GPU

The GDC trailer shows loading times of only 1 second for NVMe SSD solutions using the DirectStorage API, yet the difference compared to the default Windows API is revealed to be minimal in a separate slide.

AMD admits that switching from spatial to temporal upscaling can be a bit more taxing on the GPUs, but the new FSR 2.0 upscaler should work even on Nvidia's GTX 16-series with good results. Team Red also reveals that FSR 2.0 implementation is fastest on games that already include DLSS 2.0 support.

NVIDIA may not have released the RTX 3090 Ti yet, but it has showcased six new workstation and studio GPUs. Ranging from the RTX A500 to the RTX A5500, all the new GPU options rely on NVIDIA's Ampere architecture.

The H100 specs confirm that the upcoming RTX 4000-series will not feature a chiplet GPU design and the maximum TGP is considerably lower than the rumored 850 W. Even so, the sheer size of the die is supposed to provide enough raw performance to easily compete with AMD's upcoming RDNA3 models.

The H100 Hopper is Nvidia's first GPU with PCIe 5.0 specifications and support for HBM3 VRAM. This model is up to 6 times faster in FP workloads and offers considerably higher scalability compared to previous gen Ampere models. Nvidia is also planning to release a Hopper GPU + Grace CPU superchip and even a dual Grace superchip.

The GeForce RTX 3090 Ti has appeared sporting some shocking price tags in a leaked retailer listing. Two board variants were spotted, with both of them coming from Asus. AIB modifications will add to the price tag, and it is not known whether Nvidia will actually publish an MSRP for the long-awaited RTX 3090 Ti.

The Apple M1 Ultra has been absolutely flattened by the Nvidia GeForce RTX 3090 in a synthetic benchmark and gaming comparison. This particular M1 Ultra, in a brand-new Apple Mac Studio, was the 20-core CPU/64-core GPU model, which is as high-end as it gets. But the RTX 3090 was able to score over +158% higher than the Apple silicon in one synthetic benchmark.

FSR 2.0 offers temporal upscaling just like Nvidia's DLSS 2.0, but does not require neural network training so the technology remains open source and compatible with all GPUs on the market. As far as image quality is concerned, FSR 2.0 can be better than native resolution in some cases, but Nvidia's DLSS 2.0 still appears better overall. FSR 2.0 can also boost fps counts significantly (up to 2X compared to native resolution in performance mode).

GPU prices on eBay have dropped almost 9% in the first two weeks of March 2022 according to Tom's Hardware. Compared to December 2021, the prices have come down by an average of 25%. Both the current-gen and the last-gen GPUs experienced this price reduction.

A new rumor about the Navi 33 GPU from AMD claims that the next-gen entry-level graphics processor will be able to match, and even outperform, the flagship Navi 21 chip from the Radeon RX 6000 series. In addition, the Radeon RX 7000 series will likely require much lower power draws than the GeForce RTX 40 lineup for high-performance usage.

NVMe storage has been around for 11 years now, but its raw performance is barely utilized in applications today. Microsoft intends to change this by introducing DirectStorage - a technology that is already doing wonders on the latest Xbox consoles. The first DirectStorage API version is now available for game developers and we should see the first compatible games in the near future.

The Total Graphics Power (TGP) figure for the Nvidia GeForce RTX 4090 has supposedly been confirmed. Apparently, it stands at a toasty 600 W, considerably higher than its Ampere counterpart. In addition, an astonishing figure of 800 W or more for the potential GeForce RTX 4090 Ti has also been suggested.

The latest results of AMD's research in the domain of next-generation image upscaling technology for games will be presented at this year's GDC event. Is Team Red including dedicated hardware for image upscaling functions in the upcoming RDNA3 GPUs?

YouTuber Moore's Law is Dead has leaked a trove of new information about the Nvidia AD102 GPU that will reportedly be featured in the RTX 4090/RTX 4080. The information includes alleged TDP numbers and GDDR memory spec for the upcoming chip. The YouTuber also confirmed incredible performance numbers for Nvidia's next-gen lineup.

Embargo dates for the GeForce RTX 3090 Ti have leaked, a graphics card that NVIDIA teased in January at CES 2022. Unlike recent GeForce RTX 30 series releases, the RTX 3090 Ti's announcement, review and on-shelf embargo all end simultaneously. As such, the RTX 3090 Ti will be orderable right after NVIDIA announces the card later this month.

AMD has announced a new mid-range workstation MPX GPU for the Mac Pro. The new AMD Radeon Pro W6600X offers 32 CUs, 8 GB GDDR6 128-bit memory and up to 9.8 TF peak FP32 performance. Buyers configuring the Mac Pro can now opt for the Radeon Pro W6600X MPX card for US$700.

