GPU

The new HBM-PIM chips will deliver over twice the system performance of the current HBM2 solutions and reduce energy consumption by more than 70%. Samsung designed this new memory architecture in order to accelerate large-scale processing in data centers, high performance computing systems and AI-enabled mobile applications.

News about the RTX 3070 Ti has been quiet in recent weeks, with FinalWire adding support for the card to AIDA64 Extreme in December. Now, a retailer briefly listed an RTX 3070 Ti with 16 GB of GDDR6 VRAM on its website as part of a pre-built PC.

New findings support the idea that the PlayStation 5 does not support AMD Infinity Cache. The first SWIR die shot of the PlayStation 5 points to Sony having cut other RDNA 2 features too, including some FFUs and FMAs or FMADDs.

There have been reports of computer part retailers in Europe hiking prices of GeForce RTX 3060 AIB boards before the graphics card is even launched. Nvidia’s MSRPs for the GeForce RTX 3060 are US$329/€329/£299 but retailer prices have gone beyond double in some cases. Gamers have been blaming the hikes on cryptocurrency miners.

NVIDIA has taken matters into its hands and has intervened on how OEMs advertise its RTX 30 series laptop GPUs. OEMs will now be required to disclose their RTX 30 series laptops' TGPs and clock speeds, but NVIDIA has removed TGP values from its official release drivers.
Updated: A GeForce Hotfix Driver has re-enabled Maximum Graphics Power information for RTX 30 series laptops.

NVIDIA has formally announced that the GeForce RTX 3060 will arrive on February 25, following confirmation from various sources. The GA106 card is slated to launch for US$329 too, despite its apparent limited availability and amid concerns about price increases.

Another AMD board partner has registered Radeon RX 6700 XT SKUs with the EEC in Russia. Powercolor has registered multiple SKUs with the regulatory body, all of which have 12 GB of VRAM. However, there are also a few references to the Radeon RX 6700, which AMD appears to have designed with 6 GB of GDDR6 VRAM.

The U.S. government is ready to conduct a 100-day review that will identify potential chokepoints in the supply chain for semiconductor products, as well as for other essential goods like minerals, medical supplies and high-capacity batteries for electric vehicles. Other shortage solutions include increased semiconductor manufacturing funds and grants.

NVIDIA recently launched a DLSS plugin on the Unreal Marketplace, potentially kicking off widespread adoption of its AI-based upscaling technology. Developers should now be able to effortlessly incorporate DLSS in games that run on the Unreal engine.

Modders discovered that the MSI Leopard G76 vBIOS can be flashed on the Asus ROG Zephyrus Duo 15, allowing the RTX 3080 TGP to be raised from 115 W to 150 W. This trick can provide up to 18% increased performance, but it can also lead to serious stability issues like increased cooler noise, thermal throttling and battery problems caused by overwhelmed power supply.

The GeForce GTX 1050 Ti is being given a new lease of life, almost half a decade after its initial release. According to a new report, NVIDIA has already started resupplying the card to plug the gap currently left by the RTX 30 series.

When it comes to NVIDIA's Ampere laptop GPUs, variation in TGPs is just one conundrum. Apparently, NVIDIA is keeping mum on Dynamic Boost 2.0 clocks due to which OEMs, apart from determining TGP and Max-Q features, are also deciding on which boost clocks to advertise further complicating matters for both reviewers and end users alike.

The NVIDIA GeForce RTX 3060 is already available to purchase through official channels and at more than double its MSRP. NVIDIA's first GA106 card will be available globally, albeit probably in extremely limited quantities, from the end of February.

According to a major European retailer, the NVIDIA GeForce RTX 30 series will become even harder to purchase between now and April. Prices of NVIDIA's latest graphics cards will continue to increase too, even those bought from retailers.

AMD seems to be internally looking at a 2.5x performance increase for RDNA 3 compared to RDNA 2, flagship-to-flagship. RDNA 3 Navi 31 with 160 CUs is expected to be a multi-chip module while Navi 33 will continue to be monolithic. With Intel Xe also pegged to offer good ray tracing performance, NVIDIA Lovelace will have both Intel and AMD in its sights.

A new report from Wccftech speculates that the Nvidia GeForce RTX 3060 will be launched in February 26. Some of its key features include a GA-106-200 GPU, 12GB of GDDR6 VRAM, and 3,584 CUDA cores. Prices are expected to start at US$329.

The Nvidia GeForce RTX 3080 Ti is now speculated to come with 12GB of GDDR6X VRAM and 10,240 CUDA cores. Its 320-bit memory interface will allow for 730Gb/s of memory bandwidth, identical to that of the GeForce RTX 3080.

