Video Cards & GPUs News - Page 1
Graphics card sales are very much flagging for current-gen products, if the latest gossip from the GPU grapevine is to be believed.
Moore's Law is Dead (MLID) provided the inside scoop from various sources in the retail industry - add your own seasoning as ever - as part of the YouTuber's latest video (which focuses primarily on analyzing performance levels of the RTX 4060, though that's not what we're interested in here).
One source at a major retailer outside of the US (presumably in Europe) that MLID spoke to said that NVIDIA's RTX 3060 and RTX 3050 were still selling well at under the $300 (US) mark (the RTX 3050 especially, for those models under $250). And that on AMD's side, the RX 6700 XT and 6600 XT are selling decently.
Intel has released a new beta driver for its Arc family of graphics card, and it boosts the frame rates of Assassin's Creed Unity (an admittedly very old game) by a colossal amount.
According to Intel's release notes for driver version 188.8.131.5214, Assassin's Creed Unity (which was out in 2014) runs a rather staggering 271% faster at 1080p (on very high graphics settings), and 313% faster at 1440p (on high settings). That's four times as fast - or rather, 'up to' four times, as Intel observes.
Why is there such a huge jump here? Well, Intel has made no secret that it's (obviously) a mammoth task to optimize Arc drivers from scratch for all PC games out there, so it has concentrated on modern titles naturally (and DX12 or Vulkan).
According to a new report at Seeking Alpha, the per-wafer cost of TSMC's upcoming 2nm process technology will be USD 24,570 when production begins in 2025. This represents a 25% increase over 3nm wafer costs, the current cutting-edge node at TSMC. Wafer costs for 3nm already increased by up to 40% over 5nm, so we're getting a clear picture that more advanced chipsets will also advance in cost.
With the price of consumer-focused enthusiast CPUs and GPUs for gaming and other tasks increasing in recent times, not to mention the unprecedented price increases for consoles like the PlayStation 5 and Xbox Series X, we could see the next few generations of gaming hardware cost more than we're used to.
When it comes to GPUs for PC gaming, which feature TSMC process technology, if AMD or NVIDIA choose to use 3nm or 2nm process nodes - then it could see the price of flagship GPUs from both companies dramatically increase. This is wild to think about when the GeForce RTX 4090's MSRP is already an eye-watering USD 1,599.
As part of an AI training demonstration showcasing the prowess of NVIDIA's H100 hardware, which saw a whopping 3,584 Hopper GPUs train a GPT-3 AI model in 11 minutes, the company also revealed a roadmap for its upcoming products - including what will eventually become the GeForce RTX 50 Series.
The roadmap refers to the technology as "Ada Lovelace Next" alongside "Hopper Next" and "Grace Next," so there are no clues as to what architecture will be called. It highlights a 2025 release, so we might have to wait longer than usual before seeing the GeForce RTX 5080 and GeForce RTX 5090.
Traditionally new GeForce hardware and architecture arrived every two years, give or take a few months. The Pascal-based GeForce GTX 10 Series arrived in 2016, the Turing-based GeForce RTX 20 Series in 2018, the Ampere-based GeForce RTX 30 Series in 2020, and the Ada Lovelace-powered GeForce RTX 40 Series in 2022.
Last week MSI announced its line-up of GeForce RTX 4060 graphics cards, with models set to launch this week in the company's well-known Ventux and Gaming product lines. Today, over at Newegg, we've got pricing for three models with retail pages live ahead of launch. These are currently listed as "Out of Stock," though that will change on June 29.
And with three models listed, we have confirmation that pricing for the GeForce RTX 4060 starts at USD 299, with the more premium variants costing USD 329. At least for MSI, we expect other brands to release higher-end variants to target the USD 329 price point.
For MSI, its MSRP model is the MSI GeForce RTX 4060 VENTUS 2X BLACK 8G for USD 299, and the specs listed detail that it features 3072 CUDA Cores and a Boost Clock of 2490 MHz. VENTUS cards from MSI represent the entry-level GPUs from the company and, as we've seen from previous releases, deliver great results with adequate cooling and silent performance. These cards are also smaller than more premium models, and the MSI GeForce RTX 4060 VENTUS 2X BLACK 8G's length of 199 mm is compact.
Ahead of the GeForce RTX 4060 launch this week, we've already seen some early previews showcasing DLSS performance in Cyberpunk 2077 with ray-tracing enabled using the game's hardware-intensive 'RT Ultra' mode.
