NVIDIA’s presentation, prepared for the CES 2023 exhibition, was not opened by company chairman Jensen Huang, but by Jeff Fisher, senior vice president of games. He talked not only about video cards, but also about many smart technologies and their implementation.
NVIDIA RTX 4000 video cards for laptops
The NVIDIA RTX 4000 series of mobile graphics processors, based on the Ada Lovelace architecture, will support the 5th generation Max-Q energy-saving technology. Support includes the use of ultra-low voltage GDDR6 memory and three-speed memory controllers.
The architecture supports leading intelligent graphics processing technologies such as DLSS 3 scaling. The company introduced RTX 4090 and RTX 4080 mobile graphics cards for enthusiasts, as well as mid-range RTX 4070, 4060 and 4050 GPUs.
RTX 4090
The NVIDIA RTX 4090 notebook graphics card is based on the AD103 GPU with 9728 CUDA cores. This configuration is already used by the RTX 4080 desktop graphics card. The card is the successor to the Ampere-based RTX 3080 Ti GPU, with the same maximum TGP of 150 W (not including Dynamic Boost).
NVIDIA confirms that this card will offer 38.9 TFLOPS of single-precision computing performance, which is almost on par with the desktop RTX 4070 Ti introduced today. However, in terms of performance, it is far from the desktop RTX 4090, which is about twice as fast.
RTX 4080
The RTX 4080 graphics adapter is based on the AD104 processor with 7424 CUDA cores and 12 GB of GDDR6 memory. This model will also be tied to a TGP of 150W, but there will also be variants with low consumption – up to 60W (so be sure to read the specifications of the laptop before buying).
The flagship RTX 4080 and 4090 laptop graphics cards are claimed to be powerful enough to play on three 4K monitors at 60 frames per second.
RTX 4070
The mobile variant of the NVIDIA RTX 4070 is based on the AD106 chip with 4608 CUDA cores clocked at up to 2.175 GHz. The card will run between 35W and 115W, the same range for the RTX 4060 and RTX 4050. The 4070 model is equipped with 8GB of GDDR6 memory running over a 128-bit memory bus, just like the RTX 4060.
Younger mobile video cards of the new generation for laptops are built on the AD107 chip. The RTX 4060 will have 3072 CUDA cores, while the RTX 4050 will debut with 2560 cores. The RTX 4050 is equipped with 6 GB of GDDR6 memory and a 96-bit memory bus. The GeForce RTX 4050, 4060, and 4070 are claimed to be faster than the RTX 3080. They should be fast enough for 1440p gaming at up to 80fps, and about twice as fast as the PlayStation 5
The price of laptops with RTX 4000 GPUs will start at $999. The RTX 4070/4060/4050 graphics cards will be available from February 22, while the RTX 4080/4090 series will launch two weeks earlier on February 8. The price of laptops based on them starts at $1999.
NVIDIA GeForce RTX 4070 Ti for desktop PCs
The RTX 4070 Ti video card provides performance up to 40 Tflops. The card is based on the AD104 GPU with 7680 CUDA cores and is equipped with 12 GB of GDDR6X 21 Gbps memory. The 192-bit memory bus limits bandwidth to 504 GB/s (104 GB/s less than the RTX 3070 Ti).
The NVIDIA 4070 Ti will only be released as a partner card, which means no Founder Edition (despite renders of it). Most custom designs are equipped with three fans and occupy 2 to 3.5 slots, which is quite enough for a TDP of 285 W.
Like the previously presented top video cards of the series, the 4070 Ti is manufactured using the TSMC 4N technology process. The AD104 chip is equipped with approximately 35.8 billion transistors, the size of its crystal is about 295 mm². The VideoCards site summarized the characteristics of all presented Ada Lovelace desktop video cards in a visual table:
The card will go on sale on January 5th at a suggested retail price of $799. ASUS, Inno3D, Gigabyte, Colorful and Galax have already presented their own variants of video cards – currently known prices start from the recommended price and go up to $969.
DLSS is not only for games, but also for increasing video resolution
DLSS technology is very important for NVIDIA, Jeff Fisher started the presentation with the news about it. According to him, the next big update to DLSS 3 will be released within the next few months. Speaking about the pace of implementation of the technology, Fisher noted its announcement in 50 games that have been released or are preparing for release.
In the new release, it is planned to improve smoothing, reduce the number of artifacts in the form of halos and improve the stability of the picture in motion. The improvements will work not only with the third version, but also with DLSS 2 (fully or partially?). Games that already work with DLSS 3 will get the update.
Function RTX Video Super Resolution, or DLSS Video will allow you to scale video streams from FullHD to 4K in the appropriate quality using DLSS. Self-learning artificial intelligence will smooth and remove artifacts that occur during upscaling, allowing you to make the image better and clearer on 4K monitors.
Based on the materials of the presentation, the technology will be available in browsers, in particular, Chrome and Edge, and on RTX graphics cards. For the RTX 3000 and 4000 series, it will be launched in February, and later it will work on the RTX 2000. Whether RTX Video Super Resolution will work in other programs and with other resolutions is not yet known.
GeForce NOW – rendering on RTX 4080 and availability in cars
To create the stream of the best quality, GeForce NOW will receive servers based on the RTX 4080 video card. The system will be able to provide performance of 64 Tflops – almost twice as high as on the RTX 3080. The GeForce NOW Ultimate tariff will replace the RTX 3080 plan and will cost $20 per month.
The maximum possible frame rate in streamed games will be 240 frames per second after the update. In 4K resolution, the maximum will be 120 FPS. In addition to G-Sync, the service will receive support for Reflex technology, which reduces monitor signal delay. Support for wide-format screens with a maximum resolution of up to 3840×1600 will also be added. All innovations will appear in American and European services already this month.
GeForce NOW not only improves quality, but also expands its distribution in cars. Access to games will be provided by the NVIDIA Drive platform in the driver’s and front seats when the car is idle or at any time from the rear seats. The initiative is already being implemented by Hyundai, including Kia, BYD and Polestar.
The system will work in OC Android or using a browser. The availability of 1500 games is declared. Support for GeForce NOW cars will expand in the future. You will need a stable 4G or 5G connection to connect and play comfortably on the road
Autonomous driving platform NVIDIA Drive also paid attention to the presentation. It will receive new equipment, new sensors and more, and will also be updated for cars already in production. The system will also support virtual reality with the help of appropriate glasses or helmets.
NVIDIA intelligent technologies
Looking into the camera with NVIDIA Broadcast – Artificial intelligence adjusts the direction of the presenter’s gaze directly at the viewer when he is facing the other way.
Joint work in Omniverse Nucleus Cloud – the Omniverse work environment will allow 3D designers to work together on projects. Its support is already in the process of implementation in Blender, and is expected in other popular programs.
Sanvas 360 – obtaining a spherical map from the captured landscape using NVIDIA Canvas in popular 3D development environments (Blender, Unreal Engine, Unity).
Jensen Huang nevertheless appeared at the presentation, in a not quite usual form.
Intel introduced 35 mobile processors of the 13th generation with the number of cores up to 24 and 16 processors for desktop PCs
Sources: NVIDIA, VideoCardz