A financial blogger on Chinese social media has posted various images of Nvidia GeForce RTX 30 series laptops being used for cryptocurrency mining. The rigs on show involve dozens of gaming notebooks stacked up just for hunting down some increasingly valuable Ethereum. Some of the laptops visible are RTX 3060 machines that cost around US$1,000.

The EEC files reveal the existence of several RX 6600 XT models from ASRock, all of them coming with 12 GB of GDDR6 VRAM, This suggests that these versions may still be based on the Navi 22 SKU with 192-bit bus, but probably have low clocks that only require a single-fan cooler.

More details about the AMD Radeon RX 6700 XT have surfaced, courtesy of KOMACHI ENSAKA. This time, Gigabyte has registered multiple RX 6700 XT SKUs with the EEC, seemingly confirming the company's intent to sell AORUS, Eagle, Gaming and OC versions of the card. The listing suggests that Gigabyte plans to sell a reference model too, all of which will have 12 GB of VRAM.

AMD has promised to release more Navi 21-based reference cards throughout Q1 2021, according to a Senior Director of the company. It remains to be seen how many reference cards will hit the market though, but AMD feels that stock difficulties will stay until at least the second half of 2021. Hence, any reference cards sold this quarter will probably be resold almost instantly at a hefty premium.

A Counterpoints Research study has determined that TSMC has 28% of the market share in matured nodes (40nm and below), making it the world leader in the global semiconductor manufacturing market. UMC, SMIC and Samsung take second, third and fourth place with market shares of 13%, 11% and 10%, respectively.

Nvidia looks impressively strong in the PC video card usage chart in Steam’s latest monthly Hardware & Software survey. There are new entries for both the GeForce RTX 3060 Ti and RTX 3090, while the RTX 3080 enjoys a third month of popularity. In terms of overall percentage share, Nvidia has no real rival in the GPU space yet.

A 14C/20T Intel 12th gen Alder Lake-P processor has been spotted on Geekbench showing a boost up to 4.7 GHz. This particular entry pertains to the OpenCL benchmark of the 96-EU Xe GT2 iGPU, and the score is in line with expectations, though there definitely seems to be further scope for improvement.

The RTX 3060 is said to launch soon based on an Ampere chip variant called the GA106-300. A GPU leaks site claims to confirm its existence with a new live image showing this silicon, alongside the latest news on its supposed specs. They are said to include 12GB of RAM and fewer CUDA cores than it conceivably could.

A Gigabyte G5 KC gaming laptop with an Intel Core i7-10870H and Nvidia GeForce RTX 3060 Laptop GPU has been spotted on Geekbench turning out a terrific OpenCL score that would place it roughly between the RTX 2070 and RTX 2070 Super desktop cards in the benchmark chart. Unfortunately, the TGP of the RTX 3060 is not mentioned.

More details about the Radeon RX 6700 XT have appeared online, with the latest leak seemingly confirming several aspects about AMD's upcoming RDNA 2 cards. Not only does the leak provide images of the Radeon RX 6700 branding, but it provides details about the card's VRAM capacity, its marketed gaming resolution and its expected release dates.

Reputed benchmarks of the NVIDIA GeForce RTX 3080 Ti have surfaced, showcasing the upcoming card against the RTX 3080 and RTX 3090. It is unclear why NVIDIA would choose to release the RTX 3080 Ti on these benchmarks, though.

According to an unverified rumour from a South Korean tech forum, AMD could shift some of its GPU/APU production to Samsung. AMD wants to ramp up production, but TSMC is unable to keep up due to prior commitments.

Intel's Raja Koduri showed off a die shot of the 2-tile Ponte Vecchio Xe HPC GPU, which is said to feature seven advanced silicon technologies including the 16 compute units that are fabbed on Intel's own 7 nm process. The other key technologies in Xe HPC include Xe Link I/O, HBM2 VRAM, Rambo Cache, EMIB, 10 nm Enhanced SuperFin, and Foveros 3D packaging technology.

The laptop versions of the NVIDIA GeForce RTX 3060 are yet to land, but the first gaming benchmarks are already online. The two videos compare the GeForce RTX 3060 against the GeForce RTX 2060 in nine games, including in CDPR's latest title, Cyberpunk 2077.

