Graphics Cards - Page 89

When Intel announced that it was discontinuing the production of Intel Arc A770 16GB Limited Edition GPUs, it left very few options for those after a version of the Arc flagship with full memory capacity. The good news is that AIBs are filling that gap, with the latest coming from ASRock in the form of a new Intel Arc A770 Phantom Gaming 16GB OC graphics card.

The new GPU already has a retail listing over on Newegg, with a price of USD 329.99 - making it one of the most affordable 16GB GPUs currently available. Once it becomes available, the retail listing showed up before any official confirmation from ASRock and simply notes that the item is on "backorder" with a ship date of July 28.

The good news is that the GPU finally got an official product page on ASRock's site - which means the card is real and out later this week. The key difference between this and the 8GB Intel Arc A770 model comes down to memory capacity and a slight increase in overall bandwidth - 16GB at 17.5 Gbps versus 8GB at 16 Gbps.

During a tour of ASUS HQ at Computex 2023, we got a look at several exciting new products, with one of the standouts being its new "hidden concept" GeForce RTX 4070 GPU that was all about removing visible power cables to deliver a clean and slick look. Seeing it in person, it was surprising what a difference no visible power cables make to the look of a GPU installed in a gaming rig.

According to a new report, this concept cable-less GPU is set to become a reality, with mass production of the new GeForce RTX 4070 in white on track for later this year. Removing the standard 8-pin or 16-pin power connectors means the GPU needs to get power from an alternate source, and for ASUS's new GeForce RTX 4070, a special motherboard is required.

This new GeForce RTX 4070 features an additional PCIe connector alongside the standard PCIe Gen 4.0 x16 interface, which delivers power via a 16-pin power connector on the underside of the motherboard - with ASUS noting that it behaves in the same way as including a power connector on the GPU, delivering up to 600W of power. To demonstrate the new GPU design, ASUS paired it with a custom Z790 TUF Gaming motherboard.

NVIDIA's RTX 4060 Ti graphics card with 8GB of VRAM could be facing stock shortages, but not due to the usual reason.

Typically, this would be because there's too much demand from gamers, but in this case, it's the other end of the supply-demand equation which is suffering - NVIDIA simply isn't making enough chips.

This is happening in China, according to Board Channels forums (as spotted by Wccftech), so as ever, keep a skeptical hat on for this source which is a little sketchier than we'd ideally like.

NVIDIA and OpenAI are reportedly aiming to create a technological marvel by combining the power of more than a million NVIDIA AI GPUs and software developed by OpenAI that will link them all together.

For those that don't know, AI systems such as open AI's ChatGPT are powered by thousands of NVIDIA AI GPUs. According to reports, NVIDIA has supplied approximately 20,000 of its AI GPUs to open AI, but this won't be enough if the company wishes to keep up with its ever-expanding language models that require more and more power.

According to Wang Shah Kwong, a businessman and founder of the Chinese search engine Sogou, OpenAI is already developing a more advanced AI computing model that will have a total capacity of 10 million AI GPUs. While 10 million NVIDIA AI GPUS sounds astronomical (and it is), it's another thing entirely to actually reach that number in physical hardware.

We've seen some pretty impressive overclocking action with the GeForce RTX 4090, with American overclocker Allen "Splave" Golibersuch being the first person to achieve a recorded 4 GHz boost clock speed with a ROG Matrix GeForce RTX 4090 from ASUS not that long ago. And now the OC master is back, smashing the 3DMark Fire Strike Extreme world record with the same GPU.

The ROG Matrix GeForce RTX 4090 was paired with an Intel Xeon w9 3495X 'Sapphire Rapids-WS' CPU, W790 motherboard, and G.Skill DDR5 memory. Alongside the GPU overclocked to 3885 MHz, the VRAM was overclocked to 1563 MHz to deliver 1.2 TB/s of memory bandwidth.

Of course, cooling all of that required liquid nitrogen. The result was a new world record 3DMark Fire Strike Extreme overall score of 56,489 and graphics score of 58,589. Per the HWBOT submission, it beats the record set by German overclocker CENS of 55,809 by 1.2%.

AMD has released a new graphics driver which ushers in some pretty jaw-dropping boosts for some PC games (although why this is the case - well, we'll come back to that later).

