveiling its most powerful AMD Radeon PRO Graphics Cards for workstations and professionals built on the new RDNA 3 architecture.

The AMD Radeon PRO W7000 Series, covering the AMD Radeon PRO W7900 and AMD Radeon PRO W7800 graphics cards, are the first professional GPUs built using AMD's new chiplet design. And like the flagship consumer-focused AMD Radeon RX 7900 XTX, they also support the latest DisplayPort 2.1 standard - which has yet to be adopted by NVIDIA.

AMD promises significantly higher performance as the new workstation Radeon PRO graphics cards than the previous generation. On that front, the flagship AMD Radeon PRO W7900 features 96 Compute Units with 61 TFLOPS (FP32) performance that can deliver a 1.5X higher geoman performance on the SPECviewperf 2020 benchmark. It also sports 48GB of GDDR6 memory on a 384-bit interface with up to 864 GB/s of memory bandwidth.

Hefty graphics cards can start to perish under their own weight, with board sag breaking components, a new report underlines.

The sag - where the card is pulled down at the back with gravity, perhaps with quite some force when it comes to larger triple-slot boards equipped with giant coolers - can actually affect the VRAM modules, according to YouTuber KrisFix (who runs a German hardware repair shop).

As you can see in the above video, the repair expert shows a damaged NVIDIA RTX 2080 Ti where the memory modules, which are very close to the PCIe slot, suffer from the slight bending of the board.

A new blog post over at AMD titled "Building an Enthusiast PC" puts the majority of its focus on VRAM, especially when gaming in 1440p and 4K. With recent PC gaming community discussion around VRAM capacity and whether or not 12GB is enough for modern titles, and the launch of the GeForce RTX 4070 happening very soon - the timing of the piece is calculated.

But also timely as AMD highlights that when it comes to peak 4K memory usage in modern titles like Resident Evil 4, The Last of Us Part I, and Hogwarts Legacy with Ultra settings enabled - anything less than 16GB looks like a potential bottleneck. Of course, legitimate optimization concerns have been raised about a few of these titles, where more VRAM becomes a band-aid approach for PC games that still need development work.

Naturally, this post also shows that AMD's Radeon graphics cards have you covered when it comes to VRAM capacity. The 8GB VRAM of the Radeon RX 6600 Series is the 1080p enthusiast option, the 12GB Radeon RX 6700 Series is for 1440p, and the 16GB Radeon RX 6800 Series is for 4K.

NVIDIA's Blackwell graphics cards, expected to be GeForce RTX 5000 models, could be over twice as quick as current-gen Lovelace models - and we've heard a pile of other rumors about Team Green's next-gen GPUs to boot.

This comes courtesy of Wccftech, with the tech site keeping its ear to the ground and flagging up a bunch of info - approach all this very cautiously, naturally - from Twitter and two prominent YouTube leakers, Moore's Law is Dead (MLID) and RedGamingTech (RGT).

Bearing firmly in mind that it's still early stages for Blackwell, we hear from MLID that NVIDIA is likely to go with a monolithic GPU (single chip) for the next-gen flagship (RTX 5090). That idea has been floated in the past via the rumor mill, notably from renowned Twitter leaker Kopite7kimi, but there's a twist here - MLID doesn't completely rule out chiplets.

NVIDIA's RTX 4070 is rumored to launch later this week, and we've caught a glimpse of some pre-release pricing - and it doesn't make for pleasant reading.

We should make it clear upfront that prices from retailers that appear before a product is launched can often be wrong, and maybe just placeholders (educated guesses), but equally, they can sometimes serve as a fair indication of ballpark price tags.

Sadly, what we've witnessed from a Vietnamese retailer (Bpstore.vn) - as highlighted by I_Leak_VN on Twitter - is on the high side compared to previous leaks.

The AMD Radeon Pro W7900 Graphics Card has been spotted in Puget Bench, a benchmark tool for showcasing accurate After Effects performance. It points to an imminent or "coming soon" release for the first RDNA 3-powered workstation GPU from AMD that will compete with something like the NVIDIA RTX 6000 Ada graphics card from NVIDIA.

The graphics score of 135.3 for the AMD Radeon Pro W7900 puts it roughly 10% slower than the NVIDIA RTX 6000 Ada workstation GPU, but like the launch of the AMD Radeon RX 7900 XTX (which it's safe to assume this new GPU is based on), the assumption is that it will launch with a lower price.