Average prices for AMD Radeon RX 6000 series cards and boards from the Nvidia GeForce RTX 30 range have hit a 12-month low according to a new report. The data was taken from numerous German and Austrian retailers and could indicate that RX 6000 and RTX 30 series graphics cards may soon hit MSRP levels.

A mystery laptop running an AMD Radeon RX 6850M XT and an Intel Core i9-12900H has shown up in the Ashes of the Singularity database. Under the "Min 1080p" preset, the graphics card manages to output an average of 109 FPS and scores 10,800 overall, just a smidgen below the Nvidia GeForce RTX 3070 Ti's score.

A new leak suggests that Nvidia will launch the GeForce RTX 3090 Ti on March 29 after ironing out problems with its memory. Nvidia has also cancelled the GeForce RTX 3070 Ti 16 GB variant outright.

The Dimensity 1300 is MediaTek's new processor that confers "everyday 5G" on the devices to run on it. It is an eight-core chipset that is compatible with cutting-edge 200MP camera sensors. It supports Wi-Fi 6, Bluetooth 5.2 and a potentially high display refresh rate.

This information appears to be coming via the hacked confidential data stolen from Nvidia a few days ago. The images show driver code mentioning six Ada Lovelace SKUs, two Hopper SKUs and two more Blackwell SKUs. Judging by the number of SKUs, Ada Lovelace is clearly Ampere's successor, and we already knew that Hopper is targeting data centers, but it is still unclear what Blackwell is bringing to the table.

Benchmarks from an Asus ROG Flow X13 featuring the AMD Ryzen 7 6800HS have leaked online. Compared to the Core i9-12900H in the ROG Flow Z13, the Ryzen 7 6800HS trails considerably. However, the 16 GB of LPDDR5-6400 RAM does help the Radeon 680M in offering a significant advantage over the GeForce MX450 bringing it closer to a GTX 1650 Max-Q.

The Steam Deck has been officially released and just in time for the launch of Elden Ring. Gamers interested in buying both Valve’s handheld console and FromSoftware’s gorgeous RPG with the intent of mobile monster-slaying will be pleased to learn Elden Ring has been classified as “Deck verified”.

The latest version for AMD's System Management Unit driver is suggesting that the RDNA2 iGPUs found in the upcoming Ryzen 7000 desktop processors could run at 1.1 GHz, and, according to industry insiders, there could be a lower-end model with 4 CUs that only has a third of the Steam Deck's RDNA2 performance.

The Nvidia GeForce RTX 2050 manages to score 9,673 on the 3D Mark Fire Strike benchmark. In terms of performance, it sits between the Turing-based GeForce GTX 1650 and GTX 1660 Ti and could match the Ampere-based GeForce MX570, as they both share the same GA107 GPU.

Zotac has officially listed three Nvidia GeForce RTX 3060 Ti desktop graphics card variants featuring a GA103 GPU. Other than the GPU swap, it seems to have identical specs to its GA104-toting sibling, including 8 GB of GDDR6 VRAM and 4,864 CUDA cores.

(Update: Malware warning) A public beta of the nifty little piece of software for cryptominers will be released in the coming few days, and it could possibly impact the prices of all Nvidia's RTX GPUs with the infamous hashrate limiter.

China is reacting to new outbreaks of the Omicron variant of the Coronavirus with partial lockdowns. This could further delay the availability of laptops with 12th Gen Intel Core processors and Nvidia's Ti graphics cards, which debuted at the beginning of the year. The first factories have already been closed in Suzhou in the east of the country. Supply chain and logistics bottlenecks, a shortage of certain chip types and price increases are already on the horizon.

Multiple industry insiders and Twitter users have speculated that TSMC has run into several snags with its upcoming N3 process node. Hence, AMD may have to fabricate its Ryzen 8000 Zen 5 processors and Radeon RX 8000 RDNA 4 graphics cards on TSMC's 4 nm process.

Intel has finally commented on how many graphics cards it will ship this year. Supposedly to counteract the GPU shortage, Intel has also announced where its GPU business is headed under Project Endgame.

Lenovo's Vantage software is included from the factory in most of the OEM's laptops nowadays. While it provides the usual warranty and diagnostic information, here are some extra-neat features that may prevent you from uninstalling it immediately.

AMD's upcoming Ryzen 6000 series of processors were benchmarked across various games and synthetic benchmarks. Both the Ryzen 5 6600H and Ryzen 6800H easily trounced their 12th-generation Intel Alder Lake counterparts. The former uses an RDNA2-based Radeon 680M iGPU with 12 CUs, and the latter cuts that down to 6 CUs.