YouTuber Ancient Gameplays has been test driving the latest version, quite literally in the case of Forza Horizon 5, finding that AMD's version 23.7.1 Adrenalin release boosts the average frame rate by 67% in one case (compared to the previous driver release).

That's with the Radeon RX 7900 XT at 1080p resolution, and the 7900 XTX also witnesses a very healthy boost with the new driver to the tune of 32%. (All of this must be regarded with a little caution, as it's based off one round of testing on one particular PC configuration, of course).

The first 16GB RDNA 3 GPU from AMD for its Radeon RX 7000 Series won't be the upcoming Radeon RX 7800 rumored to launch in September. Instead, it will be the new Radeon RX 7900 GRE (Golden Rabbit Edition) for the Chinese market. According to a post from leaker @momomo_us, it's set to launch on July 28 and will be a cutdown version of the Radeon RX 7900 XT.

The Radeon RX 7900 GRE is a brand-new SKU for the line-up, with 84 Compute Units and 16GB VRAM across a 256-bit memory bus - information obtained from images showcasing the new Sapphire RX 7900 GRE NITRO+ model. As a cutdown model, the GPU will feature a slimmer build and could be some sort of Navi 31/32 hybrid.

AMD has yet to officially announce the Radeon RX 7900 GRE as a new option for Radeon GPUs, and so far, there's no indication that the GPU will be released globally, so that it could be something purely for the Chinese market. With no reference model, AMD could be testing the waters to gauge interest in a third Radeon RX 7900 variant.

NVIDIA's new GeForce RTX 4060 Ti 16GB GPU, which doubles the VRAM capacity over the existing 8GB model, has launched with very little fanfare. The main issue, relative to performance, is that the increase in memory capacity has seen an MSRP increase to USD 499 and EUR 549, compared to USD 399 and EUR 439 for the 8GB version.

And, like we've seen with other GPU launches in recent memory, the GeForce RTX 4060 Ti 16GB is currently being sold at a price lower than the MSRP set by NVIDIA. So far, there's a single example in Germany, with retailer Mindfactory offering the MSI GeForce RTX 4060 Ti 16GB Ventus 3X OC for €519, which is €30 lower than the MSRP - and even discounted from the launch price of €538.

The low price hasn't set the sales charts on fire, with Videocardz.com reporting that after 24 hours, the number of units sold was listed at 10. So then, what about the card's performance compared to the 8GB model when playing titles that could benefit from additional VRAM capacity?

Samsung has announced that it has completed development on the world's first GDDR7 ( Graphics Double Data Rate 7) DRAM for GPUs, with the new memory set to be installed in next-generation hardware from "key customers" for verification this year. This lines up with recent news that the first GDDR7-capable GPUs could hit the retail market next year.

The new GDDR7 memory spec can achieve speeds of up to 32Gbps. That's a 33 percent increase over the 24Gbps in Samsung's fastest GDDR6 memory kits. The flagship GeForce RTX 4090's GDDR6X memory tops at 21.2Gbps, 1TB/s of peak memory bandwidth. Samsung's GDDR7 memory is capable of delivering 1.5TB/s of bandwidth.

Samsung mentions PCs and game consoles in its press release; however, it also mentions high-performance computing and AI - an area that will drive the continued development of GPUs from NVIDIA, AMD, and Intel.

Earlier this week, we looked at leaked 3DMark benchmarks for AMD's upcoming new RDNA 3 GPU offerings - the Radeon RX 7700 and Radeon RX 7800 - that are on track to launch in September. Of course, all of this is rumor and unofficial, and AMD has not officially announced its plans for expanding the Radeon RX 7000 series with new models.

Currently, we have the mainstream Radeon RX 7600 in the sub-USD 300 category, with the flagship Radeon RX 7900 XT and XTX sitting in the high-end USD 899 and USD 999 price range. Everything in between is still being serviced by existing RDNA 2 products like the Radeon RX 6700 and Radeon RX 6800 - which are currently selling well thanks to sizable discounts.

Today we've got preliminary leaked pricing for the new Radeon RX 7700 and Radeon RX 7800 from Twitter user @All_The_Watts, who shared the 3DMark synthetic benchmark Time Spy results for these GPUs. And with the Radeon RX 7700 potentially launching with an MSRP of USD 449 and the Radeon RX 7800 launching with a price point of USD 549 - how does this pricing stack up when taking the benchmarks and other GPUs from AMD and NVIDIA into account?