The AMD Radeon Pro W7900 will be the successor to the AMD Radeon Pro W6000, targeting high-end workstation rigs. Although the Puget Bench listing confirms the existence of a next-generation workstation GPU from AMD, specifications and other details remain unknown.

Over at Videocardz.com, leaked images and info outlining performance and benchmarks for the GeForce RTX 4070 (presumably the Founders Edition) have appeared, confirming that overall performance without DLSS 3 and Frame Generation is on par with the GeForce RTX 3080.

The same slide also confirms that "Without Frame Gen" performance is 1.2X faster than the GeForce RTX 3070 Ti and 1.3X faster than the GeForce RTX 3070. According to the source, these official numbers from NVIDIA include DLSS 2.

The overall byline from NVIDIA for the new GeForce RTX 4070 is that the card will deliver 100+ fps performance at 1440p with ray-tracing and DLSS 3 enabled. It makes sense that official benchmarks and info from NVIDIA would focus on the brand-new GeForce RTX 40 Series exclusive DLSS 3 technology - it did the same when it came to the RTX 4090, RTX 4080, and RTX 4070 Ti.

AMD's flagship RDNA 3 GPU, the Radeon RX 7900 XTX, launched in December last year with an MSRP of USD 999. This price was for the reference model using the cooling and specs laid out by AMD, with custom partner cards arriving later with more room for overclocking and a higher price tag.

One of these cards is the ASRock Phantom Gaming Radeon RX 7900 XTX with 24GB GDDR6 memory and a higher Boost Clock of 2615 MHz using triple 8-pin power connectors. A step up from the two 8-pin power connectors of the reference models. This custom Radeon RX 7900 XTX from ASRock initially launched with a price of USD 1,119.99; its current price on online retailer Newegg is USD 959.99.

The Radeon RX 7900 XTX is a capable 4K GPU, delivering exceptional performance. Our review of the custom ASUS TUF Gaming Radeon RX 7900 XTX OC Edition produced consistently better results than the reference Radeon RX 7900 XTX. So it's safe to assume the story will be the same for the ASRock Phantom Gaming Radeon RX 7900 XTX.

The GeForce RTX 4070 launch is coming next week, with multiple models set to arrive, including a Founders Edition from NVIDIA - at least according to sources. The RTX 4070 hasn't been officially announced yet, but we're days away from seeing it in action.

The reported price point of USD 599 makes it the most affordable GeForce RTX 40 Series GPU. We've also got word that the 200W power rating will mean some models from partners will use a single standard 8-pin power connector instead of the 12VHPWR connectors used in all RTX 4090, RTX 4080, and RTX 4070 Ti GPUs.

Today comes some new info from established leaker @momomo_us that GIGABYTE plans to launch four GeForce RTX 4070 models, covering the company's Eagle, WindForce, Gaming, and Aero variants. And with a side on image showcasing all four RTX 4070 models, we can see that two will use a single 8-pin power connector while two will use the new 12VHPWR standard.

Although we've had plenty of reports outlining the USD 599 price-point for the GeForce RTX 4070, specs, and the launch happening next week - we haven't received an official announcement that the GPUs are coming from NVIDIA until now.

And the source is a bit strange, with the RTX 4070 making its first appearance in official NVIDIA benchmarks for the company's low-latency Reflex technology coming to Counter-Strike 2 (via The Verge). The long-awaited update to Counter-Strike will see the next-gen Valve Source engine-powered version of the game eventually take over from CS:GO.

And the benchmark is weird because it compares PC latency between three GPUs - the RTX 4070, RTX 3060, and the GTX 3060. Well, maybe not odd, as these older GPU options represent some of the most popular graphics cards according to the latest Steam Hardware Survey. And CS:GO is one of the most popular games on Steam, so the comparison here makes sense.

NVIDIA's RTX 4070 is rumored to be priced at $599 in the US, and part of the reasoning behind that lower-than-expected price tag pertains to the potential MSRPs of RTX 4060 models - which may also be pitched more affordably than we'd previously anticipated.

As you may have seen, YouTube leaker Moore's Law is Dead (MLID) has just published a new video exploring NVIDIA's apparent behind-the-scenes shenanigans with RTX 4070 pricing, and we cover that in full here.