Tesla's upcoming Model S 2021 cars will integrate AMD's Navi 23 GPUs to power their onboard infotainment systems. The leaked block diagram reveals that the Navi 23 will get access to 8 GB of GDDR6 14 Gbps RAM and will be able to deliver 10 TFLOPS of FP32 compute performance, which is close to what a PS5 console is capable of.

Intel has unveiled the Iris Xe desktop graphics card, better known by its codename of DG1. The Iris Xe Graphics has a 30 W TDP, 4 GB of VRAM and will only be available via OEMs. The Iris Xe DG1 will probably compete against NVIDIA's budget cards, such as the Pascal-based GeForce GT 1010.
Updated: VRAM details and Colorful statement added.

The NVIDIA GeForce RTX 30 and AMD Radeon RX 6000 series has already received price hikes following its release, and not just from scalpers reselling the likes of the RTX 3070 for well above MSRP. Now, a report claims that scarcity of GDDR6 and GDDR6X VRAM will push the price of AMD and NVIDIA GPUs up even further.

The DCS feature was discovered in a recent Linux patch and it looks like it is designed specifically for laptop GPUs, even though it is supported on the high-end RX 6800 series, as well as the upcoming RX 6700 mid-range models. DCS reduces power consumption by turning the core off when thermal and power limits are exceeded.

Lenovo's US webstore has begun what it calls a "You Pay What We Pay" event, which apparently entails the same kinds of discounts only normally available to the OEM's staff members. They include up to US$800 off an Intel 10th-gen Legion 7i laptop, as well as deals on gaming mice and headsets.

Intel's Xe DG1 dGPU will only work with select CPU and chipset combinations and will require a BIOS update in order to be recognized. Intel said that Coffee Lake-S and Comet Lake-S CPUs are supported on specific midrange and budget 300 and 400-series chipsets. This rules out any possibility of the GPU being compatible with AMD systems.

The ridiculous number would not really be a problem for end users if OEMs actually respected Nvidia's guidelines and provided clear info with the official specs. Unfortunately, very few OEMs are currently doing this, and we are seeing wild spec variations that allow lower-priced SKUs to outperform higher-priced variants.

AMD's support for USB4 and Thunderbolt is scheduled to arrive in early 2022, and, until then, laptop OEMs are trying out alternatives. One solution would be to implement a new type of connector, like Asus has done on its new ROG Flow X13 laptops. Lenovo, on the other hand, is exploring something more simple in the form of a cable that connects external video cards to the M.2 slot inside the laptop.

The new RTX 3080 cards proved to be quite the crypto mining monsters, but these high-end solutions are still in very short supply. A better and more efficient alternative could be presented by Nvidia's mobile RTX 3000 dGPUs, as suggested by the latest leaks from China. If this is indeed true, miners could eventually impact the availability of all upcoming Nvidia-based laptops.

ITHome reports that a Korean web community shared benchmark results for an under-developed AMD RDNA2-powered Exynos chip. The next-gen flagship appears to nearly double the A14's performance in the Aztec Ruins benchmark, while also exceeding it in Manhattan.

According to new rumours, AMD will be upping its GPU game massively with its next-generation cards. Apparently, Navi 31 will be the first chiplet-based GPU and one that delivers up to 160 CUs, or 10,240 stream processors. The RDNA 3 GPU may offer around 37 TFLOPs of FP32 performance, too.

Intel may launch the 12th gen Alder Lake-S family as early as September this year alongside 600-series chipset-based motherboards, just six months after Rocket Lake-S becomes available. Alder Lake-S will feature a similar Xe LP iGPU as Tiger Lake and Rocket Lake-S but now, we may also get to see a GT0.5 variant.

Fifty gaming laptops featuring Nvidia GeForce RTX 30 series GPUs from major vendors such as Asus, MSI, and Gigabyte have been listed on the Scan website. Many feature Ryzen 5000 APUs and there are a couple of Tiger Lake H-35 options too. Prices go from £998.99 (US$1,367) for an Asus TUF Dash RTX 3060 laptop up to £4,298.99 (US$5,882) for an MSI GS66 Stealth RTX 3080 laptop.

YouTube VIK-on demonstrated a way to replace the Micron 8 GB GDDR6 modules on a Palit GeForce RTX 2070 card with Samsung 16 GB ones thereby confirming that NVIDIA did have plans to launch a 16 GB RTX 2070 back then. The experiment resulted in the PC booting up, but the GPU crashed under load and posted low 3DMark Time Spy Graphics scores.

The GeForce RTX 3090 is the most power-hungry single-GPU card to date. This, however, didn't deter EK from releasing a Quantum Vector water block that promises lower temps and (possibly) higher overclocks.