Anything NVIDIA can do, ASRock can do just as sneakily it would seem - and better, to boot. At least when it comes to launching graphics cards that double up the VRAM loadout, with ASRock further showing that it doesn't have to cost the same premium that Team Green stuck on top.

You can't have missed the recent NVIDIA launch of the RTX 4060 Ti which has 16GB - compared to the original 8GB VRAM configuration - or actually, maybe you might have missed it. Seeing as Team Green made no announcement whatsoever, and didn't send out review cards, in what was the lowest of low-key releases.

And now, while ASRock hasn't actually announced a new version of its Intel Arc A770 graphics card which packs 16GB in a similar VRAM-doubling feat compared to the original A770 Phantom Gaming, we've seen the board.

It's been a long time coming, but we're getting more and more info from different sources pointing to a September launch for AMD's new mid-range RDNA 3 offerings - the Radeon RX 7700 and the Radeon RX 7800. Cards that will fill the gap between the mainstream Radeon RX 7600 and the flagship combo, which is the Radeon RX 7900 XT and 7900 XTX.

The latest comes from John Peddle Research, citing "various leaks from Asia," the cards will not only launch in September but will get an official unveiling from AMD at the upcoming Gamescom convention in Cologne, Germany, in late August. Using Gamescom as a place to announce new products is not uncommon, and we've seen NVIDIA use the venue in the past to present closer looks at its upcoming GPU releases.

Where it gets interesting is that John Peddle Research is claiming that AMD will launch "at least two" but maybe five new GPUs to fill out the RDNA 3 line-up now that the remaining stock for previous-gen RDNA 2-based Radeon RX 6000 Series graphics cards is being sold through globally.

MSI held a livestream that benchmarked the company's new MSI GeForce RTX 4060 Ti Gaming X 16GB Slim, comparing its performance to the 8GB non-slim Gaming X model on the same system across a handful of popular games at 1080p. And the results were the same across Cyberpunk 2077, running with Ultra settings to CS:GO to Hogwarts Legacy with DLSS and Frame Generation. Actually, marginally slower in most cases.

The first title benchmarked, Cyberpunk 2077, delivered 91.4 fps on the RTX 4060 Ti 8GB versus 90.6 fps on the RTX 4060 Ti 16GB. Not that it's a title that demands 16GB of VRAM to play in 1080p but the only benchmark where the new GeForce RTX 4060 Ti 16GB outperformed the GeForce RTX 4060 Ti 8GB model in MSI's 1080p benchmarks was CS:GO. In the popular competitive shooter, the frame rate increased from an already fast 419 fps up to 445 fps.

Ahead of the GeForce RTX 4060 Ti 16GB's launch, which sees the price increase from an MSRP of USD 399 to USD 499, NVIDIA's information on the card's relative performance was that, for the most part, it would be the same unless there's a very specific VRAM bottleneck. Or for those taking advantage of its creator-friendly capabilities, where more VRAM is always better.

The 16GB version of NVIDIA's RTX 4060 Ti graphics card is now on shelves, and the pricing of this new variant is pretty much as we feared - verging on the ridiculous, certainly in Europe at launch, and we don't imagine the US market will fare much better when more models emerge.

So, to Europe first, where VideoCardz spotted (via ComputerBase) that the initial asking price of the ASUS ROG STRIX RTX 4060 Ti 16GB is €659 at a German retailer (not MindFactory, but Nootebooksbillinger).

The interesting - and, well, rather ludicrous - reality is that you can purchase a baseline RTX 4070 model at the same retailer for less money. The asking price for the cheapest RTX 4070 is 10% less, in fact (well, almost - it's €599).

NVIDIA RTX IO was first revealed in 2020, during the Ampere-based GeForce RTX 30 Series era, with the new technology all about offloading texture decompression and other data tasks from the CPU onto the GPU by taking advantage of modern NVMe and PCIe Gen storage.

NVIDIA RTX IO uses the technology seen in Microsoft's DirectStorage, GDeflate, an open GPU compression standard that NVIDIA helped create. In most modern games, we see data sent from the disk drive to the CPU, which is then decompressed via system memory and sent to the GPU. With the massive gains we've seen in PCIe Gen 4 and Gen 5 storage, it's a process that doesn't take advantage of their speed and capabilities when it comes to PC gaming.