In the clip, though, MLID brings up the issue of RTX 4060 pricing as it relates to the imminent RTX 4070 and the rest of the existing Lovelace range, drawing some interesting conclusions, all of which require a hefty dose of salt (throw in some pepper, too, we reckon).

NVIDIA's supposedly imminent RTX 4070 graphics card has been the subject of more leakage which reasserts that the previously rumored $599 price is correct - but there's a sting in the tail here.

The latest video from YouTuber Moore's Law is Dead (MLID) discusses the RTX 4070 in some depth, going into more detail on the pricing and also the potential performance levels of some of these graphics cards (which is the aforementioned sting).

At the end of last month, the grapevine was suddenly alive with chatter of a $599 recommended price (in the US) for the GeForce RTX 4070, and we were pretty surprised to see the MSRP theoretically pitched at that level. Particularly when previous rumors suggested NVIDIA could be looking at a much higher price tag, closer to the RTX 4070 Ti (indeed, $750 was one idea floated).

Yesterday we got an impressive video and breakdown of the new Cyberpunk 2077 Ray Tracing: Overdrive Mode, which brings full path tracing to the game on April 11 as part of a technical preview.

The "technical preview" wording there is key, as it takes the already technically demanding (and visually impressive) Cyberpunk 2077 and gives it a cinematic makeover thanks to full ray-traced lighting. From Global Illumination (GI) to shadows, full-resolution reflections, to every direct and indirect light source in the neon-soaked Night City given the RT treatment. Shader Execution Reordering (SER), Real-Time Denoisers (NRD), and RTX Direct Illumination (RTXDI) are all being leveraged here.

As per the video, in addition to a brand-new unified pipeline to make path tracing possible in Cyberpunk 2077, the Overdrive Mode also uses DLSS 3 and Frame Generation on the GeForce RTX 40 Series to make it more than a tech demo meets slideshow. And with that, native 4K rendering for the Ray Tracing: Overdrive Mode on the GeForce RTX 4090 drops to as low as 16 frames per second.

The Intel Arc range of discrete GPUs released by Intel may have had a rocky launch. Still, recent driver updates have now positioned the Intel Arc A750 as an affordable entry-level option for high-performance 1080p gaming. And with only two players in the GPU space until now (NVIDIA and AMD), a third means more competition, especially when it's a company as large as Intel.

And speaking of large, a new report at the Taiwan-based publication Commerical Times (via Tom's Hardware) notes that Intel is placing large orders for silicon for next-generation Intel Arc graphics cards at TSMC. The follow-up to the current Alchemist range is dubbed Battlemage, and these GPUs will use the TSMC 4nm process, with the next-next-gen Celestial architecture GPUs set to use the cutting-edge TSMC 3nm process.

Though, the planned launch for these GPUs will arrive later than expected, with Battlemage set to arrive in the second half of 2024 and Celestial in the second half of 2026.

Over at Videocardz.com, a leaked slide, presumably from a presentation for the new GeForce RTX 4070 GPU, confirms several things about the GPU. One of the most interesting bits of info comes under the 'Average Gaming Power' header that lists 186W as the power draw for the GeForce RTX 4070.

We can only assume this number comes from a specific set of games running at one or multiple resolutions. Still, it points to the latest GPU in the Ada Lovelace-powered GeForce RTX 40 Series being incredibly power efficient.

Although there's no performance data with this info, the slide features comparisons between the GeForce RTX 4070 and the RTX 3070 Ti and RTX 3070. And with that, the RTX 4070 is presumably a higher-performing GPU alongside being more efficient. For comparison, the 'Average Gaming Power' for the RTX 3070 Ti is 240W, and for the RTX 3070, it's 215W.

Based on the current state and feedback surrounding the PC port of the iconic PlayStation title The Last of Us Part I on PC, the game will need more than an "optimized" driver release to get it into shape.

Still, AMD Software Adrenalin Edition 22.40.43.05 has been released, offering game optimizations for The Last of Us for Radeon GPU owners. The driver release looks to have been pushed out early, with AMD noting in the release notes that a subsequent AMD Software: Adrenalin Edition 23.4.1 driver will be made available in the future.

And with that, optimizations for this one game look to be the extent of the latest Radeon driver update - but that makes sense. Currently, AMD is offering a free copy of The Last of Us Part I for new Radeon buyers, so it's unfortunate that the game's poor state on PC is outside of its control.