Generational performance improvement estimates for the upcoming Nvidia Lovelace family and AMD RDNA3 next-gen graphics cards have seemingly left the rivals looking evenly matched. Figures were compared to the performance of the GeForce RTX 3090. A noted tipster has also signaled that GPU prices will fall “dramatically” in March.

Mobile workstations are primarily supposed to be reliable companions for professional users and offer a lot of performance at the same time. The ZBook Fury 17 G8 basically meets these criteria, but we think HP wastes a lot of potential in terms of GPU performance.

While the i7-12700H + RTX 3050 Ti combo may seem questionable, the US$1,099 price point is undoubtedly enticing. Colorful's latest X15 XS models also come equipped with 16 GB of DDR4-3200 RAM and a 512 GB NVMe SSD, plus there is an i5-12500H model for US$999, as well.

According to insider sources, the RX 6x50 XT refresh would include RX 6950 XT, RX 6750 XT and RX 6650 XT SKUs, but no RX 6850 XT. The new GPUs are rumored to get upgraded with 18 Gbps GDDR6 VRAM and the high-end cards would also see a TGP bump.

YouTuber Erdi Özüağ who has access to both the Exynos 2200 and Snapdragon 8 Gen 1 variants of the Samsung Galaxy S22 Ultra showed that the Xclipse 920 RDNA 2 GPU in the Exynos 2200 SoC puts up an underwhelming performance in benchmarks compared to the Adreno 730 in the Snapdragon 8 Gen 1. While a future OTA update may potentially reduce the performance deltas, the possibility that Samsung's 4 nm process could be having an impact on the Xclipse 920's performance cannot be ruled out.

Nvidia has officially declared that it has no new information about the GeForce RTX 300 Ti in a statement to The Verge. Initially slated for a January 2022 launch, the graphics card got pushed back due to hardware issues, presumably because of its high power requirements.

GPU prices are falling drastically and availability is the best it has ever been since January 2021 according to a report by 3DCenter. The current fall in the prices can be attributed to the recent cryptocurrency crash that has seen the price of ETH go down as low as US$2,242.

Igor's Lab reports that Nvidia is working on a GeForce RTX 3050 version. It will be powered by a GA107 GPU and draw 115 W of power. The graphics card could be launched alongside the GeForce RTX 3070 Ti 16 GB variant in the coming weeks.

The next generation of GPUs coming from both AMD and Nvidia have been broken down into subsequent chips, models, and manufacturing process by a well-known leaker. The breakdown is a best guess scenario, with one of the tipster’s followers offering suggestions of relative VRAM amounts.

Price predictions for both the Nvidia GeForce RTX 4090 and the AMD Radeon RX 7900 XT have been offered up by a YouTuber who focuses their channel mostly on graphics cards. Estimates for the RTX 4090 and RX 7900 XT come in at just under US$2,000, with the AMD Navi 31-based card expected to outperform the Lovelace AD102 GPU-powered rival.

An EEC listing for an unknown Nvidia GPU has shown up with the model number PG136C. The entry is rumored to be for the upcoming RTX 3090 Ti. The GeForce RTX 3090 Ti was previously reported to have been delayed indefinitely.

As a part of its supposed anti-bot measures, Best Buy allowed only users with a US$199 Totaltech subscription a chance at buying Nvidia RTX 3000 series graphics cards. However, there wasn't a limit on the number of purchases per account, which effectively allowed scalpers/bots to walk away with a good chunk of the GPUs.

You can now grab the NVIDIA T1000 graphics card aimed at workstation professionals for less than $400 in the U.S. The T1000 specs include four DisplayPort 1.4 outputs and support for up to five 5K or two 8K@60Hz monitors, a very slim design with one active fansink, as well as low 50W power consumption.

YouTuber Moore's Law is Dead has heard from multiple sources that the Intel Arc Alchemist graphics cards will not be launched in Q1, 2022. Intel is expected to shed more light on the lineup at PAX East, and low-end mobile SKUs could be launched first.

Nvidia's upcoming GH100 Hopper HPC accelerator has aroused quite the interest owing to recent leaks. First, it was the rumored massive die area close to 1000 mm square and now, it is the massive 140 billion transistor count. It is not yet clear how Nvidia would achieve such a high transistor density on the 5 nm package, but it is likely that Hopper will include both monolithic and multi-chip module (MCM) architectures and cater independently to HPC and DL workloads.

An AMD engineer has confirmed several details about the Radeon RX 7000 series of GPUs. As expected, the Radeon RX 7000 series will arrive as 3 GPUs, the Navi 31, Navi 32 and Navi 33. Only Navi 31 and Navi 32 will utilise Multi-Chip-Modules (MCM), too.