The result is longer load times and things like texture pop-in, even though storage speeds have doubled or tripled. This is where NVIDIA RTX IO comes in. "NVIDIA RTX IO dramatically increases IO bandwidth by letting compressed data be delivered to GPU memory with minimal staging in the CPU system memory," writes NVIDIA.

NVIDIA has quietly launched the new GeForce RTX 4060 Ti 16GB variant, which features the same hardware seen in the 8GB model but with the memory capacity doubled to 16GB. It seems that GeForce RTX 4060 Ti 16GB models have simply made their way to retail without being sent to media beforehand - so there are currently no reviews for the GPU online.

With the same underlying hardware and 128-bit memory interface, there won't be a massive difference in game performance outside of a few cases where the limitations are mainly due to running out of memory capacity. Titles like the Resident Evil 4 remake with max settings.

Interestingly, the 16GB model features the same TDP power rating, so doubling the VRAM capacity won't impact the card's impressive power efficiency. The big downside is that to get an extra 8GB of VRAM, the price for the GeForce RTX 4060 Ti jumps up from USD 399 for the 8GB model to USD 499 for 16GB. This just about puts the GPU into a brand-new pricing tier, so comparisons between it and other GPUs must be adjusted accordingly.

NVIDIA has pushed out a firmware fix for some of its older graphics cards that had a problem with DisplayPort 1.3 and 1.4 connections, a glitch that caused some nasty issues.

Affected GPU owners could find their PC locking up while booting and failing to reach the desktop. Or a lesser problem, but still annoying, was the user getting a black screen through the whole boot-up process until the desktop appeared.

NVIDIA explains that the new firmware update enables the latest DisplayPort 1.3 and 1.4 features, and without it, the mentioned problems could be encountered (although not necessarily).

With rumors swirling that AMD is planning to launch its mid-range RDNA 3 graphics cards in September, we'll be hearing much more about the planned Radeon RX 7700 and Radeon RX 7800 GPUs in the coming months. Starting today, with some supposedly leaked benchmarks for these GPUs running the popular 3DMark synthetic benchmark Time Spy coming from @All_The_Watts over on Twitter.

With the flagship Radeon RX 7900 XT and RX 7900 XTX out in the wild, which was recently joined by the mainstream Radeon RX 7600, a big gap in the line-up is ready to be filled. The consensus is that AMD has delayed the launch of its mid-range offerings for its latest generation of Radeon graphics due to existing RDNA 2 stock being available and sold at a discount.

But with stock levels dropping to the point where it's hard to find brand-new Radeon RX 6000 series cards in many regions, the time is right for AMD to launch new GPUs. And speaking of time, these Time Spy results showcase that the Radeon RX 7800 outperforms the GeForce RTX 4070 by around 5%, with the Radeon RX 7700 beating the GeForce RTX 4060 Ti by a more sizable 15%.

Want to know what's happening with AMD's RDNA 3 GPUs? If the rumor mill is right, the good news is that we should get an announcement of more than one addition to the RX 7000 range next month.

Moore's Law is Dead (MLID), a regular source of GPU leaks on YouTube, gave us the full details of where AMD is at with its remaining graphics cards for the RDNA 3 generation (add your own skepticism, though, as ever with rumors).

Gamescom 2023 in mid-August will supposedly see AMD announce the RX 7800 and 7700 - although the names of those GPUs could still change (even at this late stage).

A new report covering the DIY custom PC market in South Korea, with data gathered from Danawa Research (via @harukaze5719 on Twitter), shows the division of hardware covering GPUs and CPUs in the region for the first half of 2023. On the GPU front, it's a blowout, with GeForce GPUs taking 94% of the overall market share, followed by AMD Radeon at 5% and Intel Arc at 1%.

On the CPU side, it's a lot closer, with AMD gaining ground with a 43% share compared to Intel's 57% share. The numbers drill down from there, with the AM4 motherboard platform for AMD covering 71.3% of the Ryzen CPU market in South Korea, with the new AM5 platform and Ryzen 7000 Series covering 28.7% of the market.

With the Ryzen 7000 Series getting close to 30%, it's a strong showing for the new CPU series requiring a new motherboard and an upgrade to faster DDR5 memory. For Intel, the new 13th Gen series of CPUs (in particular, the 13th Gen Core i5) is the most popular, highlighting that PC gamers and DIY builders in South Korea like to keep their hardware up to date.