In addition to some benchmarks we reported on earlier today, veteran tech YouTube channel JayzTwoCents got his hands-on the new ASUS RTX 4060 Dual model to see how it performs with a little tweaking and overclocking.
And the results aren't exactly earth-shattering, even with a noted higher TDP on the ASUS card of up to 130W and a clock speed boost of 100 MHz.In Cyberpunk 2077, with DLSS and the RT Ultra preset selected, we see the frame-rate increase from 80.8 fps to 82.3 fps, which is the sort of overclocking you'd consider negligible at 1.2%.
After a few benchmark leaks covering some synthetic tests painting a picture of the GeForce RTX 4060 presenting a 20% raw performance uplift over the GeForce RTX 3060, it looks like we've now got some official (or at least NVIDIA-approved) looks at how the RTX 4060 performs in-game.
The game in question is Cyberpunk 2077, still one of the best-looking PC games available, and back in the headlines thanks to the upcoming September release of the Phantom Liberty expansion. NVIDIA has provided a few outlets early access to the GeForce RTX 4060, allowing them to post benchmark results using Cyberpunk 2077's hardware-intensive 'RT Ultra' setting with DLSS 2 and DLSS 3 enabled.
Over at AusGamers, the only stipulation was that DLSS needed to be set to Balanced, but other than that, there were no restrictions on what to test or compare performance with. And with that, we see in-game results compared to the GeForce RTX 3060 12GB (the STRIX OC model from ASUS) and the GeForce RTX 2060 SUPER (Founders Edition).
3D Mark synthetic benchmarks covering APIs like DirectX 11, DirectX 12, and real-time ray-tracing are often good indicators of what to expect from new GeForce graphics cards from NVIDIA. This week will see the launch of the new mainstream GeForce RTX 4060, with its wallet-friendly price-point of USD 299 and promise of more efficient and better-than GeForce RTX 3060 performance with the DLSS 3 and Frame Generation bonus.
With cards like the GeForce GTX 1060, RTX 2060, and RTX 3060 representing the most popular gaming GPUs out there, the odds are that the upcoming GeForce RTX 4060 will become one of the biggest sellers for the new Ada Lovelace generation. And with that, there's a lot of interest in seeing how it performs.
Outlet Videocardz.com managed to get its hands-on some early benchmark results for the card, covering 3DMark scores from FireStrike (DirectX 11), TimeSpy (DirectX 12), and Port Royal (RT). Taking all of the numbers in, we learn that regarding these specific benchmarks, the GeForce RTX 4060 performs roughly 23% faster than the GeForce RTX 3060 12GB.
AFOX might not be a household name like other graphics card makers, but it's a company that has been in the game for over two decades - so it's a veteran regarding NVIDIA GeForce products. And its latest is very interesting, with the AFOX RTX 4090 (AF4090-24GD6XH4) presenting the flagship Ada GPU from NVIDIA in two-slot form with an old-school blower fan for cooling.
Using a blower fan also means it's shorter, length-wise, than the GeForce RTX 4090 Founders Edition by over 30 centimeters. There aren't many GeForce RTX 4090 models out there that aren't 3.5 slots thick and long, too, so something like the AFOX RTX 4090 will be of interest to those with limited space in their case.
Specs-wise, there's no OC action here; it follows the reference design with a 2230 MHz base clock, a 2520 MHz boost clock, and a 450W power rating delivered over a single 16-pin power connector. This makes sense because blower-style fans for cooling are a relatively old-school way of cooling a GPU. They expel/blow hot air out of the rear of the GPU and are generally noisier than modern cooling solutions.
Earlier this year, we saw pictures of a four-slot Founders Edition for a potential GeForce RTX 4090 TI or TITAN-class Ada GPU that featured an interesting vertical orientation of the I/O ports. This indicated that the internal hardware would be configured vertically and positioned in a certain way to account for the extra cooling required.
Reports at the time indicated that dual 12VHPWR 16-pin connectors would be required to deliver over 800W of power. On top of that, it looks like this GeForce RTX 4090 TI or TITAN-class Ada GPU would also ship with 48GB of VRAM on a 384-bit memory bus.
This week comes new images courtesy of @ExperteVallah over on Twitter showcasing the cooler the rumored "Titan Ada" graphics card. Sitting next to a Founders Edition model from the GeForce RTX 30 Series makes it look absolutely huge, thanks to the quad-slot thickness. The images also reinforce the vertical configuration for the I/O ports, though these images are for a prototype without internal hardware.